<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Amin Amjadi on Medium]]></title>
        <description><![CDATA[Stories by Amin Amjadi on Medium]]></description>
        <link>https://medium.com/@maamjadi?source=rss-88281f25dc2b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*tTm0nsJRdm4lHNGP9ww6bg.jpeg</url>
            <title>Stories by Amin Amjadi on Medium</title>
            <link>https://medium.com/@maamjadi?source=rss-88281f25dc2b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 11:43:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@maamjadi/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Trade Show Lead Capture: The Booth Playbook]]></title>
            <link>https://medium.com/gravisun/trade-show-lead-capture-the-booth-playbook-1c231960e0e2?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c231960e0e2</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Fri, 24 Jul 2026 15:14:00 GMT</pubDate>
            <atom:updated>2026-07-28T06:03:22.042Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Z9EmQ8zgEVI2twEI7EdycA.png" /></figure><p><strong>A trade show booth is the most expensive lead source you own — and the leakiest.</strong> You pay for the floor space, the build, the shipping, the flights, the hotel, the swag, and three days of your best people being off their desks. Then you watch a meaningful share of the leads evaporate, not because the conversations were bad, but because nobody had a system for the 90 seconds <em>after</em> each handshake.</p><p>Here’s the uncomfortable part: the booth is usually fine. The traffic is fine. The pitch is fine. What fails is the transfer — getting a real human conversation into a durable, owned, followed-up record before the next person walks up. That transfer is a process problem, and process problems are fixable.</p><p>Booth lead capture isn’t won at the booth. It’s decided in the prep, executed in a tight on-floor flow, and cashed in the 24 hours after the doors close. Here’s the full playbook.</p><h3>What does a booth lead actually cost you?</h3><p>Work it out once and you’ll never treat capture as an afterthought again.</p><p>Take your <strong>all-in</strong> show cost — not just the space. Booth space, stand design and build, shipping and drayage, electricity and wifi, flights, hotels, per diems, printed collateral, giveaways, and the loaded salary cost of every person on the floor for the days they’re there (plus travel days). That’s your real number, and it’s usually two to three times what people quote when they say “the booth cost us X.”</p><p>Now divide it by the number of conversations you actually expect to have. As a purely illustrative example: a show that costs <strong>$30,000</strong> all-in and produces <strong>300 real conversations</strong> puts you at <strong>$100 per conversation</strong> — before a single follow-up email is written. If a third of those conversations never become a retrievable record, you didn’t lose 100 leads. You spent $10,000 on nothing, and you have no way to know it happened.</p><p>That’s the frame worth holding for the rest of this post: at a trade show, <strong>an uncaptured conversation is the single most expensive item on the invoice.</strong> It costs exactly as much as a captured one and returns nothing. Every recommendation below is aimed at that one number.</p><h3>Why do trade show leads leak?</h3><p>Booth leaks are specific and predictable. Nearly all of them fall into six buckets:</p><ol><li><strong>The stockpile.</strong> Cards go into a pouch, a pocket, a fishbowl, or the drawer under the counter, to be “sorted out Monday.” By Monday the stack is out of order, the context is gone, and a few cards are missing entirely. This is the <a href="https://lynqu.com/blog/stop-losing-leads-at-capture">capture gap</a>, and at a booth it runs faster than anywhere else — because the next conversation is already starting.</li><li><strong>One device, six reps.</strong> The show’s rented lead-retrieval scanner lives on a lanyard around one person’s neck. When they’re mid-demo, everyone else defaults to paper. Capture quietly becomes optional for five out of six people.</li><li><strong>The organizer’s data dump.</strong> Many rented scanners hand you a CSV at the end of the show: name, company, email, badge type. No note, no intent, no owner, no memory of who said what. It’s a mailing list, not a set of leads — and it arrives <em>after</em> the moment when it could have been enriched with context.</li><li><strong>Context evaporation.</strong> A record with no note is only half a lead. Three days and 200 conversations later, “Sarah, ops director” is indistinguishable from every other Sarah, and the follow-up defaults to a generic blast — the exact email that gets ignored.</li><li><strong>No owner.</strong> The list comes back and nobody’s name is on any row. Two reps email the same promising contact, forty others hear from nobody, and the argument about territory happens <em>after</em> the leads have already gone cold.</li><li><strong>The Monday cliff.</strong> Even a perfectly captured list rots if it sits until the team is “back and caught up.” Day-one leads are four days old by then, and the competitor who emailed on Tuesday night has already booked the call.</li></ol><p>Notice what these have in common: none of them are motivation problems. Nobody chooses to lose a lead. Each one is a <em>friction</em> problem at the exact moment attention is scarcest — which means each one is solved by a decision made before you ever got on the plane.</p><h3>Before the show: the decisions that make capture automatic</h3><p>This is the stage every guide skips and every winning team obsesses over. Walk onto the floor with all of these already settled.</p><ul><li><strong>One capture flow, briefed to everyone.</strong> Agree on a single motion — <strong>capture → tag → one-line note</strong> — and make sure every person working the stand can perform it identically, including the sales engineer who’s only there Thursday and the founder who “just wants to say hi to a few people.” One flow means no lead is lost to “I thought someone else got them.”</li><li><strong>The fields you’ll capture, and no more.</strong> Name, company, email, and one intent tag is enough. Every extra field you demand at a booth is friction paid 300 times. You are not building a complete customer record on the floor; you are building a record good enough to have an intelligent conversation on Tuesday.</li><li><strong>A tag list that’s about the <em>next action</em>, not the vibe.</strong> “Hot / warm / cold” feels informative and isn’t — one rep’s hot is another’s warm, and neither tells Monday-you what to do. Use tags that name the next step: demo, pricing, send-info, partner, not-a-fit. Now the follow-up sorts itself, and not-a-fit keeps your list honest instead of inflating the headline number.</li><li><strong>Ownership and routing, decided in advance.</strong> Who owns which leads — by territory, product line, or simply who captured them? Deciding <em>after</em> the show is how leads sit in no-man’s-land for a week. Capture-based attribution (the rep who captured the lead keeps the credit) is the simplest rule that works, and it has the useful side effect of making capture personally worth doing.</li><li><strong>Your digital card ready to share.</strong> A <a href="https://lynqu.com/blog/how-to-use-qr-codes-to-share-your-contact-details-instantly">QR code or tap that opens your card in any browser</a> makes the exchange two-way and instant, with nothing for the visitor to install. Put the QR on the counter, on the stand graphics, and on the back of every badge lanyard your team wears.</li><li><strong>A capture path that survives a bad network.</strong> Convention center wifi is famously terrible and hall cellular is worse when 20,000 people are on it. Test your capture method <em>in the hall</em> during setup, not in the hotel lobby. Know what your fallback is before you need it — and if a rep does fall back to paper, they scan those cards during the next lull, not next week.</li><li><strong>Follow-up templates written before you fly.</strong> Draft two or three now — one for each major intent tag — so the post-show job is <em>personalizing</em> a message, not writing from a blank page at 11pm. This single hour of prep is the difference between following up on Tuesday and following up “soon.”</li><li><strong>The event set as the lead source, and the CRM connection tested.</strong> Tag every lead with the show name at the moment of capture. This is the only thing that makes <a href="https://lynqu.com/blog/networking-roi-measurement">event ROI measurable</a> afterwards, and it takes about a minute to set up beforehand — versus an unwinnable archaeology project later.</li><li><strong>A consent line everyone can say in one breath.</strong> If you’re scanning badges or cards, be straightforward about what happens next: <em>“I’ll send you the deck and follow up by email — okay?”</em> It’s good manners, it’s good practice under GDPR-style regimes, and it doubles as a soft qualifier — anyone who hesitates was never going to reply anyway. Note what they agreed to, and honor it.</li><li><strong>Roles and shifts, on paper.</strong> Who’s on the floor in each slot, who’s demoing, and who’s the “floater” watching capture quality and relieving anyone who’s stuck. Booth fatigue is real by hour six of day two, and capture discipline is always the first thing to go.</li></ul><p>Prep is the cheapest leverage in the entire event: an hour of decisions that protects a five-figure spend.</p><h3>During the show: a 30-second capture flow that scales</h3><p>On the floor you’re capturing volume under time pressure, so optimize for speed and consistency, not completeness. The whole flow should fit in the gap between one conversation ending and the next beginning.</p><p><strong>The beat-by-beat:</strong></p><ol><li><strong>Capture on the spot</strong> — scan their badge or business card, or share your digital card so their details save to you. Right then. Not into a pile, not “after this rush.” Later is where leads die, and at a booth “later” arrives within about twenty seconds.</li><li><strong>Tag the intent</strong> while the conversation is still in your head. One tap: demo, pricing, send-info, not-a-fit.</li><li><strong>Add a one-line note.</strong> Not minutes — one line. <em>“Rolling out 40 seats in Q3, hates their current tool.”</em> This single sentence is what makes your follow-up read like a continuation of a conversation instead of a mail merge, and it’s worth more than every other field combined.</li><li><strong>Move on.</strong> Total elapsed time: under thirty seconds, done standing up.</li></ol><p><strong>Two questions that qualify without killing the conversation.</strong> You are not running discovery at a booth — you’re triaging. Two questions do almost all the work:</p><ul><li><em>“What brought you to the show?”</em> (tells you the problem they’re actually shopping for)</li><li><em>“What are you using for this today?”</em> (tells you fit, urgency, and who you’re up against)</li></ul><p>Anything more belongs on the follow-up call. The most expensive mistake on a show floor is spending twenty-five minutes with one polite visitor while eight qualified buyers walk past your booth — trade show time is a fixed, perishable, brutally scarce resource, and a friendly non-buyer will consume all of it if you let them.</p><p><strong>Run the floor, not just the booth:</strong></p><ul><li><strong>Everyone captures.</strong> If only one person has the scanner, capture becomes that person’s job — and it stops happening when they’re busy. Every rep should be able to capture from their own phone.</li><li><strong>Keep a shared count visible.</strong> A running “leads captured today” number keeps energy up during the dead hours and shows the manager who’s swamped and who needs a hand. Visible progress reliably produces more progress — and it surfaces the rep who’s had forty conversations and captured six.</li><li><strong>Triage at the end of each day, not the end of the show.</strong> Fifteen minutes over dinner: skim the day’s captures, fix anything missing a note, and send the same-day follow-ups to the handful of genuinely hot leads. Day-one leads should never wait until day three.</li></ul><p>If you do this right, there is no “data entry” phase after the show. By the time you’re breaking down the stand, every conversation already exists as a structured, tagged, owned lead.</p><h3>After the show: the 24-hour conversion window</h3><p>Booth leads decay faster than almost any other kind, because your visitor had fifty conversations too. The rules are simple, and speed dominates all of them — <a href="https://lynqu.com/blog/fastest-way-to-follow-up-with-sales-leads">speed of response is one of the strongest levers you control</a>.</p><p><strong>Tier the follow-up instead of blasting the list:</strong></p><ul><li><strong>Tier A — genuinely hot (</strong><strong>demo, </strong><strong>pricing).</strong> Same day where possible, within 24 hours at the latest, written individually. Reference the specific thing you discussed, propose a concrete time, and attach only what they actually asked for. These are a small number of emails; treat them like the five-figure opportunities they are.</li><li><strong>Tier B — real but early (</strong><strong>send-info, </strong><strong>partner).</strong> Within 24-48 hours. Use the pre-written template, but personalize the first line with the note you captured. One specific sentence rescues an otherwise-templated email.</li><li><strong>Tier C — </strong><strong>not-a-fit.</strong> Send a short, gracious note or nothing at all, and keep them out of the sales cadence. Protecting the list from noise is what keeps everyone trusting the tier system next time.</li></ul><p><strong>Then close the loop properly:</strong></p><ul><li><strong>Every lead lands in the pipeline</strong> at a defined stage, with exactly one owner. A lead with no owner and no stage isn’t a lead; it’s a note.</li><li><strong>Set the next step at the same moment as the first touch</strong> — a task, a reminder, a scheduled second email. Follow-up that depends on memory competes with the 400 unread emails that piled up while you were away, and loses.</li><li><strong>Run a 30-minute retro within the week</strong>, while it’s fresh: what converted, which tags were useless, which staff slots were dead, whether the booth position earned its price. Write it down for next year’s show. Nobody remembers this in eleven months.</li></ul><h3>What should you measure?</h3><p>Track five numbers, and track them per show so they compound into something you can plan with:</p><ol><li><strong>Conversations vs. captures.</strong> The gap between them is your leak, and it’s the only one of these numbers most teams never look at.</li><li><strong>Follow-up rate within 24 hours.</strong> The single best predictor of whether the show pays off.</li><li><strong>Cost per captured lead</strong> — all-in show cost ÷ leads captured.</li><li><strong>Sourced pipeline</strong> — opportunity value tagged to the show, which is why the source tag has to be set at capture.</li><li><strong>Cost per opportunity and closed revenue</strong> — the number your CFO actually wants, and the one that decides whether you rebook the stand.</li></ol><p>Tag every lead with the event at the moment of capture and all five fall out of the data automatically. Skip that one step and you’ll spend February reconstructing October from memory. If you want to sanity-check the spend before you commit to next year’s floor space, run the numbers through the <a href="https://lynqu.com/tools/roi-calculator">ROI calculator</a> and read the <a href="https://lynqu.com/blog/networking-roi-measurement">full measurement framework</a>.</p><h3>The mistakes that cost the most</h3><ul><li><strong>Staff on their phones</strong> — the universal signal that the booth is closed. Visitors will not interrupt you.</li><li><strong>Sitting down.</strong> Nobody approaches a seated stranger behind a counter.</li><li><strong>Doing discovery on the floor.</strong> Triage, then book the call.</li><li><strong>One scanner for the whole team.</strong> Capture stops being everyone’s job the moment it’s technically one person’s.</li><li><strong>Unbriefed helpers.</strong> The engineer or exec who’s there for a day captures nothing because nobody showed them the flow in the ninety seconds it takes.</li><li><strong>Collecting business cards into a bowl for a prize draw and calling it lead capture.</strong> It’s a list of people who wanted an iPad.</li><li><strong>Waiting for Monday.</strong> Four days of decay applied to your most expensive leads of the year.</li></ul><h3>A quick example</h3><p>Picture a three-person team working a 200-booth expo. Foot traffic is relentless. On the old plan, each rep collects cards into a lanyard pouch and plans to “sort them out Monday.” By Monday the stacks are mixed together, half the context is gone, three cards never made it home, and two reps email the same promising lead while a hundred others hear nothing at all.</p><p>Now run the playbook. Before the show, the team agrees on the capture → tag → note flow, settles ownership by territory, writes two follow-up templates, tests capture in the hall during setup, and agrees on a one-sentence consent line. On the floor, each conversation is captured the moment it ends, tagged by next action, and given a one-line note — from all three phones, not one shared scanner. A shared counter ticks past 180 by the end of day two, and the manager notices one rep is drowning and reroutes traffic. Each evening, fifteen minutes over dinner clears the day’s hot leads with individual emails.</p><p>Monday morning there is no shoebox. Two hundred-plus leads are already in the pipeline — owned, tagged, and largely already followed up — and the team is booking calls while their competitors are still transcribing cards.</p><p>Same booth. Same budget. Same people. The difference is a system that existed <em>before</em> the doors opened.</p><p><em>(Illustrative example for clarity, not a specific customer.)</em></p><h3>The pre-show checklist</h3><p>Print this and work down it in the week before you fly:</p><ol><li>One capture flow agreed and demonstrated to <strong>every</strong> person who will stand in the booth, including part-timers.</li><li>Fields fixed at name, company, email, one tag — nothing more.</li><li>Tag list defined by next action (demo, pricing, send-info, partner, not-a-fit).</li><li>Ownership and routing rules written down.</li><li>Digital card live, QR code printed for the counter and the stand graphics.</li><li>Capture tested <strong>in the hall</strong> during setup; fallback agreed.</li><li>Two or three follow-up templates written and saved.</li><li>Event set as the lead source; CRM connection tested with a dummy lead.</li><li>Consent sentence agreed and practiced.</li><li>Shift rota set, with a designated floater watching capture quality.</li></ol><p>Ten decisions. Perhaps an hour. It is the highest-return hour of the entire event.</p><h3>Where a tool fits — and an honest caveat</h3><p>A lead capture platform like is built for exactly this shape of problem: <a href="https://lynqu.com/solutions/smart-scan">AI scanning</a> for badges and paper cards straight from any rep’s phone, intent tags and notes captured in the same motion, a shared pipeline where every lead has an owner and a stage, capture-based attribution so the rep who did the work keeps the credit, a live team view of what’s been captured today, and a push into your CRM — Salesforce, HubSpot, Zoho, Pipedrive, or Dynamics 365 — so nothing needs re-typing. It turns the leaky ninety seconds after each handshake into a captured, owned lead. You can <a href="https://lynqu.com/business/lead-capture">start free</a> and run it at your next show, and if you’re comparing options, we keep an <a href="https://lynqu.com/compare/best-event-lead-capture-apps">honest comparison of event lead capture apps</a>.</p><p>Honest caveat: no tool will fix a booth in a dead aisle, a stand nobody stops at, or a pitch that doesn’t land. Those are real problems and they’re upstream of anything software can touch. What a tool fixes is the leak — making sure the leads you <em>did</em> earn, with the money you already spent, don’t quietly disappear between the handshake and Monday. That’s the part most exhibitors lose, and it’s rarely the part they think they’re losing.</p><h3>FAQ</h3><p><strong>What’s the best way to capture leads at a trade show?</strong> Capture each visitor the moment the conversation ends — scan their badge or card, tag their intent, and add a one-line note — using a single capture flow your whole team agreed on before the show. On-the-spot capture from every rep’s own phone beats stockpiling cards to process later, which usually never happens, and it preserves the context that makes follow-up work.</p><p><strong>How do you prepare for trade show lead capture?</strong> Settle ten things before you fly: one capture flow, the two or three fields everyone collects, a tag list based on next actions, ownership and routing rules, your digital card and QR code, a capture method tested in the hall, pre-written follow-up templates, the event set as the lead source in your CRM, a one-sentence consent line, and a shift rota with a floater watching capture quality.</p><p><strong>How soon should you follow up with trade show leads?</strong> Within 24 hours, and same-day for anyone who asked about a demo or pricing. Your visitor had dozens of conversations too, so you’re competing with fading memory as much as with other vendors. Triage each evening rather than waiting for the end of the show — day-one leads shouldn’t sit until day three.</p><p><strong>How do you qualify leads at a booth without wasting time?</strong> Ask two questions — “What brought you to the show?” and “What are you using for this today?” — then tag and move on. Booth time is fixed and perishable; deep discovery belongs on the follow-up call. The costliest floor mistake is spending twenty-five minutes with one polite visitor while qualified buyers walk past.</p><p><strong>Do we need the show’s rented badge scanner?</strong> Often not, and it’s worth checking what you’d actually get. Rented lead-retrieval devices typically produce a name-and-email CSV with no notes, no intent, and no owner — a mailing list rather than a set of leads — and they concentrate capture on whoever is holding the device. Scanning paper cards or exchanging digital cards from every rep’s phone captures context at the same time and works for the visitors who never scan a badge. Some shows do restrict badge data to their own device, so confirm with the organizer before you build your plan around it.</p><p><strong>Do visitors need an app to share their details with us?</strong> No. A digital business card opens in any browser from a QR code or a tap, and card and badge scanning happen on your side. The visitor installs nothing — which matters, because anything requiring a download will be declined at a busy booth.</p><p><strong>What about consent and GDPR when scanning badges?</strong> Say plainly what happens next -<em>”I’ll send you the deck and follow up by email — okay?”</em> — record what they agreed to, and honor it. Beyond being required in GDPR-style regimes, it’s a useful filter: anyone who hesitates was unlikely to reply. Check your show’s own terms too, since badge data usually comes with organizer conditions attached.</p><p><strong>How many people should work the booth?</strong> Enough that no visitor waits and nobody works more than a few hours without a break — capture discipline is the first thing that slips when staff are tired. Whatever the headcount, every person on the stand needs to be able to capture a lead themselves, including executives and engineers who are only there for one day.</p><p><strong>What should we measure to know if the show was worth it?</strong> Five numbers: conversations versus captures (your leak), 24-hour follow-up rate, cost per captured lead, sourced pipeline, and cost per opportunity or closed revenue. All five depend on tagging every lead with the event at the moment of capture — a one-minute setup beforehand that’s impossible to reconstruct afterwards.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/trade-show-lead-capture-booth-playbook"><em>https://lynqu.com</em></a><em> on July 24, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1c231960e0e2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/trade-show-lead-capture-the-booth-playbook-1c231960e0e2">Trade Show Lead Capture: The Booth Playbook</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Stop Losing Leads Between First Contact and Follow-Up | Lynqu]]></title>
            <link>https://medium.com/gravisun/how-to-stop-losing-leads-between-first-contact-and-follow-up-lynqu-8ce28c9f6ec1?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/8ce28c9f6ec1</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Tue, 30 Jun 2026 05:37:13 GMT</pubDate>
            <atom:updated>2026-06-30T09:11:04.587Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*i9vvxzGN9SD5jRBBqujyeg.png" /></figure><p><strong>Leads almost never vanish in one dramatic moment — they leak.</strong> Quietly, a few at a time, through the same handful of gaps between “nice to meet you” and the first real follow-up. Nobody decides to lose them; they just slip through the cracks while attention is somewhere else, and by the time anyone notices, the trail has gone cold.</p><p>That’s actually good news. A leak has a <em>location</em>. If leads were lost at random you’d be stuck, but they’re not — they escape at predictable points you can find and seal. You don’t stop losing leads by trying harder. You stop by making leakage hard to do by accident.</p><h3>Why do leads slip away between first contact and follow-up?</h3><p>Because the moment of contact and the moment of follow-up are almost never the same moment — and everything that can go wrong, goes wrong in the gap between them. At first contact, the lead is the hottest it will ever be: present, interested, easy to reach. Then real life intervenes. You move to the next conversation, the next meeting, the next inbox fire. The lead doesn’t get rejected; it gets <em>postponed</em> — and postponed leads quietly become lost leads.</p><p>The gap isn’t one hole. It’s five, and most teams are leaking through several at once.</p><h3>Where exactly do leads get lost? The five leak points</h3><p>Find which of these is yours before you fix anything — most “we’re bad at follow-up” problems are really one or two specific leaks:</p><ol><li><strong>The capture gap — the lead is never written down.</strong> It lived on a paper card, a half-remembered name, or a “I’ll add them later” that never happens. A lead you didn’t capture can’t be followed up by anyone. This is the biggest and most invisible leak, because you never even know what you lost.</li><li><strong>The context gap — you have the name but not the conversation.</strong> The contact exists, but there’s no note about what you discussed or why they mattered. So the follow-up would have to be generic — and generic feels like work, so it keeps getting put off until it’s too late to send.</li><li><strong>The ownership gap — everyone assumes someone else has it.</strong> On a team, a lead lands in a shared inbox or a shared pile and stalls in the space between “I thought you were handling that” and silence. No name on it means no one on it.</li><li><strong>The timing gap — intent decays before anyone acts.</strong> The lead was captured, with context, owned — and still got followed up a week later, by which point the person has moved on. Speed is its own leak: being <em>eventually</em> organized isn’t the same as being <em>fast</em>.</li><li><strong>The tracking gap — you can’t see who’s been contacted.</strong> With no single view of status, “I’ll get to them” is invisible. Leads drop silently because nobody can tell the difference between <em>contacted</em>, <em>waiting on them</em>, and <em>forgotten</em>.</li></ol><p>Notice these stack. A lead survives capture, then dies in the context gap. Another survives context, then dies because no one owned it. Plugging one leak just exposes the next — which is why piecemeal fixes feel like whack-a-mole.</p><h3>Why “I’ll just be more organized” never fixes it</h3><p>Because willpower doesn’t scale and it doesn’t survive a busy week. The plan “I’ll follow up diligently when things calm down” assumes a calm week that never comes — and it puts the entire system inside one person’s memory, which is exactly the thing that fails first.</p><p>Good intentions hold for about ten leads. Past that, the only thing that reliably works is removing the chances to leak in the first place: capture that happens automatically, context that’s attached on the spot, ownership that’s assigned by default, a clear trigger for <em>when</em>, and a list that shows status at a glance. The goal is a process where losing a lead requires effort, not where saving one does.</p><h3>How do you stop losing leads between first contact and follow-up?</h3><p>Seal each leak at its source. One fix per gap:</p><ul><li><strong>Capture at the moment of contact, not “later.”</strong> The only reliable time to record a lead is while you’re standing in front of them. Scan their card or share yours so the contact is saved before you move on — closing the capture gap before it opens.</li><li><strong>Attach the context immediately.</strong> A one-line note and an intent tag (hot / warm / nurture) the moment you meet, so the follow-up can be specific without an archaeology dig through your memory a week later.</li><li><strong>Give every lead one named owner.</strong> On a team, the moment a lead arrives it should belong to exactly one person. Shared responsibility is no responsibility.</li><li><strong>Set a timing trigger, not a someday.</strong> Decide the rule in advance — “hot leads hear back within the hour, everyone else within 24 hours” — so <em>when</em> isn’t a judgment call you make while exhausted. (For why the clock matters this much, see <a href="https://lynqu.com/blog/fastest-way-to-follow-up-with-sales-leads">the fastest way to follow up with sales leads</a>.)</li><li><strong>Work from one list with visible status.</strong> Every lead in one place, each marked contacted / waiting / done — so “who still needs a reply?” is a finishable list, not a vague guilt you carry around.</li></ul><p>Do these five and the gap between first contact and follow-up stops being a place where leads disappear. Once they’re captured, owned, and visible, the <em>act</em> of following up — what to actually write — gets much easier too; there are <a href="https://lynqu.com/blog/how-to-follow-up-after-a-networking-event">copy-paste templates and a 24-hour rule here</a>.</p><h3>A 60-second way to find your own leak</h3><p>You don’t need a CRM audit. Think back to the last event, week, or busy stretch and ask, in order:</p><ol><li><strong>How many people did you actually talk to?</strong> Not a guess — a real number.</li><li><strong>How many could you name and reach right now, without searching?</strong> The drop from question 1 to 2 is your <strong>capture gap</strong>.</li><li><strong>Of those, how many have a note about <em>why</em> they mattered?</strong> The drop here is your <strong>context gap</strong>.</li><li><strong>How many actually got a follow-up?</strong> The drop here is your <strong>ownership + timing gap</strong>.</li><li><strong>Could you tell me, today, which ones are still waiting on you?</strong> If not, that’s your <strong>tracking gap</strong>.</li></ol><p>Wherever the biggest drop is — that’s the leak to seal first. Most people are surprised it’s earlier in the funnel than they assumed: the lead they “lost in follow-up” was usually lost at capture.</p><h3>A quick example</h3><p>Imagine a rep back from a regional conference who’d swear they had “maybe twenty good conversations.” Run the audit. They can name nine. Of those nine, four have any context beyond a name. Two got a follow-up — the two most obvious ones — and they couldn’t say whether either had replied.</p><p>Eighteen of twenty leads didn’t die in follow-up; most were gone before follow-up was even possible. The lead they remember as “I should have emailed them back” was never the real loss — the real loss was the eleven they couldn’t even name.</p><p>Run it the sealed way instead. Each conversation captured on the spot — a scan, a one-line note, a hot/warm tag — every lead owned, every one visible in a single list. The follow-up the next morning isn’t heroic; it’s just working down a complete, tagged list while everyone’s still a fresh memory. Twenty leads kept instead of two, because the leaks were closed before the busy week could drain them.</p><p><em>(Illustrative example for clarity, not a specific customer.)</em></p><h3>Where a tool fits — and an honest caveat</h3><p>This is exactly the problem a lead-capture and engagement platform like is built to remove. Capture each lead the instant you meet them — <a href="https://lynqu.com/solutions/smart-scan">scan their card</a> or share your digital one — with a note and a tag attached on the spot; keep everyone in one list with a clear owner and a visible status; and reconnect in a tap. The leak points close by default instead of relying on you to remember. You can <a href="https://lynqu.com/business/lead-capture">start free</a>.</p><p>Honest caveat: a tool can’t make a lead want to buy, and it won’t write the genuine, personal message that earns a reply — that part is still yours. What it removes is the <em>leakage</em> — the un-captured card, the context you can’t reconstruct, the lead with no owner. It makes keeping a lead the easy default. Whether you then do something worthwhile with it is, rightly, up to you.</p><h3>FAQ</h3><p><strong>Why do I keep losing leads between first contact and follow-up?</strong> Because the two moments are separated by a busy gap, and leads leak through predictable points in it: they’re never captured, captured without context, owned by no one, followed up too late, or lost because no one can see their status. Most teams are leaking through more than one of these at once.</p><p><strong>Where do leads actually get lost?</strong> At five points: capture (never recorded), context (no note on why they mattered), ownership (everyone assumes someone else has it), timing (intent decays before anyone acts), and tracking (no visible status, so leads drop silently). The earliest gap — capture — is usually the biggest and the least noticed.</p><p><strong>How do I stop leads from falling through the cracks?</strong> Seal each gap at its source: capture every lead at the moment of contact, attach a note and intent tag immediately, give each lead one named owner, set a timing rule in advance, and work from a single list with visible status. The aim is a process where losing a lead takes effort, not where saving one does.</p><p><strong>Isn’t this just a matter of being more disciplined?</strong> No — discipline holds for about ten leads and then collapses under a busy week. Relying on memory and good intentions <em>is</em> the leak. A simple system that captures, assigns, and tracks leads automatically beats willpower because it doesn’t depend on a calm week that never arrives.</p><p><strong>How can I tell which leak is costing me the most?</strong> Run a quick audit: how many people you talked to, how many you can name and reach now, how many have any context, how many got a follow-up, and how many you can currently say are still waiting on you. The biggest drop between two steps is your worst leak — and it’s usually earlier in the funnel than people expect.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/stop-losing-leads-between-first-contact-and-follow-up"><em>https://lynqu.com</em></a><em> on June 30, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8ce28c9f6ec1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/how-to-stop-losing-leads-between-first-contact-and-follow-up-lynqu-8ce28c9f6ec1">How to Stop Losing Leads Between First Contact and Follow-Up | Lynqu</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Fastest Way to Follow Up With Sales Leads (Methods That Work) | Lynqu]]></title>
            <link>https://medium.com/gravisun/the-fastest-way-to-follow-up-with-sales-leads-methods-that-work-lynqu-0f0209ad9f45?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/0f0209ad9f45</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Thu, 18 Jun 2026 09:00:26 GMT</pubDate>
            <atom:updated>2026-06-18T09:11:10.438Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Gf3cZWqceMMlr06d_oapsA.png" /></figure><p><strong>The fastest way to follow up with a sales lead is to remove every step between “the lead appears” and “the lead hears from you.”</strong> That’s the whole game. Most deals aren’t lost to a better pitch — they’re lost in the gap between a raised hand and a reply, while the lead’s attention quietly moves on.</p><p>Speed is the rare sales lever that’s both high-impact and almost free: you don’t need a better script or a bigger team, just to stop letting fresh, interested leads cool. Here’s how to actually be fast, the methods that hold up, and how fast “fast” needs to be.</p><h3>Why does speed matter so much when following up with sales leads?</h3><p>Because intent has a short shelf life. The moment someone fills in a form, scans your card, or asks “how does this work?” is the hottest that lead will ever be — and from there both their <em>memory</em> of you and their <em>intent</em> to act decay fast. A day later they’re buried in their own work; a week later they barely remember reaching out.</p><p>The research backs the instinct: Harvard Business Review’s well-known study on lead response found that companies contacting a web lead within an hour were dramatically more likely to qualify it than those who waited even a few hours, and far more again than those who let it sit a day. The point isn’t a magic number — it’s the <em>shape</em> of the curve: your odds fall off a cliff early. So speed isn’t about looking eager; it’s about catching the lead while they still care.</p><h3>What’s the fastest way to follow up with a sales lead?</h3><p>In one line: <strong>capture the lead the instant you get it, get alerted in real time, send a pre-written first touch you personalize in seconds, and let them act in one tap.</strong> Each piece kills a delay — the lead never sits in your memory or an unchecked inbox, you’re not writing from a blank page, and the lead can reply or book without hunting for your details.</p><p>Fast follow-up is mostly mechanics. The people who do it aren’t more disciplined — they removed the steps in advance.</p><h3>The fastest follow-up methods that actually work</h3><p>Ranked by how quickly they get a real human exchange going:</p><ol><li><strong>An instant auto-acknowledgment.</strong> A short automated “Got it — [Name] will be in touch within the hour” buys time and sets an expectation. Honest caveat: it’s a placeholder, not the follow-up — it earns you minutes, it doesn’t replace a person.</li><li><strong>A pre-written first touch you personalize in seconds.</strong> Draft 2–3 templates ahead of time (hot inbound, warm in-person, light nurture), then swap in one specific detail and send. Specific-and-fast beats polished-and-slow every time.</li><li><strong>Reply on the channel they already used.</strong> Don’t make a lead switch context — someone who messaged on WhatsApp or scanned your card often replies faster to a text than a formal email. Match the channel to how you met.</li><li><strong>Real-time notification with a clear owner.</strong> Speed dies in shared inboxes where everyone assumes someone else has it. The moment a lead arrives, one named person should be pinged and own the first response.</li><li><strong>One-tap reconnect built into the message.</strong> A link to your digital card or calendar lets the lead act immediately, while they’re still warm — instead of “I’ll find time to reply” (they won’t).</li></ol><h3>How fast is fast enough?</h3><p>Match the speed to the heat of the lead:</p><ul><li><strong>Hot inbound</strong> (demo request, pricing question): <strong>within minutes</strong>, ideally the same hour. This is where speed-to-lead wins or loses deals.</li><li><strong>Warm, met-in-person</strong> (event, meeting, card scan): <strong>within 24 hours</strong>, while you’re still a vivid memory rather than a half-remembered name.</li><li><strong>Lighter / nurture</strong>: fast still helps, but consistency matters more than raw speed.</li></ul><p>The trap is letting <em>perfect</em> kill <em>fast</em>. A message with one real detail sent now beats a beautifully crafted one sent in three days. Speed first, then specificity — never the reverse.</p><h3>The mistake that makes “fast” impossible</h3><p>Trying to be fast without preparation. If your plan is “I’ll reply quickly when I get a minute,” you’ve already lost: leads are scattered across a form tool, your phone, and paper cards (so just <em>finding</em> them takes longer than the lead will wait), every reply starts from a blank page, and nobody clearly owns the first response — so leads stall between “someone will handle that” and silence.</p><p>Fast follow-up isn’t a personality trait. It’s a small amount of setup — templates, capture, alerts, ownership — done before the leads arrive.</p><h3>A quick example</h3><p>Imagine a rep who lands a strong demo request plus five solid conversations in one afternoon at a trade show. The instinct is to “follow up back at my desk” — which means a week of catch-up, six contacts blurred into one memory, and only the most obvious lead ever hearing back.</p><p>Run it the fast way instead. Each conversation is captured on the spot with a one-line note and a hot/warm tag, and a notification flags the demo request immediately. Within the hour the rep sends a personalized first touch from a pre-written template — one specific detail each, a calendar link in every message. Six warm replies that evening instead of one the following week. Nothing clever happened; the leads just heard back <em>while they still cared</em>.</p><p><em>(Illustrative example for clarity, not a specific customer.)</em></p><h3>Where a tool fits — and an honest caveat</h3><p>This is exactly what a lead-capture and engagement platform like is built for: capture each lead the instant you meet them (scan a card or share your digital one), keep the contact and context in one place with a clear owner, and reconnect in a tap — so the gap between “lead appears” and “lead hears from you” shrinks to minutes. You can <a href="https://lynqu.com/business/lead-capture">start free</a>.</p><p>Honest caveat: no tool makes a <em>slow team</em> fast on its own, and it won’t write a genuine message for you. What it removes is the friction — the digging, the retyping, the “who’s got this?” — so being fast becomes the easy default instead of a heroic effort. The judgment and the personal touch are still yours.</p><h3>FAQ</h3><p><strong>What’s the fastest way to follow up with a sales lead?</strong> Capture the lead instantly, get a real-time alert, send a pre-written first message that you personalize in seconds, and include a one-tap way to reconnect (a digital card or calendar link). The goal is to shrink the gap between when the lead appears and when they hear from you.</p><p><strong>How quickly should you follow up with a sales lead?</strong> For hot inbound leads, within minutes — ideally the same hour, since your odds of qualifying drop sharply after that. For leads you met in person, within 24 hours while you’re still a fresh, specific memory.</p><p><strong>Does responding faster to leads really win more deals?</strong> Yes. Research on lead response times (including a widely cited Harvard Business Review study) consistently shows that contacting a lead within the first hour makes you far more likely to qualify it than waiting even a few hours. Intent and memory both decay quickly.</p><p><strong>How can I follow up with leads faster without sounding generic?</strong> Prepare 2–3 first-touch templates in advance, then personalize each with one specific detail from your conversation. That requires capturing a quick note when you meet the lead — so speed and specificity work together instead of competing.</p><p><strong>Why do fast follow-ups fail in practice?</strong> Usually because there was no prep: leads are scattered across tools and paper, every reply starts from a blank page, and no one clearly owns the first response. Fast follow-up is a system you set up beforehand, not willpower in the moment.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/fastest-way-to-follow-up-with-sales-leads"><em>https://lynqu.com</em></a><em> on June 18, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0f0209ad9f45" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/the-fastest-way-to-follow-up-with-sales-leads-methods-that-work-lynqu-0f0209ad9f45">The Fastest Way to Follow Up With Sales Leads (Methods That Work) | Lynqu</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Capture & Manage Leads at Events (2026 Guide)]]></title>
            <link>https://medium.com/gravisun/how-to-capture-manage-leads-at-events-2026-guide-89fb3be73766?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/89fb3be73766</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Fri, 05 Jun 2026 11:00:21 GMT</pubDate>
            <atom:updated>2026-06-10T14:57:30.309Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NGT58fg1d14g2fR0Ucav1A.png" /></figure><p>For most teams the honest answer is “a handful, maybe.” The rest are sitting in a jacket pocket, a tote bag, or a photo roll — slowly going cold. That gap between <em>meeting</em> someone and <em>following up</em> is where most event budgets quietly leak away.</p><p>The good news: it’s a process problem, not a people problem. With a simple, repeatable system you can capture every lead at the moment you meet them, follow up before they forget you, and prove what the event was actually worth. Here’s how.</p><h3>Why do most event leads never turn into pipeline?</h3><p>It’s rarely one big failure — it’s four small ones stacked on top of each other:</p><p>The cost isn’t abstract. A lead that’s followed up within 24 hours is far more likely to convert than one you reach a week later — and every cold lead is money you already paid to acquire (flights, booth, time) and then let evaporate. <strong>Losing a lead is more expensive than never meeting them</strong>, because you’ve paid the cost without getting the return.</p><h3>The shift: from collecting cards to managing leads</h3><p>The fix is to stop thinking about <em>cards</em> and start thinking about a <em>lifecycle</em>. Every introduction should move through four stages:</p><ol><li><strong>Capture</strong> — get the contact into one place the moment you meet them.</li><li><strong>Engage</strong> — add context and follow up while the conversation is fresh.</li><li><strong>Manage</strong> — route the lead to an owner and a pipeline stage.</li><li><strong>Measure</strong> — track what converted, so you can prove (and repeat) what worked.</li></ol><p>A business card is just the <em>first touch</em> in that lifecycle — useful, but only if everything after it actually happens. Let’s walk through each stage.</p><h3>How should you prepare before an event?</h3><p>The teams that capture cleanly at the booth are the ones who decided <em>how</em> before they arrived. A few minutes of prep removes the friction that kills capture in a noisy hall:</p><ul><li><strong>Agree on the fields and tags every rep captures.</strong> Pick the 2–3 fields (e.g. name, company, email) and the intent tags (hot / warm / “nice to meet”) that everyone uses, so the data is consistent enough to act on later.</li><li><strong>Set up your digital card and QR ahead of time.</strong> Have your card ready to share in one tap or scan, so swapping details is a single motion instead of a fumble.</li><li><strong>Decide ownership and routing up front.</strong> Who owns the leads from this event, and how are they split across the team? Settling this before the show means no lead lands in a no-man’s-land afterward.</li><li><strong>Pre-write the follow-up.</strong> Draft the template you’ll personalize within 24 hours, so the post-event scramble is editing, not writing from scratch.</li><li><strong>Brief the team on one 30-second capture flow.</strong> Scan → tag → one-line note. When everyone runs the same motion, nothing slips through because “that wasn’t my job.”</li></ul><p>Prep is the cheapest leverage you have: it costs minutes before the event and saves the leads that would otherwise vanish during it.</p><h3>What should you do with business cards after a networking event?</h3><p>Don’t let them sit. The single highest-leverage habit is to <strong>digitize every card before you leave the venue</strong> — or that same evening at the latest, while you still remember who’s who.</p><p>Practical options, fastest to slowest:</p><ul><li><strong>Scan them with an AI card scanner.</strong> Point your phone at a card or badge; modern tools use AI vision to read the name, title, company, email, and phone, then save it as a digital contact in seconds. This is the only approach that scales when you’re holding a stack of 50.</li><li><strong>Add a one-line note to each contact immediately</strong> — where you met, what they cared about, the next step you promised. This 5-second habit is what turns a name into a real lead.</li><li><strong>Tag by intent</strong> (hot / warm / “nice to meet”) so your follow-up list writes itself.</li></ul><p>Avoid the two losing moves: typing cards into a spreadsheet days later (it won’t happen), or doing nothing and “getting to them eventually” (they go cold).</p><h3>How do you capture leads at a trade show or conference booth?</h3><p>At a booth you’re capturing volume under time pressure, so design for speed:</p><p>The goal is simple: by the time you tear down the booth, every conversation already exists as a structured lead, not a rubber-banded pile.</p><h3>How do you follow up so leads don’t go cold?</h3><p>Speed and relevance win. A few rules that consistently outperform a generic blast:</p><p>If your capture step saved context and tags, this step is almost effortless: you already know who’s hot, what they wanted, and who owns them.</p><h3>How do you manage and route leads across a team?</h3><p>Capture and follow-up are individual habits; <em>management</em> is where teams win or lose:</p><h3>How do you measure the ROI of an event?</h3><p>If you can’t measure it, you can’t defend the budget. Track, at minimum:</p><p>Once “the conference” becomes a number on a dashboard instead of a vibe, two things happen: you can prove the wins, and you can cut what doesn’t work. Both make next year’s events better.</p><h3>What this looks like in practice</h3><p>Picture a three-person team working a 200-booth expo over two days. Here’s the difference the system makes.</p><p><strong>The old way:</strong> they come home with a tote bag of ~180 business cards. In the weeks after, someone types maybe 30 into the CRM before the day job takes over. A few get a generic “great to meet you” email. By the time anyone asks “did the expo pay off?”, nobody can say which deals — if any — came from it.</p><p><strong>With a capture-first system:</strong> they agree on four fields and three tags before they go. On the floor, every badge gets scanned between conversations — 180 leads, captured on the spot, visible as a live count on a shared view. Each evening they add a one-line note and a hot/warm tag while the conversations are fresh. Within 24 hours, the ~40 hot leads get a personal, context-aware follow-up from whoever owns them; the warm ones drop into a sequence. A week later the pipeline tells the story: 180 captured → 165 followed up → 22 opportunities → a handful of deals, measured against the event’s all-in cost.</p><p>Same event, same conversations. The only difference is that the second team built a path for every lead to travel — so none of them quietly disappeared.</p><h3>The 6-step event lead checklist</h3><p>Copy this for your next event:</p><ol><li>☐ Before: agree on the fields + tags every rep captures.</li><li>☐ At the event: scan every badge/card on the spot; share your card via QR/tap.</li><li>☐ Same day: add a one-line note + intent tag to each lead.</li><li>☐ Within 24h: assign an owner and send a personal, context-aware follow-up.</li><li>☐ Week 1: move each lead through pipeline stages; sync to CRM.</li><li>☐ After: review leads captured → followed up → converted, vs event cost.</li></ol><h3>Where a tool helps (and where it doesn’t)</h3><p>You can run this system with a notebook and discipline — but at volume, a dedicated tool removes the steps that usually break. A lead-capture platform like is built around exactly this lifecycle: AI scanning to capture any badge or card, a contact hub to tag and note, a shared pipeline with capture-based attribution, and automatic CRM sync — so the leads you meet become tracked, followed-up pipeline instead of a pile of cards. You can <a href="https://lynqu.com/business/lead-capture">start for free</a> and bring a card you already use.</p><p>To be clear, no tool follows up <em>for</em> you or replaces a good conversation. What it does is make sure the lead is still there, in context, when you’re ready to act — which is the part teams most often get wrong.</p><h3>FAQ</h3><h3>What’s the best way to capture leads at an event?</h3><p>Capture each contact the moment you meet them — scan the badge or card with an AI scanner and add a quick note — rather than collecting cards to process later. On-the-spot capture is the only method that survives a busy booth.</p><h3>What should I do with business cards after a networking event?</h3><p>Digitize them the same day while you still remember the conversations: scan each card into a contact app, add a one-line note and an intent tag, then follow up within 24 hours. Don’t leave them for a “spreadsheet later” that never happens.</p><h3>How fast should I follow up with event leads?</h3><p>Within 24 hours. Follow-up speed is one of the biggest predictors of whether a lead converts, because you’re still a fresh, specific memory rather than a forgotten stranger.</p><h3>How do I measure whether an event was worth it?</h3><p>Track leads captured, follow-up rate, conversion to opportunity, and pipeline/revenue sourced — then compare that to the event’s all-in cost. Attribute leads to the rep who captured them for clean team metrics.</p><h3>Do I need an app to share or capture a card?</h3><p>No. A good digital card is shared via a QR code or tap and opens in any browser — the person receiving it doesn’t need to install anything. You only need a capture tool on your side.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/how-to-capture-and-manage-leads-at-events"><em>https://lynqu.com</em></a><em> on June 5, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=89fb3be73766" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/how-to-capture-manage-leads-at-events-2026-guide-89fb3be73766">How to Capture &amp; Manage Leads at Events (2026 Guide)</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Digital vs Paper Business Cards: What’s the Real Difference (and Which Should You Use)?]]></title>
            <link>https://medium.com/gravisun/digital-vs-paper-business-cards-whats-the-real-difference-and-which-should-you-use-61b7ccfe4e7a?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/61b7ccfe4e7a</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Mon, 01 Jun 2026 12:16:26 GMT</pubDate>
            <atom:updated>2026-06-01T12:16:26.115Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mT-D2BLHG-aeFWNUdfl7YA.png" /></figure><h3>I tested digital business cards against paper for a year. Paper won exactly twice.</h3><p><em>The two situations where paper still beats digital — and the comparison most people get wrong everywhere else.</em></p><blockquote><em>Disclosure: I’m a co-founder of </em><a href="https://lynqu.com/"><em>Lynqu</em></a><em>, a digital business card platform. I have an obvious stake in the answer — so I tried to write the version of this piece I’d want to read if I weren’t already in the industry, including the parts where paper still wins.</em></blockquote><p>You meet someone interesting at a conference. You both reach for your wallets. They hand you a paper card; you hand them yours. Three days later that card is in a drawer with eleven others, and neither of you can quite remember why you wanted to follow up.</p><p>Replay the same moment with a digital card. One tap, contact saved directly into the phone, with your website, your calendar link, and your LinkedIn already attached. You can see the next day that they actually opened it.</p><p>That’s the real difference between digital and paper business cards — and it’s not the one most people argue about. The argument is usually framed as <em>traditional vs. modern</em>. The practical question is much simpler: what happens to your card after the handshake?</p><p>I spent the last year running an unscientific but honest comparison — handing out both formats, tracking what came back as actual follow-ups, paying attention to the moments where each one felt right. Here’s what I found.</p><h3>The job a business card is actually hired to do</h3><p>Before comparing formats, it helps to strip the format away. A business card — paper or digital — is hired to do three things:</p><ol><li>Transfer your contact info into the other person’s phone, reliably and accurately.</li><li>Make a memorable first impression that says something about who you are.</li><li>Give them a reason to reconnect, and make reconnecting frictionless.</li></ol><p>The interesting question isn’t <em>which one is better in the abstract.</em> It’s <em>which one does the job better in the moments that matter to you.</em> That’s the lens for everything below.</p><p>Most “paper vs. digital” arguments skip straight to features and pricing. Features and pricing are downstream. If the job is “transfer info into the other person’s phone,” the format that puts it there in one second — without typing — has a structural advantage. If the job is “feel premium at a black-tie event,” a beautifully printed letterpress card might still win. Same product category. Completely different criteria.</p><h3>Eight things people actually use a card for</h3><p>Here’s how the two formats score against the jobs that come up most often. I’m intentionally avoiding a table — Medium turns them into mush — so each item is paired below.</p><p>Getting into the recipient’s phone contacts.<br>Paper: manual typing later, which almost never happens. Digital: one tap or QR scan, saved instantly. Not close.</p><p>Surviving longer than a week after the meeting.<br>The most-cited industry estimate puts paper retention at around 12% within a week (Statistic Brain). Digital cards live in the contact app permanently. They don’t get tossed.</p><p>Updating your title, email, or phone number.<br>Paper: reprint the whole batch. Digital: edit once and every shared link updates everywhere — including the card someone saved last year.</p><p>Working when you’ve run out / forgotten them.<br>Paper doesn’t. Digital is always in your pocket.</p><p>Carrying more than name + number.<br>A paper card squeezes onto a 3.5&quot;×2&quot; rectangle. A digital card has effectively unlimited fields — calendar, portfolio, services, payment links, latest case study.</p><p>Consistent branding across a team.<br>Paper drifts with each reorder cycle. Digital cards on a team plan come from one template, so the whole company looks like the same company.</p><p>Telling you whether the lead actually engaged.<br>Paper tells you nothing. Digital cards report views, taps, and saves — which is the difference between guessing whether networking works and knowing.</p><p>Feeling tactile and premium in formal settings.<br>Paper wins here. Even a high-end NFC card can’t quite replicate the weight of a letterpress card on heavy stock. This is real and worth being honest about.</p><p>Six out of eight lean digital. Two lean paper — and they’re not throwaway points.</p><h3>Where paper genuinely won (the two times)</h3><p>I said in the headline that paper won exactly twice in my year of testing. Here’s where, both times:</p><p>The first was at a small private dinner with about ten people, hosted by an investor. The room was warm, conversation was good, and at the end someone handed out their card. It would have been jarring to pull out a phone. The physical card carried something the moment needed — a small ritual of mutual respect, a tactile object that lingered after the dinner ended.</p><p>The second was at a corporate event in Tokyo. The Japanese <em>meishi koukan</em> — the formal business card exchange — is not just an information transfer. It’s a small ceremony: presented with two hands, received with two hands, read carefully, placed thoughtfully on the table during the meeting. Trying to substitute a phone screen would have been a category mistake. Paper isn’t the inferior version of digital there. It’s a different object doing a different job.</p><p>There are a few other situations I didn’t personally hit but where I’d still keep a paper card on hand:</p><ul><li>Backup for dead batteries, no signal, or phone-down policies — hospitals, certain manufacturing floors, secure facilities.</li><li>Branding objects in themselves — a photographer’s letterpressed card or a chef’s embossed mini-menu is a marketing asset, not just contact info.</li><li>Audiences who genuinely prefer it — a subset of older or less tech-comfortable clients will appreciate the physical card. Pretending otherwise is unhelpful.</li></ul><p>None of those is <em>“make paper your primary card again.”</em> They’re <em>“yes, also carry a few.”</em></p><h3>What digital does that paper physically can’t</h3><p>These aren’t preferences. These are things paper <em>cannot</em> do, no matter how nicely it’s printed:</p><ul><li>Reach the contact app without typing. Anyone who has typed someone else’s paper card into their phone knows it almost never happens. Digital cards save in one tap or scan.</li><li>Stay current after you change roles. Your paper card from two years ago — the one with the old phone number — is still in someone’s drawer, misinforming people on your behalf. A digital card link updates everywhere, instantly.</li><li>Carry your full professional context. Calendar link, portfolio, services, payment link, social profiles, reviews, latest deck. None of that fits on a paper rectangle.</li><li>Get forwarded. A saved digital card can be sent to a colleague who might need you. Paper stops in the wallet it lands in.</li><li>Tell you anything. You hand out 500 paper cards and learn nothing. Digital cards report views, taps, and saves — which is the difference between guessing and knowing.</li></ul><p>A way to frame the gap: a paper card <em>ends</em> when you hand it over. A digital card <em>starts</em> when you hand it over.</p><h3>The hybrid most people I know actually land on</h3><p>After a year of running this, I’m not anti-paper. I’m anti-paper-as-the-default. The setup that’s worked for me, and for most of the people I’ve watched try this, is a hybrid:</p><ul><li>Digital as the default share. Used in roughly 90% of the moments — the coffee, the conference floor, the casual intro.</li><li>A small stack of paper for the moments paper is genuinely better. The 10% above. The formal dinner, the regulated context, the no-signal venue.</li><li>The paper that remains has a QR code on it pointing to the digital card. So even the paper hand works harder than it used to — the recipient can save the contact in one scan instead of typing.</li></ul><p>The “either/or” framing dissolves once you see them as layers. You’re not abandoning a habit you’ve had for twenty years. You’re upgrading it. And your print order drops by 60–80% along the way.</p><h3>A four-question check before you decide</h3><p>If you’re trying to decide for yourself, run these in order. The first “yes” is your answer:</p><ol><li>Do you ever change your title, phone number, employer, or office? → You need a digital card. Reprinting on every change is a tax you don’t have to pay.</li><li>Do you meet more than ~3 new professional contacts per month? → Digital pays for itself on contacts saved alone.</li><li>Do you care whether your follow-ups actually happen? → You need digital. Paper can’t tell you who opened, saved, or shared your card.</li><li>Do you only network 2–3 times a year, in formal-only settings, with audiences who won’t use a phone? → Stay paper, or use a free digital plan as backup.</li></ol><p>For almost everyone I talk to, questions 1–3 are all yes. Which is why the hybrid setup is what tends to stick.</p><h3>How to test it without throwing away your paper cards</h3><p>The lowest-risk way to evaluate this is staged, not all-or-nothing:</p><ol><li>Create a free digital card. Should take ~10 minutes on any reputable platform.</li><li>Use it as your default for one month. Tap or QR-scan when you meet someone new. Keep paper in your wallet for the moments that feel right.</li><li>At the end of the month, look at two numbers. How many people saved your digital card (it will tell you). How many paper cards came back as actual follow-ups. The gap is usually pretty clear.</li></ol><p>If digital is doing the job, reduce your next paper order by 60–80%. If it isn’t — say you work in a context where paper genuinely fits better — you’ve spent zero money and lost nothing.</p><h3>A few questions I get every time I write about this</h3><p>Are paper business cards dead in 2026? No, but their role has narrowed. They still have a place in formal and ceremonial settings, in some cultural contexts, and as tactile branding objects. What’s changed is that paper is no longer the default — most professional contexts now expect a digital exchange or a QR code on the paper card.</p><p>Do digital business cards work without internet? The first share usually needs a brief connection (to fetch the contact page), but once the recipient has saved your contact, it’s stored locally like any other contact. Some platforms also support offline-first NFC handoff for venues with no signal.</p><p>Are digital cards professional enough for senior or executive meetings? Yes — and increasingly more polished than paper for senior audiences. A well-designed digital card carries calendar links, your LinkedIn, your latest deck, and your contact details in one tap. That’s more useful to a busy executive than something they have to re-type later.</p><p>Can I keep paper and add digital on top? Yes. Print a QR code on your paper card pointing to your digital one. Recipients tap the digital version straight into their phone, or hold onto the paper for the tactile memory. Both worlds, no choosing.</p><h3>The bottom line</h3><p>After a year of running both side by side, my honest summary is this: paper still has a job, but it’s a much narrower one than most people assume. The hybrid — digital as default, paper as a small considered tool for specific moments — is what’s worked for me and almost everyone I’ve watched try this.</p><p>The lowest-risk way to find out which side of the line you’re on is to test it for a month with the platform of your choice and look at the numbers. Whichever platform you pick, the framework above applies the same way.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/digital-vs-paper-business-cards"><em>lynqu.com/blog/digital-vs-paper-business-cards</em></a><em>. If you want to try the platform I work on, </em><a href="https://lynqu.com/guest-card/new"><em>Lynqu has a free tier</em></a><em> — no card required. Plenty of good alternatives exist too; the framework above works for any of them.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=61b7ccfe4e7a" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/digital-vs-paper-business-cards-whats-the-real-difference-and-which-should-you-use-61b7ccfe4e7a">Digital vs Paper Business Cards: What’s the Real Difference (and Which Should You Use)?</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Are Digital business cards worth it for small businesses?]]></title>
            <link>https://medium.com/gravisun/i-ran-the-real-numbers-on-business-cards-for-a-5-person-team-c0bcd0a51c4f?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/c0bcd0a51c4f</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Sat, 30 May 2026 07:49:18 GMT</pubDate>
            <atom:updated>2026-05-30T11:38:44.923Z</atom:updated>
            <content:encoded><![CDATA[<h3>I ran the real numbers on business cards for a 5-person team. Paper lost — but not for the reason you’d guess.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tGevgdZIVoEBzxobKx79bw.png" /></figure><p><em>Hidden reprint costs, the 88% trash rate, and the one situation where paper still beats digital.</em></p><blockquote><em>Disclosure: I’m a co-founder of </em><a href="https://lynqu.com/"><em>Lynqu</em></a><em>, a digital business card platform. I have an obvious stake in the answer — so I tried to write the version of this piece I’d want to read if I weren’t already in the industry, including the parts where paper still wins.</em></blockquote><p>You’re standing at the checkout for another box of business cards. 500 cards, four-color print, decent stock — about $80. You ordered the last batch eight months ago. Two people on your team changed roles since then, your office moved last month, and the new hire still hasn’t had any cards printed at all.</p><p>That’s the moment most small business owners start Googling whether digital business cards are actually worth it.</p><p>The short answer for a team of 2–10 is <em>almost always yes</em> — but the interesting part is why. It isn’t the price tag. It’s everything paper cards quietly cost you that never shows up on the invoice.</p><h3>What you’re actually spending on paper</h3><p>Most small businesses underestimate the total cost of paper cards by a factor of two or three, because the print bill is the only line item they see. Walk through the real number for a 5-person team:</p><ul><li>The base print run — 500 cards per person, printed once a year: ~$400 (5 × $80).</li><li>Reprints when details change — someone’s role, email, phone, or office updates: +$80–$160/year.</li><li>Cards printed for people who leave within 6 months: +$80–$160/year, straight to the bin.</li><li>Cards literally thrown out — industry estimates put this at ~88% within a week: a sunk cost on most of the print run itself.</li><li>Time spent reordering, distributing, and managing: roughly 4–8 hours/year of someone’s actual workweek.</li><li>Brand inconsistency from staggered reorder cycles: hard to price, but real — different fonts, slightly different logos, mismatched colors across the team.</li></ul><p>Add it up and you’re at roughly $600–$800/year in direct costs for a small team — not counting time or opportunity cost.</p><p>The deeper number — the one most owners don’t want to look at — is the percentage of those cards that ever turn into anything. Industry research consistently puts the “card actually saved into a phone” rate in the low double digits. The other 80–90% is a marketing budget you set on fire at networking events.</p><h3>What digital costs, honestly</h3><p>Pricing in this category is converging. For small businesses (2–10 employees) the landscape is roughly:</p><ul><li>Free plans. Most reputable digital business card platforms offer a usable free tier. Fine for solo founders or for trialing the experience before committing.</li><li>Pro / Individual. Typically $5–$8/user/month. Removes branding limits, adds analytics, custom domains on the card, and richer templates.</li><li>Business / Team. Typically $8–$15/user/month. Adds shared brand control, central admin, team templates, CSV imports, SSO.</li></ul><p>For a 5-person team on a mid-tier plan, you’re looking at roughly $300–$700/year — usually less than what they’re already spending on paper. And the comparison isn’t apples to apples, because the digital version does things paper physically can’t.</p><h3>The ROI math: when does it pay back?</h3><p>A digital business card pays for itself the moment one extra lead per team member per year actually saves your contact. That’s a low bar. The mechanism is unglamorous but effective:</p><ul><li>Tap or QR works in about a second, which means the share happens <em>before</em> the conversation ends. Paper exchanges interrupt the conversation; digital ones don’t.</li><li>The contact goes straight into the phone, not a wallet that gets emptied on Sunday.</li><li>Updates are instant. Change a phone number once and every previously shared card is now correct — even the one someone saved two years ago.</li><li>You can see who actually saved your details, which is a research insight you’ve never had from paper.</li></ul><p>Even at a conservative 10% lift in retained contacts per networking event, the math works out for any small business spending more than ~$200/year on print.</p><h3>What digital does that paper structurally can’t</h3><p>This is where the cost-benefit conversation usually flips. Paper can’t:</p><ul><li>Update in place. New role, new number, new logo? Reprint vs. one click.</li><li>Carry your whole context. A digital card holds your website, calendar link, portfolio, services, social profiles, payment links, and reviews — same tap.</li><li>Tell you anything. You drop 500 paper cards into the world and learn precisely nothing. Digital cards report views, taps, and saves.</li><li>Stay current for someone who took it three years ago. That old paper card is wrong. Your old digital card link is still right.</li><li>Get re-shared past the room you’re in. A saved digital card can be forwarded. Paper stops at the wallet it lands in.</li></ul><p>For service businesses, real estate, agencies, consultants, and B2B sales of any kind, that compounding effect is the actual product.</p><h3>When paper still wins (the honest part)</h3><p>I’m not going to pretend paper is dead. There are real situations where it’s still the right call:</p><ul><li>Older or less tech-comfortable clients in some industries genuinely prefer something tactile. Don’t pretend otherwise.</li><li>Regulated sectors — legal, medical, certain financial niches — sometimes have formal expectations around physical credentials.</li><li>Black-tie / formal events where pulling out a phone breaks the tone of the room.</li><li>Backup for dead batteries, no signal, or phone-down policies in hospitals and some manufacturing floors.</li><li>Branding objects in themselves. A letterpressed card for a luxury brand is a marketing asset, not just a contact mechanism.</li></ul><p>The realistic answer for most small businesses isn’t <em>replace paper entirely.</em> It’s <em>carry digital as the default and a small stack of paper for the edge cases.</em> You’ll likely cut your print order by 60–80%.</p><h3>A simple decision framework</h3><p>If you’re trying to decide, work through these in order. The first “yes” you hit is your answer:</p><ol><li>Do you reorder cards more than once a year because details change? → Digital pays for itself just on reprint savings.</li><li>Does your team meet more than ~5 new people per month each? → Digital pays for itself on lead capture alone.</li><li>Do you want any visibility into whether networking is actually working? → You need digital. Paper can’t tell you anything.</li><li>Are brand consistency or rapid hiring real concerns? → Digital team plans solve both in one move.</li><li>Do you only need a card 2–3 times a year, in formal settings, for an audience that won’t tap a phone? → Stay paper, or use a free digital plan as backup.</li></ol><p>For the vast majority of small businesses I talk to, questions 1–3 are all “yes” — which is why the answer to “is it worth it” lands the way it does.</p><h3>A few questions I get every time I write about this</h3><p>Are digital business cards professional enough for client meetings? Yes. Most professional contexts now expect digital exchange, and a well-designed digital card looks more polished than a generic paper one. The exceptions are the formal industries above.</p><p>Does the recipient need the same app? No. They don’t need to install anything — most digital cards work via NFC tap or QR scan and open in the recipient’s browser, saving directly into their phone’s native contacts.</p><p>What happens to the data when someone leaves the team? On a business-tier plan, admins can reassign or deactivate cards centrally — something paper definitionally can’t do. Old paper cards exist in the wild forever.</p><p>Is the free plan enough? For a one-person operation, often yes. For teams, the paid tiers earn their keep through shared branding, analytics, and admin controls — which is where most of the real ROI is.</p><p>How long does it take to switch a team over? Typically under an hour for a 5-person team, including importing existing contacts and assigning cards. The biggest variable is design — templates get you to “done” much faster than building cards from scratch.</p><h3>The bottom line</h3><p>For a small business with 2–10 employees, digital business cards usually cost less than paper, do more than paper, and give you information paper never could. The honest exceptions exist — they’re narrower than most owners assume.</p><p>The lowest-risk way to test the call is to try a free plan with one or two people for a month, see whether saved-contact rates and follow-ups actually move, and decide from there. If it works, roll it to the team. If it doesn’t fit your industry, you’ve lost nothing but an hour.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/are-digital-business-cards-worth-it-for-small-businesses"><em>lynqu.com/blog/are-digital-business-cards-worth-it-for-small-businesses</em></a><em>. If you want to try the platform I work on, </em><a href="https://lynqu.com/pricing"><em>Lynqu has a free tier</em></a><em> — no card required. Plenty of good alternatives exist too; the framework above works for any of them.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c0bcd0a51c4f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/i-ran-the-real-numbers-on-business-cards-for-a-5-person-team-c0bcd0a51c4f">Are Digital business cards worth it for small businesses?</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Digital Business Card Category Stopped Trying. Here’s What’s Actually Worth Using in 2026.]]></title>
            <link>https://medium.com/gravisun/the-digital-business-card-category-stopped-trying-heres-what-s-actually-worth-using-in-2026-21b33aa158c9?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/21b33aa158c9</guid>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[networking]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[apps]]></category>
            <category><![CDATA[saas]]></category>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Tue, 05 May 2026 17:53:38 GMT</pubDate>
            <atom:updated>2026-05-05T17:54:15.441Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QfHodZd8lFVjhljtjIHU7Q.png" /></figure><h3>After two months of testing every major app on the market, I’m done pretending they’re all the same.</h3><p>The pitch is always identical: “Replace your paper cards. Share your contact info with a tap.” Every app in this category for the last five years has run that same one-line pitch — and most of them have stopped innovating past it.</p><p>I spent the last two months actually using all seven of the major players. I paid for the paid tiers, used each as my primary card at real events, handed my phone to non-technical friends to test how the recipient experience felt, and verified pricing on every vendor’s page in March 2026. What I found is that the category is more interesting than the consensus suggests, but not in the way the marketing claims.</p><p>The mistake everyone makes is treating “best digital business card app” as a single question. It’s seven different questions wearing the same hoodie. Each app on this list is genuinely the best at <em>something</em> — and most of them are mediocre at everything else.</p><blockquote><strong><em>Full disclosure before we go further:</em></strong><em> I’m the founder of </em><strong><em>Lynqu</em></strong><em>, one of the seven apps in this comparison. I’m telling you upfront so you can weight my opinion accordingly. The reason I’m publishing this anyway is that I genuinely couldn’t find an honest cross-category comparison when I was researching the market in 2024 — most “best of” lists are affiliate roundups that test nothing. So I made the one I wanted. I’ve named competitors as the better pick in three categories below, including ones where Lynqu is a direct alternative. If you’d rather skip my product entirely, the article still works — pick the second-best option in each category.</em></blockquote><h3>TL;DR — winners by category</h3><p>If you don’t want to read 2,500 words:</p><p><strong>Best all-in-one platform → Lynqu.</strong> Card + email signature + virtual background + AI scanner in one app.</p><p><strong>Best for wide adoption → Blinq.</strong> Largest user base, most polished free plan.</p><p><strong>Best for events &amp; lead capture → Popl.</strong> NFC hardware + 4,000+ integrations.</p><p><strong>Best for enterprise security → Mobilo.</strong> SOC 2 Type II + deep CRM.</p><p><strong>Best for customization → HiHello.</strong> Templates + contact enrichment.</p><p><strong>Best premium NFC hardware → V1CE.</strong> Metal/bamboo cards, lifetime replacement.</p><p><strong>Best simplicity → Wave Connect.</strong> Cleanest onboarding.</p><p>Now the long version, with the reasoning.</p><h3>Why “the category stopped trying” — and where it didn’t</h3><p>Here’s what I mean by “stopped trying.” Five of the seven apps below have shipped roughly the same feature set for three years. The free tiers exist mostly as conversion funnels (one app caps free contacts at five — <em>five</em>). Privacy is treated as a checkbox, not a feature. Multi-language support is “we have English.” The receiving experience — what happens when a non-technical person scans your QR — is an afterthought on most of them, even though that’s literally the entire user-facing surface of the product.</p><p>Two apps in this list (Blinq and Popl) have grown to multi-million user bases despite this. The argument from the consensus has been that the category is “solved” because the leaders are big. The argument I’d make after testing everything is that the category is <em>underserved</em> because the leaders stopped iterating.</p><p>That’s the gap each of the seven apps below fills differently. Here’s how.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZqUgd_HNXIjbF3NW4U2LJw.png" /><figcaption><a href="https://lynqu.com/">lynqu.com</a></figcaption></figure><h3>1. Lynqu — Best all-in-one platform</h3><p><em>Pricing: Free / $5.99/mo (Pro, $4.99 annual) / $4.99 per user (Business, $3.99 annual) / Custom (Enterprise)</em></p><p>I’ll keep this brief because I built it and you should weight my opinion accordingly. The pitch is that Lynqu is one app instead of three: digital business card, email signature generator, Zoom-style virtual background, and an AI scanner that reads paper cards and event badges. Most of the other six on this list either pick one of those lanes (Popl is a card + lead capture tool; Blinq is a card + CRM tool) or charge for them à la carte.</p><p>Two things I’d defend even on a competitor’s blog:</p><p><strong>Privacy modes</strong>. Three states (Public, Protected, Private) on the same card. The same QR code can show your full info to a paying client and a stripped-down version to someone you just met at a bar. None of the other six on this list do this in any meaningful way — they’re all “share or don’t.”</p><p><strong>Multi-language out of the box</strong>. 12 locales from day one, including Japanese, Polish, Hungarian, and Brazilian Portuguese. Every other app on this list is English-only. If half your clients aren’t in the US, that gap is bigger than it sounds.</p><p><strong>Where Lynqu loses:</strong> smaller user base than Blinq (we launched in 2026; they launched in 2017). No physical NFC cards — if you want a metal tap-to-share card, see V1CE or Mobilo below. The free tier is genuinely useful, but the Business tier doesn’t have the integration depth of Popl or Mobilo yet (it’s coming, but it’s not there today).</p><p><strong>Skip Lynqu if:</strong> you need NFC hardware for now, you’ve standardized on a CRM stack with 4,000+ Zapier-style integrations, or you specifically want the highest-adopted app for network-effect reasons.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b3VjfpuLDsMFhma2pSep1A.png" /><figcaption><a href="https://blinq.me/">blinq.me</a></figcaption></figure><h3>2. Blinq — Best for wide adoption</h3><p><em>Pricing: Free / $9.99/mo (Premium, $7.33 annual) / $6.99 per user (Business, $4.99 annual) / Custom</em></p><p>Blinq is the safe default. 2.5M users, Fortune 500 logos, 150K+ App Store reviews at 4.9 stars. If your boss asks “what app are you using for your card,” saying “Blinq” requires zero explanation.</p><p>The free plan is genuinely generous — virtual backgrounds and email signatures are included, which most competitors paywall. The receiving experience is clean: when someone scans your QR, they figure out how to save the contact without you having to coach them. That sounds like a low bar until you watch a normal human try to save a contact from one of the worse apps on this list.</p><p><strong>Where Blinq loses:</strong> the card itself is sparse. No services with booking links, no photo galleries, no audio clips, no document attachments. If you’re a freelancer or consultant who wants the card to <em>do</em> more than display contact info, you’ll outgrow Blinq quickly. English-only interface. Premium at $9.99/mo is the most expensive consumer tier in the category.</p><p><strong>Pick Blinq if:</strong> you want the safest, most-recognized choice and your card just needs to be a digital vCard with a logo on it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Yb8Roj55ukffuZCAajZh1g.png" /><figcaption><a href="https://www.hihello.com/">hihello.com</a></figcaption></figure><h3>3. HiHello — Best for customization</h3><p><em>Pricing: Free (4 cards) / $8/mo (Professional, $6 annual) / $6 per user (Business, $5 annual) / Custom</em></p><p>HiHello has been around since 2018 and it shows in the polish. Their card templates are the cleanest in the category. The Professional plan adds AI contact enrichment, which automatically fills in a lead’s company, role, and LinkedIn after they save your card.</p><p>Where HiHello earns its place: if you really care about how your card <em>looks</em> and you’re willing to spend time designing it, this is the most flexible canvas. Watch app, widgets, Live Activities — all the iOS chrome is there.</p><p><strong>Where HiHello loses:</strong> it’s the most expensive Pro plan in the category at $8/mo. Analytics are paywalled. The free plan caps you at 5 scans/month, which is not enough if you actually attend events. English-only. The card content surface is narrower than Lynqu’s — no services/bookings, no audio, no documents.</p><p><strong>Pick HiHello if:</strong> card aesthetic is your top priority and you’ll be spending real time customizing rather than just dropping in your details.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kg_rgt2p7rBJjdSBi1rCzA.png" /><figcaption><a href="https://popl.co/">popl.co</a></figcaption></figure><h3>4. Popl — Best for events &amp; lead capture</h3><p><em>Pricing: Free (5 contacts) / $7.99/mo (Pro) / $14.99/mo (Pro+) / Custom</em></p><p>If you live at trade shows, this is your app. Popl bills itself as a “go-to-market platform” and the orientation shows: 4,000+ integrations, AI badge scanning that’s been refined over years of event use, physical NFC products (cards, wristbands, phone tags) that ship with their own branding.</p><p>The lead capture workflow is genuinely the best on this list. Scan a badge, the contact lands in your CRM with a follow-up sequence triggered, and you’re already onto the next handshake.</p><p><strong>Where Popl loses:</strong> the free plan caps contact storage at <strong>5 contacts</strong>. Five. If you’re testing the app at a single conference, you’ll hit that limit during the first hour of the first day. Paid plans run $7.99–$14.99/mo, which is the steepest in the category. No email signatures, no virtual backgrounds. The receiving experience tilts toward “lead capture” rather than “person-to-person exchange,” which is the right call for sales teams and the wrong call if you’re a consultant or solo professional.</p><p><strong>Pick Popl if:</strong> you’re a sales rep or event-focused team that needs CRM-routed lead capture and physical NFC hardware.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XXfGlAjFliEorjSMRpPluA.png" /><figcaption><a href="https://www.mobilocard.com/">mobilocard.com</a></figcaption></figure><h3>5. Mobilo — Best for enterprise security</h3><p><em>Pricing: Free (digital) / $4 per user (Teams) / $5 per user (Enterprise) — NFC cards $5–$69 one-time</em></p><p>Mobilo is the boring-in-a-good-way choice for IT-led rollouts. They’re SOC 2 Type II certified, integrate with Salesforce / HubSpot / Pipedrive, and offer SSO and directory sync. The pricing model is unusual: you buy the NFC card once ($5–$69) and the digital profile is free for individuals. Teams pay per seat.</p><p>If you’re an enterprise IT buyer, Mobilo’s procurement story (one-time hardware purchase, predictable per-seat software, SOC 2 documentation) is easier to get past your security review than most competitors’.</p><p><strong>Where Mobilo loses:</strong> no email signature, no virtual background, no AI follow-ups, no dark mode (yes, in 2026), no multi-language. The card itself is functional but not delightful. The full experience really does require buying NFC cards.</p><p><strong>Pick Mobilo if:</strong> you’re an enterprise procuring for sales teams and your security team’s checklist is the gating factor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vd0dPyOWWjdVgixmOSStzA.png" /><figcaption><a href="https://v1ce.co/">v1ce.co</a></figcaption></figure><h3>6. V1CE — Best premium NFC hardware</h3><p><em>Pricing: Free (digital) / $79+ (NFC cards) / $197 flat (Teams)</em></p><p>V1CE is the only one on this list where the physical card is the product and the software is the supporting cast. The cards themselves are gorgeous — metal, bamboo, custom prints — and they offer <strong>free lifetime replacement</strong> if your card breaks or fades, which is the kind of warranty no one else on this list will touch.</p><p>The web profile pages are also among the most visually polished. If your brand needs the card itself to be a status object, V1CE delivers.</p><p><strong>Where V1CE loses:</strong> $79 minimum for an NFC card and $197 flat for Teams is steep. No AI scanner, no email signatures, no virtual backgrounds. Web-based — no native iOS/Android app. Limited integrations.</p><p><strong>Pick V1CE if:</strong> the physical card is the actual marketing artifact (luxury brands, high-end consultants, anyone whose card needs to feel expensive in someone’s hand).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KTEic1wqCyMLADuGj4g7rw.png" /><figcaption><a href="https://wavecnct.com/">wavecnct.com</a></figcaption></figure><h3>7. Wave Connect — Best simple option</h3><p><em>Pricing: Free / $9/mo (Pro, $7 annual) / $7 per user (Teams, $5 annual)</em></p><p>Wave Connect is the app I’d recommend to a relative who asked “what’s the easiest one.” The onboarding is the simplest in the category — you can have a working card in under two minutes, and the interface is genuinely minimalist instead of the pretend-minimalism that tends to mean “we hid the features.” NFC products are available. Teams pricing is competitive.</p><p><strong>Where Wave Connect loses:</strong> the feature surface is narrower than the others. No email signatures, no virtual backgrounds, no AI enrichment, no multi-language. It’s the app for people whose answer to “what features do you need?” is “as few as possible.”</p><p><strong>Pick Wave Connect if:</strong> you want the lowest-friction setup and you’re allergic to feature creep.</p><h3>A decision matrix that actually helps</h3><p>The “best” app depends on what you’re optimizing for. If you only have 30 seconds, here’s the matrix I wish I’d had when I started this category:</p><p><strong>Pick by overall fit:</strong></p><ul><li><em>One app that replaces three</em> (card + signature + virtual background) → <strong>Lynqu</strong></li><li><em>The safest, most-adopted choice</em> → <strong>Blinq</strong></li><li><em>The best card aesthetic with deep customization</em> → <strong>HiHello</strong></li><li><em>NFC hardware + event lead capture + 4,000 integrations</em> → <strong>Popl</strong></li><li><em>SOC 2 + enterprise CRM + IT-friendly procurement</em> → <strong>Mobilo</strong></li><li><em>A premium physical card that doubles as a status object</em> → <strong>V1CE</strong></li><li><em>The simplest possible setup</em> → <strong>Wave Connect</strong></li></ul><p><strong>Pick by specific must-have feature:</strong></p><ul><li><em>Multi-language card for international clients</em> → <strong>Lynqu</strong> (only one with 12 locales)</li><li><em>Privacy modes — different versions for different audiences</em> → <strong>Lynqu</strong> (only one that does this)</li><li><em>AI follow-up reminders after meetings</em> → <strong>Lynqu</strong></li><li><em>AI contact enrichment</em> → <strong>Blinq, HiHello, Popl</strong></li><li><em>Apple/Google Wallet on the free plan</em> → <strong>Lynqu, Blinq</strong></li><li><em>Unlimited contacts on the free plan</em> → <strong>Lynqu, Blinq</strong></li></ul><h3>How I tested</h3><p>For two months, I:</p><ul><li>Created a complete card on every platform (real photo, real bio, real links)</li><li>Paid for the paid tier on each</li><li>Used each as my primary card for at least one in-person event or coffee meeting</li><li>Tested the recipient experience by handing my phone to non-technical friends and watching them try to save the contact (the most underrated benchmark in this category)</li><li>Verified pricing from each vendor’s pricing page in March 2026</li><li>Checked App Store and Google Play ratings the same week</li><li>Compared free-plan limits by actually hitting them</li></ul><p>Where Lynqu’s data appears, it’s from internal product analytics. Where competitor data appears, it’s from public sources (their pricing pages, their app store listings, their published case studies). I have not made up numbers and I have linked back to sources on Lynqu’s <a href="https://lynqu.com/compare/best-digital-business-cards">comparison pages</a> for anyone who wants to fact-check.</p><h3>FAQ</h3><p><strong>Are digital business cards actually replacing paper?</strong> In my experience, no — they’re supplementing it. Most professionals I talked to keep a small paper stack for specific situations (a partner who explicitly asks, an old-school client) and use digital for everything else. Paper is becoming the exception, not the default.</p><p><strong>Do recipients need to download an app?</strong> On all seven of these platforms, no. The recipient gets a web page with a “save contact” button. Some apps push the recipient harder toward downloading their app to save you back — that’s the worst pattern in the category, and Lynqu, Blinq, and HiHello all avoid it. Popl is the most aggressive about app prompts.</p><p><strong>Which one has the best free plan?</strong> Lynqu and Blinq, narrowly. Lynqu’s free plan includes wallet integration, unlimited contacts, and three privacy modes. Blinq’s free plan includes virtual backgrounds and email signatures. Both are usable as a permanent free tool. Popl, HiHello, and Wave Connect’s free plans are designed to push you toward paid faster.</p><p><strong>Which one is best for international teams?</strong> Lynqu is the only one of the seven with multi-language support out of the box (12 locales). The other six are English-only. If you have team members or clients outside the US/UK/AU, this is more decisive than people realize.</p><p><strong>What about the apps not on this list?</strong> There are at least a dozen smaller players (Beaconstac/Uniqode, DOT, Linq, Wave Cards, Boop, etc.) that I tested but didn’t rank. Most are clones of one of the seven above with worse execution. A few — Linq specifically — have genuinely nice physical card aesthetics but thin software. If a name keeps coming up in your research that I haven’t covered, ping me in the comments and I’ll give you a one-paragraph honest take.</p><h3>My honest take, as someone with skin in the game</h3><p>If you’re a solo professional, freelancer, or consultant in 2026 — pick <strong>Lynqu</strong> if you can stomach the bias of getting the recommendation from the founder. The all-in-one bundle (card + signature + virtual background + AI scan) genuinely saves you from buying three tools, and the privacy modes are the feature I miss most when I have to use a competitor for testing.</p><p>If you’d rather not give your business to the obviously biased recommendation, pick <strong>Blinq</strong>. It’s the safest choice for a reason. The card surface is shallower than Lynqu’s, but the brand recognition + maturity might matter more to you than the feature gap.</p><p>If you’re a sales team or live at events, <strong>Popl</strong> is the right answer regardless of what I’d prefer.</p><p>If you’re a brand that needs the physical card itself to be a marketing artifact, <strong>V1CE</strong>.</p><p>The other three (HiHello, Mobilo, Wave Connect) are best if you have specific requirements they uniquely solve. Otherwise, the four above will cover 90% of professionals.</p><blockquote><em>Amin Amjadi is the founder of </em><a href="https://lynqu.com/"><em>Lynqu</em></a><em>, an all-in-one digital business card platform built in Europe and available in 12 languages. He writes occasionally about building SaaS into “saturated” markets, what incumbents get wrong, and what it actually takes to ship a polished consumer-grade product as a small team.</em></blockquote><blockquote><em>If this comparison was useful, the longest-form version (with feature-by-feature tables, methodology details, and per-platform deep dives) lives at </em><a href="https://lynqu.com/compare/best-digital-business-cards"><em>lynqu.com/compare/best-digital-business-cards</em></a><em>. It’s free, no signup, no email gate.</em></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=21b33aa158c9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/the-digital-business-card-category-stopped-trying-heres-what-s-actually-worth-using-in-2026-21b33aa158c9">The Digital Business Card Category Stopped Trying. Here’s What’s Actually Worth Using in 2026.</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Measure Networking ROI: A Practical Framework | Lynqu Blog]]></title>
            <link>https://medium.com/gravisun/how-to-measure-networking-roi-a-practical-framework-lynqu-blog-95fbab37ad33?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/95fbab37ad33</guid>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Tue, 21 Apr 2026 05:30:00 GMT</pubDate>
            <atom:updated>2026-06-05T10:16:22.320Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1qhu4XI1LgA1HdG_bFoN2Q.png" /></figure><p>Almost every professional you ask will tell you that networking is one of the most valuable parts of their work. Almost no one can tell you, with any precision, what it has produced for them in the past year. The activity is real; the measurement is anecdotal. The result is a category of work that everyone agrees matters but no one feels confident defending when it gets cut.</p><p>The reason is partly cultural. Networking has always been a soft activity in industries that measure hard outputs. But it is also methodological: the right measurements are subtle, the wrong measurements are misleading, and most professionals have never been trained on either. Surveys of B2B sales organizations consistently find that only a minority track networking-sourced pipeline as a distinct category, despite networking being self-reported by the majority of senior reps as their highest-yield lead source.</p><p>This guide is a practical framework for measuring networking ROI-at the individual level, the team level, and the organizational level. It distinguishes between vanity metrics (numbers that feel productive but tell you little) and value metrics (numbers that actually inform decisions). And it gives you a workable system for capturing the data without building a new analytics function.</p><h3>Why Networking ROI Is Hard to Measure</h3><p>The standard ROI formula is straightforward: outcome value divided by cost. The difficulty in networking is that both the outcome and the cost are time-shifted, partially attributable, and often invisible.</p><h3>Time-Shifted Outcomes</h3><p>A relationship formed at a conference in March may produce a deal in October. The connection between the two is real, but in any monthly or quarterly view, the activity and the outcome show up as separate events. Most attribution systems are not patient enough to connect them, and most reps do not log the linkage themselves.</p><h3>Partial Attribution</h3><p>The deal that closes in October was probably influenced by the conference connection in March, but also by the LinkedIn post the buyer saw in June, the case study they read in August, and the introduction from a mutual connection in September. Naive attribution gives 100% credit to the last touchpoint. Sophisticated attribution gives weighted credit across the journey, but the data to do this well is rarely captured at the networking-event level.</p><h3>Invisible Outcomes</h3><p>Some networking outcomes are not deals at all. They are warm introductions to other prospects, hires made, partnerships formed, advice received, or reputation built. These have real economic value but rarely appear in any analytics system. They are usually mentioned anecdotally and forgotten when budgets get reviewed.</p><p>The combined effect is that the organizations and individuals who network most effectively often have the worst documentation of the value they produce. The activity speaks for itself in the rep’s own pipeline; it does not speak for itself in a CFO’s spreadsheet.</p><h3>The Three-Layer Measurement Framework</h3><p>A workable measurement system has three layers, each answering a different question. Most teams skip directly to layer three (pipeline value) and find it useless because they have not built layers one and two. The layers are sequential.</p><h3>Layer 1: Activity (What Did I Actually Do?)</h3><p>The most basic layer. Counts of networking activities undertaken. Examples:</p><ul><li>Number of contacts captured at events.</li><li>Number of new connections made via warm introduction.</li><li>Number of meaningful conversations beyond a 30-second exchange.</li><li>Number of LinkedIn connection requests sent and accepted.</li><li>Number of digital business card shares.</li></ul><p>Activity metrics are the easiest to capture and the easiest to misuse. They are necessary because without them you cannot tell whether the system is being used; they are insufficient because they tell you nothing about quality.</p><p>Treat activity metrics as a hygiene check, not a success metric. If activity is zero, something is wrong. If activity is high but layers two and three are flat, you are doing something busy but not useful.</p><h3>Layer 2: Engagement (Did Anything Happen After?)</h3><p>The middle layer. The proportion of activity that produced any subsequent meaningful action. Examples:</p><ul><li>Save rate: of contacts who received your card, what fraction added you to their phone or CRM.</li><li>Reply rate: of follow-ups you sent post-event, what fraction received a reply.</li><li>Meeting conversion: of new contacts, what fraction agreed to a follow-up meeting within 30 days.</li><li>Connection acceptance rate on LinkedIn after meeting in person.</li></ul><p>Engagement metrics tell you whether your activity is producing relationships rather than just exchanges. They are the layer where most of the actionable signal lives. A rep with 200 captured contacts and a 5% meeting conversion is producing 10 meetings; a rep with 100 captured contacts and a 25% meeting conversion is producing 25. The second rep is networking better, even though the first is networking more.</p><p>This is the layer where the difference between a good networker and a mediocre one becomes visible.</p><h3>Layer 3: Outcome (What Did the Relationships Produce?)</h3><p>The top layer. The downstream value created by the relationships that engagement produced. Examples:</p><ul><li>Pipeline value attributed to networking-sourced contacts.</li><li>Closed-won deals attributed to networking-sourced contacts.</li><li>Meetings sourced through warm introductions from existing contacts.</li><li>Hires made through the personal network.</li><li>Partnerships formed.</li><li>Inbound opportunities (calls, emails, referrals) sourced from people in the network.</li></ul><p>Outcome metrics are the most valuable but also the most time-shifted and most attribution-sensitive. They should be reviewed quarterly or annually rather than weekly. Trying to measure them in shorter cycles produces noise and discourages the patient work that produces the best outcomes.</p><p>The right pattern is: track activity weekly, engagement monthly, and outcomes quarterly. Each layer informs the one above it.</p><h3>The Specific Metrics Worth Tracking</h3><p>Across the three layers, here are the specific metrics that tend to produce the most insight relative to the effort of capturing them. Treat the benchmark ranges below as starting calibration points based on what consistently shows up across B2B teams-not absolute targets. Your industry, sales motion, and event mix will shift the right numbers up or down.</p><h3>Contacts Captured per Networking Hour</h3><p><strong>Definition:</strong> Total new contacts added to your system divided by hours spent in networking-eligible activities (events, conferences, planned coffees, intentional outreach).</p><p><strong>Why it matters:</strong> A simple efficiency metric. Most professionals dramatically overestimate how many contacts they make per hour of networking. The number is almost always lower than memory suggests, and the gap between perception and reality is itself useful feedback.</p><p><strong>Benchmark:</strong> 4 to 8 captured contacts per hour at a busy conference is good. 1 to 3 per hour at a typical networking event is normal. Below 1 per hour is a signal that the events you are attending are not networking-rich, regardless of how enjoyable they are.</p><h3>Save Rate</h3><p><strong>Definition:</strong> Of the people who received your contact information, the fraction who saved it (added you to their contacts, accepted a connection request, or otherwise indicated they wanted to maintain the relationship).</p><p><strong>Why it matters:</strong> Distinguishes between a contact you handed your card to and a contact you actually formed. A high save rate suggests your introduction is landing well; a low save rate suggests the contact is forgettable or your card is unappealing.</p><p><strong>Benchmark:</strong> 60% to 80% save rate is typical for warm meetings; 30% to 50% for incidental conference exchanges; below 20% suggests something is wrong with the introduction.</p><h3>Time to First Meaningful Reply</h3><p><strong>Definition:</strong> The median time between the initial contact (event introduction, cold outreach with a warm reference) and the first substantive reply from the contact.</p><p><strong>Why it matters:</strong> A sensitive indicator of how warm the relationship actually is. Cold contacts produce replies measured in days or weeks if at all. Warm contacts produce replies within hours. The metric correlates strongly with downstream conversion.</p><p><strong>Benchmark:</strong> Under 24 hours: very warm. 24 to 72 hours: warm. 72 hours to 7 days: lukewarm. Beyond 7 days: cold.</p><h3>Meeting Conversion Rate</h3><p><strong>Definition:</strong> Of new contacts captured, the fraction who agreed to a follow-up meeting within 30 days.</p><p><strong>Why it matters:</strong> The clearest signal that a contact is moving toward becoming a relationship rather than a stagnant entry in your CRM.</p><p><strong>Benchmark:</strong> 15% to 30% from cold conference networking is good. 40%+ from intentional, targeted networking with named accounts is good. Above 50% suggests the targeting is excellent or the bar for “contact” is too high (you are only counting people who were already going to meet you).</p><h3>Pipeline Attributed to Networking</h3><p><strong>Definition:</strong> Total open pipeline value where networking activity is the recorded source of the original contact.</p><p><strong>Why it matters:</strong> The single most credible metric for defending networking budgets to finance. Tied directly to opportunity value.</p><p><strong>Benchmark:</strong> For sales teams, networking should produce 20% to 40% of pipeline in B2B segments where personal relationships matter. Below 10% is a signal that either the team is not networking enough, or the attribution is broken.</p><h3>Closed-Won Attributed to Networking</h3><p><strong>Definition:</strong> Closed deals where networking is the recorded source of original contact, valued at the closed amount.</p><p><strong>Why it matters:</strong> The truest ROI metric. Lagging by 6 to 18 months from the original contact, depending on sales cycle length, but the most defensible measure of dollars produced.</p><p><strong>Benchmark:</strong> The closed-won attribution rate should approximately match the pipeline attribution rate. If pipeline attribution is 30% and closed-won is 5%, networking-sourced deals are converting worse than other sources-which is itself useful information.</p><h3>Building the Measurement System</h3><p>The metrics are useless if the data is not captured. Here is the practical capture system that tends to work.</p><h3>Capture at the Source</h3><p>The data needs to enter the system at the moment of contact, not days later. This is the single biggest reason most measurement systems fail: by the time the rep tries to log the contact, half the context has been forgotten and the entry is incomplete or never happens.</p><p>The fix is structural. The capture mechanism (a digital business card, an event app, a CRM mobile interface) has to be in the rep’s pocket and easier to use than the alternative. <a href="https://lynqu.com/business/lead-capture">Lynqu’s lead-capture flow</a> records contact details into a structured record at the moment of share, with the event and source automatically tagged. The rep does not have to remember to log anything; the structure is created by default. (For sales-team-specific use cases, see <a href="https://lynqu.com/blog/sales-teams-digital-business-cards">how sales teams close more deals with digital business cards</a>.)</p><h3>Tag the Source Automatically</h3><p>Every captured contact should know where it came from. “Conference XYZ 2026,” “LinkedIn DM,” “Warm intro from Alice,” “Email signature click.” This tagging is what makes attribution possible later. UTM parameters on QR codes and tracking links produce the source data automatically; manual tagging on rep entries fills in the gaps.</p><h3>Sync to the CRM</h3><p>The captured contact has to flow into the system where the rest of the pipeline lives. A contact that exists only in the rep’s digital card and never reaches Salesforce or HubSpot will not appear in any pipeline analysis, regardless of how excellent the original capture was.</p><h3>Add Light Annotation</h3><p>One field worth adding to every captured contact: a one-sentence note about the conversation. “Discussed Q3 expansion in EMEA, send case study X.” The note is the difference between a contact you remember and a contact you have to reconstruct. It also serves as the basis for follow-up content; reps with strong note-taking habits convert at noticeably higher rates than reps without.</p><h3>Review Quarterly, Not Weekly</h3><p>The whole system should be reviewed at the team level once a quarter, not more often. Activity is reviewed weekly because the cycle is short. Engagement is reviewed monthly because conversations move on month-long timeframes. Outcomes are reviewed quarterly because that is the cycle on which deals close. Trying to do all of this weekly produces noise and discourages the patient activities that drive long-term results.</p><h3>What Vanity Metrics Look Like</h3><p>Some of the most commonly reported networking metrics are vanity metrics. Recognizing them is half the battle.</p><h3>Total LinkedIn Connections</h3><p>The count of connections in a network tells you almost nothing about that network’s value. A network of 5,000 mostly-irrelevant connections is worse than a network of 500 closely-aligned ones. The metric optimizes the wrong thing: volume rather than relevance.</p><h3>Number of Events Attended</h3><p>Attendance is presence; presence is not productivity. The metric should always be paired with engagement and outcome data. Otherwise “attended 12 conferences” sounds productive while being meaningless.</p><h3>Number of Cards Handed Out</h3><p>This was always a vanity metric but is even more so in the digital age, where the marginal cost of sharing is zero. A rep who shared their card 500 times last quarter has done very little if 480 of those shares produced no save and no follow-up. The metric to watch is the save rate, not the share count.</p><h3>Cost per Contact</h3><p>Almost always a misleading metric in isolation. Some of the most expensive contacts (named-account meetings at exclusive events) produce 100x the value of cheaper contacts (random conference exchanges). Cost should be evaluated against the eventual outcome, not against the count of contacts produced.</p><h3>Common Mistakes</h3><h3>Measuring Only What Is Easy to Measure</h3><p>Activity metrics are easy. Outcome metrics are hard. The temptation is to focus on what the system can produce automatically, which leads to a system that optimizes for measurable activity instead of valuable activity. Resist this. The metric you most want is usually one of the harder ones.</p><h3>Single-Touch Attribution</h3><p>Treating the source of the original contact as 100% responsible for the eventual deal ignores everything that happened in between. Use multi-touch attribution where possible, or at minimum acknowledge that the network gets credit for an opportunity that may have been closed by other touchpoints.</p><h3>Confusing Lag With Failure</h3><p>A relationship formed at a conference in March may not produce visible value until October. A monthly review that does not see immediate output may conclude the conference was a waste, when in fact the value is still developing. Always look at networking outcomes on at least a 6-month rolling window.</p><h3>Forgetting the Non-Pipeline Outcomes</h3><p>Even the most rigorous pipeline-based attribution misses warm introductions, hires, partnerships, advice, and reputation effects. A complete networking ROI view includes these as separate outcomes, even if they are not dollar-quantified. Otherwise the people who network most effectively are systematically undervalued.</p><h3>Letting Perfect Be the Enemy of Good</h3><p>The ideal measurement system has multi-touch attribution, six months of trend data, manually annotated context on every contact, and weekly review. Most organizations cannot build this immediately. A simpler version-activity, save rate, meeting conversion, and pipeline attribution-captured consistently, produces 80% of the value with 20% of the effort. Start there.</p><h3>The Individual vs Team View</h3><p>The metrics above apply to both individual contributors and teams, but the diagnostic questions differ.</p><p><strong>For individuals:</strong> Are my activity, engagement, and outcome metrics consistent over time, or trending in a particular direction? Where is my conversion lowest, and what could I change about my approach? Which of my contacts have produced multiple downstream introductions, and what made those relationships different?</p><p><strong>For teams:</strong> Which reps consistently outperform on engagement metrics regardless of activity volume? What are they doing that others can learn from? Which events produce the highest yield per dollar invested, and which are vanity attendance? How does networking attribution compare to other lead sources at similar maturity?</p><p>The team view enables resource decisions: where to invest event budget, who to send, what to coach. The individual view enables professional development: where each rep is strongest and weakest, what specific behaviors correlate with their best outcomes.</p><h3>The 90-Day Implementation Plan</h3><p>If you are starting from no measurement at all, here is a sequence that produces a working system within a quarter.</p><p><strong>Days 1 to 14:</strong> Audit current state. What is captured today, where, by whom. Identify the gap between capture and CRM. Set up source tagging on the <a href="https://lynqu.com/business/analytics-crm">analytics and CRM integration</a> if it is not already in place.</p><p><strong>Days 15 to 30:</strong> Define the four to six metrics you will track. Resist the urge to track more. Establish baselines for each metric using whatever historical data exists.</p><p><strong>Days 31 to 60:</strong> Run the system. Capture contacts at every event with full source tagging. Sync to CRM. Add the one-sentence note discipline. Have reps review their own activity weekly.</p><p><strong>Days 61 to 90:</strong> First quarterly review. Compare actual numbers against baselines. Identify the metric that needs the most improvement and design one focused experiment to address it.</p><p>By day 90, you have a working measurement system, real data, one targeted improvement experiment in flight, and a clear story to tell finance about what networking is producing. For event-heavy teams specifically, the <a href="https://lynqu.com/blog/conference-networking-guide-2026">conference networking guide</a> covers the capture-and-attribution rituals that produce the cleanest data.</p><h3>What This Earns You</h3><p>Networking ROI is not unmeasurable; it is undermeasured. The combination of activity, engagement, and outcome metrics, captured at the source and reviewed on the right cycle, produces a defensible picture of what networking actually contributes.</p><p>The professionals and teams who measure consistently tend to also network consistently, and the two reinforce each other. Once you can see the numbers, the bias toward investment in the activity that produces them follows naturally. Once the activity is consistent, the numbers improve.</p><p>Start with the simplest possible system. Capture the right data at the moment of contact. Review it on the right cycle. The compounding return is not just in the deals that close-it is in the institutional clarity about which networking activities are worth doing more of, and which ones are just busy. That clarity is itself one of the most valuable outputs any measurement system can produce.</p><p><em>Originally published at </em><a href="https://lynqu.com/blog/networking-roi-measurement"><em>https://lynqu.com</em></a><em> on April 21, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=95fbab37ad33" width="1" height="1" alt=""><hr><p><a href="https://medium.com/gravisun/how-to-measure-networking-roi-a-practical-framework-lynqu-blog-95fbab37ad33">How to Measure Networking ROI: A Practical Framework | Lynqu Blog</a> was originally published in <a href="https://medium.com/gravisun">Gravisun</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Deep Dive Into Google’s ML Kit and Apple’s Core ML: Which one to use?]]></title>
            <link>https://medium.com/good-audience/deep-dive-into-googles-ml-kit-and-apple-s-core-ml-which-one-to-use-fe8540f8cb63?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe8540f8cb63</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[coreml]]></category>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Mon, 30 Jul 2018 22:48:12 GMT</pubDate>
            <atom:updated>2018-07-31T02:53:11.246Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qsIxQl5qR1MXUtYho-aijg.jpeg" /></figure><p>If you decided to integrate machine learning into your new mobile app and you don’t know which service to use? or you just simply want to have more knowledge over the new Apple’s Core ML and Google’s ML Kit then this article is perfect for you.</p><p>This article is the second part of the <strong>“Which one to use?”</strong> series. In <a href="https://blog.goodaudience.com/online-vs-offline-machine-learning-which-one-to-use-934801e0a418">Previous article</a> we talked about how machine learning works and what are the possible approaches to start with machine learning. In this article we want to focus more on the recently introduced SDKs from Google and Apple.</p><p>Recently Google and Apple introduced their SDKs which allow developers to integrate machine learning into their apps easier than ever. Their attempt is to address the challenges mobile developers faced every day when trying to leverage deep learning models from mobile devices. As mobile apps have become the great contributor of data point, machine learning has become critical to mobile development. So it is more essential than ever for such a services to exist to let all range of developers (whether an experience ML developer or an amateur person) to be able to start with no problem.</p><p>Today, the way most mobile apps leverage deep learning capabilities is by exclusively calling centralized APIs hosted on a cloud platform. While that model is relatively simpler from a developer standpoint, it has some tangible limitations which we have discussed closely on the <a href="https://blog.goodaudience.com/online-vs-offline-machine-learning-which-one-to-use-934801e0a418">previous article</a>, but some of the most notable limitation are:</p><ol><li><strong>Latency: </strong>Executing computational intensive deep learning models imposes a latency penalty that sometimes is prohibited for mobile apps. For instance if your app should handle some kind of realtime realtime predictions on the big data types, it would be impossible to use a centralized API on a cloud.</li><li><strong>Long Training Cycles:</strong> Mobile devices are great collectors of data that can be used on the training of deep learning models. However, the lifecycle of collecting data from a mobile app, training a deep learning model on a cloud platform and using the model from the mobile application is both computationally expensive and relatively long.</li><li><strong>Mobile Adaptability: </strong>Most deep learning runtimes have been natively designed to produce models that require large hardware infrastructures to execute. As a result, adapting many of the existing deep learning models to mobile applications remains incredibly challenging.</li></ol><p>The starting sentences of “<a href="https://www.youtube.com/watch?v=ejrn_JHksws">Google, Introducing ML Kit</a>” are:</p><blockquote>In today’s fast-moving world, consumers have come to expect mobile apps to not only be intuitive, but also to be able to provide powerful features and adapt to new information. That’s why machine learning has become critical to mobile development. developer’s are increasingly relaying on machine learning to enhance their app’s user experience and only with <strong>fine-tuned machine learning models</strong> can they deliver those powerful features to delight users…</blockquote><p><strong>Fine-tuned machine learning models</strong>: In some cases Apple’s Core ML might be the service which offers those fine-tuned machine learning models that you need and in other cases Google’s ML Kit, depending on your app and use case scenario.</p><h3>Core ML</h3><p>Core ML was first introduced at WWDC conference in June 2017 in the form of a framework with the launch of iOS 11 to help the developers and make it possible to for them to integrate trained machine learning models into their iOS, macOs and tvOS apps easier than every. Trained model are loaded into Apple’s Xcode IDE and packaged in the app bundle.</p><p>Core ML supports:</p><ul><li><a href="https://developer.apple.com/documentation/vision">Vision</a> for image analysis</li><li><a href="https://developer.apple.com/documentation/naturallanguage">Natural Language</a> for natural language processing</li><li><a href="https://developer.apple.com/documentation/gameplaykit">GameplayKit</a> for evaluating learned decision trees</li></ul><p>You should note that Core ML only works with Apple ecosystem and it cannot be used for Android apps.</p><p>Unlike Google’s ML Kit, <strong>Core ML is running strictly on-device</strong> (offline) to ensures the privacy of user’s data which has its own <a href="https://blog.goodaudience.com/online-vs-offline-machine-learning-which-one-to-use-934801e0a418">benefits and drawbacks</a>. The main reasons why Apple chose on-device approach over in cloud is:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2pqFvB-j6PSBq21aFOReAg.png" /><figcaption>Picture from <a href="https://devstreaming-cdn.apple.com/videos/wwdc/2018/708ax570wjtpxze/708/708_whats_new_in_core_ml_part_1.pdf?dl=1">Apple’s presentation slides</a></figcaption></figure><p>A year later, in WWDC 2018 Apple introduced Core ML 2 . During the coference Apple said:</p><blockquote>“Core ML 2 is 30 percent faster thanks to batch prediction, and it can compress machine learning models by up to 75 percent with the help of quantisation.”</blockquote><p>The new version of Core ML can update models from a cloud service like Amazon Web Services or Microsoft Azure at runtime. It also comes with a converter that works with Caffe, Caffe 2, Keras, sci-kit-learn, XGBoost, LibSVM and TensorFlow Lite.</p><p>And they also introduced Create ML. It’s a new GPU-accelerated tool for native AI model training on Mac computers. Since it’s coded in swift, developers can use drag-and-drop programing interfaces like Xcode Playground to train models. The trained models can perform tasks like recognizing images, extracting meaning from text or finding relationships between numerical values.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*6wMICtfMnW-I6tyQs7rJIA.png" /><figcaption>Picture from <a href="https://developer.apple.com/documentation/create_ml">Apple’s documentation of Create ML</a></figcaption></figure><p>Apple says the developers of Memrise, a language-learning app, previously took 24 hours to train a model using 20,000 images. CreateML and Core ML 2 reduced it to 48 minutes on a MacBook Pro and 18 minutes on an iMac Pro. Furthermore, the size of the model was reduced from 90MB to just 3MB.</p><p>But the question is, how do these really work?</p><h3>Create ML</h3><p>let’s say you want to create the model that can recognizes the certain type of fruits in a picture like banana, strawberry and orange.</p><p>So the question for the picture below would be like,<strong> is there any banana?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*ekiawPMQ-yO6A6gZEV5u-w.png" /></figure><p>~ <strong>2007</strong></p><p>Hope computers can answer some day, but for now — just ask someone.</p><p>~ <strong>2012</strong></p><p>Wow — Computers <em>can </em>answer this! Cutting-edge deep learning research is exciting.</p><p><strong>2018</strong></p><p>Train a model with Create ML — the UI is great!</p><p>So how to start? Very simple. we just need to prepare your required data, in this case different pictures of bananas, strawberries and oranges and group them each in a separated folder in the way that all the pictures of bananas should be in banana folder and so on. After that we need to put all of our folders in one folder and we can call it “Training Data”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/538/1*z7OGnay_KKpK9Hcj8ZESEA.png" /></figure><p>Then we need to open Xcode and then File/New/Playground and then choose we choose macOS and we can start with the blank playground and set the name of our file and Create.</p><pre><strong>import</strong> CreateMLUI</pre><pre><strong>var</strong> builder = MLImageClassifierBuilder()<br>builder.showInLiveView()</pre><p>Now if we run the code we get this view:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7VmEufrCVBlq53AoiRg_3A.png" /></figure><p>and after that we can drag and drop our “Training Data” folder to specified area and in matter of seconds we will have our trained model created. Then to evaluate our model we can create another folder with different set of pictures of the same fruits in same way as before and test our model before saving it to check if you are satisfied with the results and it is accurate enough.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dzsX1U1XmwosNB_vcnIahg.png" /></figure><h3>Core ML 2</h3><p>This year, Apple was focused on 3 main points for helping Core ML developers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6TKWfIiq40PctDoLf0XiuA.png" /></figure><p>So now let’s explore there three points!</p><h4>Model Size</h4><p>One of the main things about Core ML as we have discussed previously is that it is on-device and totally offline. So the way it works is, the app should have the models in it’s bundle so it would be able to use them for predictions. But one of the biggest drawbacks with this method is <strong>increase in app size</strong>.</p><p>So Imagine you have an app that works perfectly and all the customers are happy and then you decide to integrate machine learning to your app to increase the satisfaction of your customers. But sometimes this implementation means increase of 100MB in size of your app or if you want to have more accurate and better machine learning sometimes even +250MB.</p><p>So with Core ML 2 Apple decided to give developers the tools to <strong>quantize</strong> their Core ML models. Quantization is techniques used to store and calculate numbers in a more compact form. This can lead to a reduced runtime memory usage and faster calculations.</p><p>There are 3 main factors when it comes to Core ML app size:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZxAHEj0cG9YuQOsOJtrQVA.png" /><figcaption>Picture from <a href="https://devstreaming-cdn.apple.com/videos/wwdc/2018/708ax570wjtpxze/708/708_whats_new_in_core_ml_part_1.pdf?dl=1">Apple’s presentation slides</a></figcaption></figure><p>Quantization for models is done by reducing the size of the weight. In Core ML, the models were stored in 32-bit models. But <strong>Core ML 2 now supports 16-bit Floating Point and all level of quantization including down to 1-bit</strong> which can greatly reduce the size of the models.</p><blockquote>In case you aren’t familiar with what weights are, here’s a really good analogy. Say that you’re going from your house to the supermarket. The first time, you may take a certain path. The second time, you’ll try to find a shorter path to the supermarket, since you already know your way to the market. And the third time, you’ll take an even shorter route because you have the knowledge of the previous 2 paths. Each time you go to the market, you’ll keep taking a shorter path as you learn over time! This knowledge of knowing which route to take is known as the weights. Hence, the most accurate path, is the one with the most weights!</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vrp55b9hQ7E6mZHlMWMBHQ.png" /></figure><p>But we should have in-mind that this reduction in size doesn’t come without tradeoffs. When we quantize our model it means in simple words that the new quantized model is approximation of the original one which means it won’t be as accurate as the original model. But it’s not a bad thing since it totally depends on our model and how these approximations effect our over all result so for example let’s look at the picture below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bj9aLSrTvSZ0JMulkK-NFA.png" /></figure><p>In this picture we can see that 8-bit version of the model is giving the same output as for the 32-bit version even-though the size decreased from 6.7MB to 1.7MB and we can see the output starts to change from 4-bit which might be still acceptable.</p><h4>Performance</h4><p>Since we are running the ML computations on-device, we want it to be fast and accurate.</p><p>Let’s consider a realtime style transferrer app that can transform a certain image into different styls.</p><p>If we were to look into the neural network of style transfer, this is what we would notice. There are a certain set of inputs for this algorithm. Each layer in the neural network adds a certain transformation to the original image. This means that the model has to take in every input and map it out to an output and make a prediction from that. The prediction then helps in the creation of the weights. What would this look like in code?</p><pre><em>// Loop over inputs</em></pre><pre><strong>for</strong> i <strong>in</strong> 0..&lt; modelInputs.count {<br>modelOutputs[i] = model.prediction(from: modelInputs[i], options: options)<br>}</pre><p>In the above code, you see that for each input, we are asking the model to generate a prediction . However, iterating over every input can take a long time.</p><p>To solve this problem, Apple has introduced a <strong>new API called Batch</strong>. Unlike a <em>for</em> loop, a batch in machine learning is when all inputs are fed to the model as soon as it can be don’t and it won’t need to wait like it used to in a <em>for</em> loop method, so it loops to the next input after first input is done. This can take much less time and also less code.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*M_sy-NpSXfC-IFupNKl_1A.png" /><figcaption>For Loop Vs. Batch</figcaption></figure><p>And if we write the above <em>for</em> loop code with the new Batch API, it will be like:</p><pre>modelOutputs = model.prediction(from: modelInputs, options: options)</pre><h4>Customization</h4><p>When it comes to neural network, there are consist of many layers and there might be cases in which we want to convert a neural network from some service like Tensorflow or Keras to Core ML. Even though now that Core ML comes with a converter that works with many different services, there might be a scenario where Core ML simply does not have the tools to convert the model correctly. Don’t worry if don’t understand this, let’s take a look at the example below to see what it means.</p><p>So for example imagine you have a model using a Convolutional Neural Network (CNN). CNNs consists of series of layers which are highly optimized and when you convert a neural network from one format to Core ML, you are transforming each of those layers. But there might a scenario where Core ML does not provide the tools to convert a layer. Before Core ML 2, you could not do anything about it but now in this version of Core ML apple engineers have introduced the<strong> </strong><strong>MLCustomLayer</strong>protocol which solves this issue and allows developers to create their own layers in Swift..</p><p>This topic is very complicated and it usually takes a skilled data scientists or a specialist to deal with these things. This is beyond the scope of the tutorial so we won’t delve into it.</p><h3>ML Kit</h3><p>ML Kit was first introduced at Google I/O in May 2018 in the form of a framework and it is Google’s answer to bringing deep learning universe to Android and iOS developers. With ML Kit, Google aims to simplify the execution of deep learning models in mobile applications and it includes a-bunch of ready to use deep learning APIs that that are divided into 5 categories:</p><ul><li><a href="https://firebase.google.com/docs/ml-kit/label-images">Image Labeling</a></li><li><a href="https://firebase.google.com/docs/ml-kit/recognize-text">Text Recognition</a></li><li><a href="https://firebase.google.com/docs/ml-kit/detect-faces">Face Detection</a></li><li><a href="https://firebase.google.com/docs/ml-kit/read-barcodes">Barcode Scanning</a></li><li><a href="https://firebase.google.com/docs/ml-kit/recognize-landmarks">Landmark Recognition</a></li></ul><p>Google is also working on Smart reply that they said will be later.</p><p>One of the main and biggest advantages with ML Kit would be the fact that there are many options and it can be used anywhere. Unlike Apple, it is cross-platform and can be used for both iOS and Android. Also with ML Kit you can either call the APIs from the cloud or perform the calls all locally inside the device (offline).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*siJEVUhVeFpZ_VVNK-lUtg.png" /></figure><p>At the conference, Google introduced ML Kit with these four main factors that can be seen in the picture above. So now let’s talk about each and look at them one by one.</p><h4>iOS and Android SDK</h4><p>ML Kit uses the standard Neural Networks API on Android and Metal on iOS. And now let’s see how to start</p><ul><li>On iOS:</li></ul><p>First we need to include the ML Kit libraries in the app’s Podfile:</p><pre>pod &#39;Firebase/Core&#39;<br>pod &#39;Firebase/MLVision&#39;<br>pod &#39;Firebase/MLVisionLabelModel&#39; # for on-device model</pre><p>Then for instance for on-device image labeling we can call the APIs like:</p><pre>vision = Vision.vision()<br>let labelDetector = vision?.labelDetector()</pre><pre>// set up image metadata...<br>let image = VisionImage(image: uiImage)</pre><pre>labelDetector?.detect(in: image) { (labels, error) in<br>   guard error == nil, let labels = labels, !labels.isEmpty else {<br>      //...<br>      return<br>   }<br>   // handle extracted entities<br>}</pre><ul><li>On Android:</li></ul><p>First we need to include the ML Kit libraries in the app’s build.gradle:</p><pre>dependencies {<br>  // ...<br>  implementation &#39;com.google.firebase:firebase-core:15.+&#39;<br>  implementation &#39;com.google.firebase:firebase-ml-vision:16.+&#39;<br>  implementation &#39;com.google.firebase:firebase-ml-vision-image-label-model:15.+&#39;<br>}</pre><p>Then for instance for on-device image labeling we can call the APIs like:</p><pre><strong>FirebaseVisionLabelDetector</strong> detector = FirebaseVision.getInstance().getVisionLabelDetector();</pre><pre>FirebaseVisionImage image = FirebaseVisionImage.formFilePath(context, uri);</pre><pre>Task&lt;List&lt;<strong>FirebaseVisionLabel</strong>&gt;&gt; result = <br>    detector.detectInImage(image)<br>    .addOnSuccessListener(<br>        new OnSuccessListener&lt;List&lt;<strong>FirebaseVisionLabel</strong>&gt;&gt;() {<br>            @Override<br>            public void onSuccess(List&lt;<strong>FirebaseVisionLabel</strong>&gt; labels) {       //handle extracted entities      }<br>        })<br>    .addOnFailureListener(...);</pre><p>The above code is for on-device. But what if we want to call the APIs from the cloud? No problem, easily we just need to change the parts which are bold in the above code and the rest is the same:</p><pre><strong>FirebaseVisionCloudLabelDetector</strong> detector = FirebaseVision.getInstance().getVisionLabelDetector();</pre><pre>FirebaseVisionImage image = FirebaseVisionImage.formFilePath(context, uri);</pre><pre>Task&lt;List&lt;<strong>FirebaseVisionCloudLabel</strong>&gt;&gt; result = <br>    detector.detectInImage(image)<br>    .addOnSuccessListener(<br>        new OnSuccessListener&lt;List&lt;<strong>FirebaseVisionCloudLabel</strong>&gt;&gt;() {<br>            @Override<br>            public void onSuccess(List&lt;<strong>FirebaseVisionCloudLabel</strong>&gt; labels) {       //handle extracted entities      }<br>        })<br>    .addOnFailureListener(...);</pre><h4>Base APIs and custom model support</h4><p>Now, what about custom models? If ML Kit base APIs don’t cover you use cases, it would be possible to bring your own existing <strong>TensorFlow Lite models</strong>. It is just enough to upload the TensorFlow Lite models to the service. Then it acts as an API layer to your custom model.</p><p>Also in addition to deploying your own custom model, Google is releasing an experimental model compression flow that can helps to reduce model size. To have early access to this experimental feature you can sign up now in <a href="https://docs.google.com/forms/d/e/1FAIpQLSd7Uzx6eepXeF5osByifFsBT_L3BJOymIEjG9uz1wa51Fl9dA/viewform">here</a>.</p><h4>On-device and Google Cloud AI interface APIs</h4><p>As we discussed before with ML Kit, it is possible to perform the calls both locally on-device (offline) and from the cloud (online). Unsurprisingly, <strong>there is a trade-off here</strong>. The models that run on the device are smaller and hence offer a lower level of accuracy but if you decide to use the models from the cloud, neither model size nor available compute power are an issue therefore though models are larger and hence more accurate.</p><p>(you can read about the benefits and trade-offs of the two method to see which one is more suitable for your use case from <a href="https://blog.goodaudience.com/online-vs-offline-machine-learning-which-one-to-use-934801e0a418">this article</a>)</p><p>So, For instance in the case of image labeling, while the <strong>on-device model features about 400 labels, the cloud-based one features more than 10000</strong>.</p><p>As an example, Google showed the case below where we can see the clear difference between the accuracy of the two methods:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S-vrmlRcvrYWZjPq548DOg.png" /></figure><h4>Deeply integrated into Firebase</h4><p>ML Kit is introduced as a part of Firebase cloud services and one of the main advantages is that it allows developers to easily start developing their apps. Also since it is a part of Firebase services, it <strong>seamlessly works with other firebase services</strong>.</p><h3>Conclusion</h3><p>Choosing between Core ML 2 and ML Kit mostly depends on your preference and use case scenario.</p><p>Core ML 2 doesn’t support Android and it only runs on-device (offline) and in the other hand, ML Kit is still in beta and it is not as developed and shaped as Core ML 2.</p><p>Developers familiar with Google’s Firebase are more likely to prefer ML Kit. Likewise, longtime Xcode users will probably tend toward Core ML 2.</p><h3>References:</h3><ul><li><a href="https://medium.com/google-cloud/technology-fridays-deep-dive-into-google-ml-kit-dc81624d093e">Medium.com</a></li><li><a href="https://techcrunch.com/2018/05/08/googles-ml-kit-makes-it-easy-to-add-ai-smart-to-ios-and-android-apps/?guccounter=1">Techcrunch.com</a></li><li><a href="https://venturebeat.com/2018/06/05/apples-core-ml-2-vs-googles-ml-kit-whats-the-difference/">Venturebeat.com</a></li><li><a href="https://www.appcoda.com/coreml2/">Appcoda.com</a></li><li><a href="https://developers.google.com/ml-kit/">developers.google.com</a></li><li><a href="https://developer.apple.com/documentation/coreml">developer.apple.com</a></li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2F7b0181%3Fas_embed%3Dtrue&amp;dntp=1&amp;url=https%3A%2F%2Fupscri.be%2F7b0181%2F&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/748a06e9019942f2677bba917ef55804/href">https://medium.com/media/748a06e9019942f2677bba917ef55804/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fe8540f8cb63" width="1" height="1" alt=""><hr><p><a href="https://medium.com/good-audience/deep-dive-into-googles-ml-kit-and-apple-s-core-ml-which-one-to-use-fe8540f8cb63">Deep Dive Into Google’s ML Kit and Apple’s Core ML: Which one to use?</a> was originally published in <a href="https://medium.com/good-audience">Good Audience</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Online vs. Offline Machine learning: Which one to use?]]></title>
            <link>https://medium.com/good-audience/online-vs-offline-machine-learning-which-one-to-use-934801e0a418?source=rss-88281f25dc2b------2</link>
            <guid isPermaLink="false">https://medium.com/p/934801e0a418</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[neural-networks]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Amin Amjadi]]></dc:creator>
            <pubDate>Sat, 07 Jul 2018 23:01:35 GMT</pubDate>
            <atom:updated>2018-07-11T18:10:40.309Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FricT0qL99gpk7ne3aNV7A.png" /></figure><p>So you have decided to implement machine learning into your app and you want to know all of your options? or you just want to learn more about all the different approaches? then this article is made for you.</p><p>If you are reading this article probably you know or have some ideas of what exactly is machine learning but let’s review it, shall we?</p><p>Machine learning has many definitions but the most accepted and occurate one is:</p><blockquote>Machine learning is the ability for computers to use <strong>statistical techniques</strong> in order to “learn” with <strong>data</strong>, <strong>without being explicitly programmed</strong>.</blockquote><p>This definition has 3 main parts: Statistical techniques, Data and not being explicitly programmed. But what do these really mean?</p><p>Well, based on the definition of machine learning for most solutions, they combined of 3 main steps:</p><ol><li>Providing the <strong>Data</strong></li><li><strong>Training a model</strong> from the provided data</li><li>Using the trained model to <strong>make predictions</strong></li></ol><p>let’s say an example to make this more understandable. For instance you decided to create an app that can recognize different types of flowers such as Aconitum, Anthurium, Narcissus… etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/225/1*p9n9dXl5lJP4RNPO_OJsyA.png" /><figcaption>Narcissus</figcaption></figure><p>For that, first you need to gather a data set including the pictures of all the different flowers that the app should recognize.</p><p>Then in the next step you need to create a model with a specific purpose of recognizing the types of the flowers and gives the exact type of information that app requires.</p><p>At last with the information provided by the model the app can predict the answer.</p><p>They are many important decisive factors when it comes to implementing the machine learning: <strong>How accurate it should be? What is the size of the data set needed? Should it require network connectivity? What sort of model you need?… etc</strong>. and then based on your app decide if you want to:</p><ol><li><strong>Use an existing model</strong> or training your own?</li><li><strong>Train on your own computer or use the cloud</strong>?</li><li>Access the trained model and <strong>do the predictions locally on device (offline) or in the cloud (online)</strong>?</li></ol><p>So now let’s start comparing the different approaches in order of the questions above, one by one.</p><h3><strong>Use an existing model</strong></h3><p>One of the main questions is <strong>whether you need to train your own model or that is not necessarily compulsory for you</strong> and you can use one that is already exist.</p><h4>Quickest and easiest way to start</h4><p>If you don’t need to have your own model then one of the quickest and easiest ways is to use someone else’s model that is already trained and ready to use. (There is another easier and cheaper way to use offline machine learning SDK like Core ML and ML Kit which we will discuss in details later)</p><p>There are many companies that are providing machine learning packages that are designed for a specific purpose such as image classification, text analyses, speech recognition… etc. that you can make use of.</p><p>Some of the most popular service providers are:</p><ul><li><a href="https://www.clarifai.com/">clarifai</a></li><li><a href="https://cloud.google.com/vision/">Google cloud vision API</a>, <a href="https://firebase.google.com/products/ml-kit/?gclid=CjwKCAjwg_fZBRAoEiwAppvp-V1XP3oJr_NbAuC0WrHoj4VwshY8oKPUQDSuxykM7oWG04b-n5GQABoCbZMQAvD_BwE">ML Kit for Firebase(beta)</a></li><li><a href="https://azure.microsoft.com/en-us/services/cognitive-services/">Microsoft Azure cognitive services</a></li><li><a href="https://aws.amazon.com/rekognition/">AWS Rekognition</a>, <a href="https://aws.amazon.com/lex/">Lex</a>, <a href="https://aws.amazon.com/polly/">Polly</a></li><li><a href="https://www.ibm.com/watson/">IBM Watson</a></li></ul><p>If these services provide one of the specific tasks that your app suppose to do, they definitely worth to take the look at and to be considered.</p><p><em>You should note that by using these services you cannot access their models directly and you can only use them through an API.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fG4twtCgTCHQ1yw-sQgKCA.png" /><figcaption>architecture of how fully managed machine learning works</figcaption></figure><p><strong>How to work with this? </strong>Your app simply just need to send an HTTPS request to the web service with the required data, like the picture of the flower in our example and then in the matter of seconds the service replies with the requested information.</p><p><strong>How to keep the model up-to-date?</strong> One of the biggest advantages of using a web service is that you don’t need to be worried about the model and deal with all the headaches that will come with re-training your model or updating it as it is all <strong>being done automatically by the web service itself</strong>. They are constantly updating their models therefore the best part of it is that you don’t have to know anything about machine learning in order to use these services.</p><p><strong>Do I have to pay?</strong> Of course, usually with these services <strong>you need to pay per request</strong>. Most of these services provide an SDK which makes it really easy for you to do the call to service’s API endpoint.</p><p><strong>What do I need to know to start?</strong> nothing, that’s the best thing about it. You don’t need to have any kind of background in machine learning and you can simply just call the API and get the information you are looking for. <strong>You just need to read the documentation of the web service you choose</strong> to know how to call the API end points.</p><p><strong>Advantages:</strong></p><ul><li><strong>It is very simple and easy to use</strong> and you don’t need to have any kind of background in machine learning.</li><li>You don’t need to deal with all the headaches that will come with how to re-train or update your model and <strong>it is done all automatically</strong>.</li></ul><p><strong>Disadvantages:</strong></p><ul><li><strong>You need to pay for the service</strong>, something around 1$ per 1000 requests.</li><li>These services only work with the most common data such as pictures, videos and speech therefore if your app uses some kind of unique data type then it won’t be the best to use this option.</li><li>You cannot have direct access to the trained data model. what it means is that if you decide to inference locally on the device, you cannot.</li><li><strong>There is a small delay</strong> when you send a request until you get the result.</li><li>Your app <strong>won’t work without the network connectivity</strong>.</li></ul><p><em>You should note that with some of these services you are actually allowed to do a bit of training. For instance with Clarifai you can create your own model by uploading custom training images. Also with ML Kit for firebase(in beta) you can upload your own existing model by TensorFLow Lit.</em></p><h3><strong>Train on your own computer or use the cloud?</strong></h3><p><em>(if you already have your training model and don’t need to create one, skip this part.)</em></p><h4>Training your model</h4><p>Training is usually a really difficult and expensive process to do and it is only worth to do and make sense for big applications and businesses. For small applications it usually more often to use someone else’s model which we discussed in previous part. But if your data is unique or the existing solutions doesn’t satisfy you then you will need to train your own data model.</p><p>The accuracy and success rate of your training model depends on the amount of data you provide for your training model. Basically in most cases as much more data you provide the chance that the training model makes the right prediction increases more.</p><p>Also it depends on your app and the range of your possible predictions and that how important it is to have the right answer all the time in a cost of much more work.</p><p>Usually training the model with the provided data set is done in three ways: in <strong>your own computer</strong>, <strong>in the cloud </strong>or<strong> on the device itself</strong>.</p><p><strong>When to use my own computer and when the cloud?</strong> Well it totally depends on the amount of data provided.</p><ul><li>Small data set: when you are having a small data set you can train your model on your own computer or any other machine that you have.</li><li>large data set: as it requires much more power, you need a huge machine with multiple GPUs and it is really a job of for a high-performance computer.</li></ul><p>Providing that high-performance machines to do such a work is usually really expensive and out of reach for most developers unless you have your own data center. For that reason for most cases it would make more sense to just rent a computer power as they are many cloud platform that are happy to do so. But in a long run if you re-train your model often or the data changes alot then for big businesses it actually might be better to provide a data center.</p><p>At the end it all depends on your situation to decide whether <strong>renting would be cheaper than actually buying?</strong></p><p>Then <strong>What about training on device? </strong>in the case of small data set, if all the data needed for making a trained model is available on user’s device then there is an option to do all the work on locally on the device itself. But again it is only possible for small data set, if it doesn’t require much computing power. As the computing power of the smart phones is increasing day by day and by introduction of new SDKs such as Core ML and ML Kit which have been introduced recently it might be even the best option to use this method.</p><h4>Training in the cloud</h4><p>If you decide to choose the cloud option for training your model then you would facing with 2 options from it.</p><p>Either you can use <strong>hosted machine learning</strong> or you can use <strong>the general purpose cloud computing</strong>.</p><p>In most cases renting a cloud seems like much better and cheaper option compare to actually buying a high-performance computer. When you choose a cloud then your hands are open and it is very flexible depending on your needs. For instance if you want more compute power you can just simply upgrade your current package for that period of time.</p><p>So now first let’s start with the most affordable a suitable option which is renting a hosted machine learning.</p><p><strong>Renting a hosted machine learning:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1OkyXKFCrX4snhwEsgjJ9Q.png" /><figcaption>architecture of how hosted machine learning works</figcaption></figure><p>If the first option which was using a fully managed machine learning service isn’t suitable for you because you have special data that the service cannot provide then this option might be the best option for you.</p><p>This is a case between using a fully managed machine learning and doing everything by yourself on the cloud.</p><p>In this scenario <strong>you don’t need to have any knowledge about how machine learning works and how to train your model</strong>. You just simply upload your data set to the service and then it will handle it all automatically and let’s you to use your training model by calling it through an API.</p><p>What it means is that in most cases you <strong>won’t be able to download your training model</strong> and therefore you can’t have inference on the device (offline) and you are locked into that service and you can only use that service to be able to access your trained model.</p><p><em>You should note that </em><strong><em>there is an exception</em></strong><em> to this which is Google’s new </em><a href="https://cloud.google.com/ml-engine/"><em>Cloud Machine Learning engine</em></a><em>. Unlike all the other service providers, this cloud service let’s you to actually not only to train your model but to also be able to export it, which let’s you do predictions offline.</em></p><p><strong>Advantages:</strong></p><ul><li>You don’t need to deal with training the data. <strong>Just upload the data that’s it</strong>.</li><li>It’s <strong>very simple to integrate</strong> these services into your app.</li></ul><p><strong>Disadvantages:</strong></p><ul><li>In most cases <strong>you are locked into their service</strong> and cannot export your trained data. which means you won’t be able to do inference on device locally.</li><li>You not only need to pay for the computing a training model per hours for your usage but also you need to buy storage to upload your data, SQL service…etc. and you need to <strong>pay for each of these services individually</strong>.</li><li><strong>It is less flexible</strong> as in some cases, the service provider only supports a limited number of models that you can choose from.</li></ul><p><strong>General purpose cloud computing:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Lp9hJWYm1fGWMdfSMCd_QQ.png" /><figcaption>architecture of how general purpose cloud computing works</figcaption></figure><p>In this option you can rent a cloud computing service from one of the service providers and then <strong>just pay for computing hours</strong>.</p><p>You are able to do any kind of task and you don’t have any limitations with this option. You can run your favorite training software and upload your own data set.</p><p>Some of the Service providers are:</p><ul><li><a href="https://cloud.google.com/compute/">Google Compute Engine</a></li><li><a href="https://aws.amazon.com/ec2/">Amazone EC2</a></li><li><a href="https://azure.microsoft.com/en-us/services/virtual-machines/">Azure Virtual Machine</a></li></ul><p>When you are choosing this option <strong>you need to know what you are doing</strong> and how to train a model as your hands are open and if you don’t know what to do it will be just waste of money.</p><p>Also when using a cloud computing service, not only you need to pay for the compute hours for your training model but also you need to pay for the storage to upload your data.</p><p>After your training model is complete you can delete the instances and use your training model <strong>either locally on device (offline) or in the cloud (online)</strong>.</p><p><strong>Advantages:</strong></p><ul><li><strong>It is very flexible</strong>, your hands are open and you can use any kind of training software to train any type of data.</li><li>You only need to pay for compute hours and since training can be done in a period of time, you only need to pay for short period of time and then when you wanted to update or re-train your model pay again.</li></ul><p><strong>Disadvantage:</strong></p><ul><li>You really need to have full confidence on what you are doing and <strong>have complete knowledge </strong>about it. Otherwise you should hire a person to do the training for you.</li><li>Not only you should pay for the service but <strong>you need to also pay for the storage</strong> for your data to upload.</li></ul><h4>Training on your own computer</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qKEf7RAYMIfcOLVXlN_dyQ.png" /><figcaption>architecture of how training or your own computer works</figcaption></figure><p><strong>Similar to the process done in the cloud computing</strong>, for training on your own computer the only difference would be that you use one or more of your computers physically . Similarly your hands are open to run any of your preferred training softwares for any type of data that you have.</p><p>It is very flexible but again you should have in mind that when you are choosing this option you need to have knowledge about model training and that how machine learning really works.</p><p>Also it will be really suitable option for you<strong> if your data is sensitive</strong> and you need to have full insurance that your data is safe. Even-though the cloud companies usually has really good security for your data it won’t be as safe as having it locally on your own computer.</p><p><strong>Advantages:</strong></p><ul><li><strong>It is very flexible</strong>, your hands are open and you can use any kind of training software to train any type of data.</li><li>You have <strong>complete control over all of your data</strong>.</li><li>You <strong>won’t pay for renting a cloud computer</strong> or a cloud storage</li></ul><p><strong>Disadvantage:</strong></p><ul><li>You really need to have full confidence on what you are doing and <strong>have complete knowledge </strong>about it. Otherwise you should hire a person to do the training for you.</li><li>You need to buy a high-performance computer if you don’t have access to one.</li><li>You need to pay for the software and all the electricity bills and everything that keep your computer running.</li></ul><h4>Summary for training options</h4><p>At the end it all comes to you that <strong>whether you have knowledge of how to train a model</strong> or not, and <strong>whether you want to have your model in your hands</strong> which ables you to implement it in anyway you want (online or offline on the device).</p><p><em>There is an exception that is how often would you like to re-train your model. If you are planing to do online training then you should really think about whether it would be better option to use the cloud or training on your own computer would be cheaper after all.</em></p><p>So based on those two main questions let’s compare them.</p><p><strong>Whether you have knowledge of how to train a model?</strong></p><p><strong>If you don’t know</strong> how to train a model and you don’t have any background in that area then you should consider using <strong>hosted machine learning</strong> as you only need to upload your data set to the service and then it takes care of training the model by itself. By using these services you will be locked into their system and you can only access your model through API requests but the <strong>exception to this would be </strong><a href="https://cloud.google.com/ml-engine/"><strong><em>Cloud Machine Learning engine</em></strong></a>.</p><p><strong>If you do know</strong> how to train a model then your options are:</p><ul><li>Training on your own computer</li><li>Use a general purpose cloud computing for training</li></ul><p>And based on their advantages and disadvantages decide which one is more suitable for your based on your use.</p><p><strong>whether you want to have the option to export your model?</strong></p><p>If you need to have the ability<strong> to export your trained model</strong> which let’s you to be able to use it in any way you like (either locally on device or in the cloud) then your options are:</p><ul><li>Training on your own computer</li><li>Use a general purpose cloud computing for training</li></ul><p>And<strong> if it is not your main concern</strong> then the hosted machine learning can also be considered.</p><h3><strong>Do the predictions locally on device (offline) or in the cloud (online)?</strong></h3><p>When it comes to use your trained model to do the predictions you have two options:</p><ul><li>Locally on device (offline)</li><li>In the cloud (online)</li></ul><p>Certainly choosing the right option in this part is one of the most important decisions when it comes to implementing your machine learning as this decision might have big impact in the <strong>speed, power, privacy and cost</strong>.</p><p>Each of these services have their own tradeoffs. For instance when by choosing the cloud option the app should be connected to the network for it to work and in the other hand if you do the predictions locally on device you are always limited to the hardware restrictions can cannot perform any type of machine learning due to the RAM and CPU limitations.</p><p>But let’s first see <strong>what are these options really? </strong>and<strong> how each of them works?</strong></p><p>As you have seen the first steps for implementing a machine learning is gathering all of the data that you need to use for your model and by the options we discussed train a model. These are the common steps between both inferences on device(offline) and in the cloud(online).</p><p>From here if you choose:</p><ul><li><strong>On device (offline)</strong>, then it means in most cases, once you have your training model in use for the predictions then you cannot update that model easily. Over time when your model got out-of-date and didn’t work perfectly in the way that it should, you need to re-train your model with more/newer data and then update the app which includes the new model.</li><li><strong>In the Cloud</strong>, You don’t need to be worried about re-training your model and how to publish the new model. Your model is continuously being updated. Also if you decided to re-train your model because of any reason then it won’t be a problem and you can easily just update the model in the cloud and everyone can get benefit from it.</li></ul><h4>Locally on device (offline)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/251/1*gXxZh7DuGYFeHm5CpSOYmA.png" /><figcaption>architecture of how inference locally on device works</figcaption></figure><p>Whether choosing this option is totally up to the case of your usage. If your model is not that heavy and huge then inference locally on device can be an option for you to consider.</p><p>One of the biggest examples of this would be the Core ML framework from Apple. When you<a href="https://developer.apple.com/documentation/coreml"> reference to the documentation of this framewrok on their website</a>, you can see this part,</p><blockquote>Core ML is optimized for on-device performance, which minimizes memory footprint and power consumption. Running strictly on the device ensures the privacy of user data and guarantees that your app remains functional and responsive when a network connection is unavailable.</blockquote><p>This text is referring to two of the main reasons why to use this method: <strong>ensuring the privacy of user’s data</strong> and <strong>no need for network connection</strong>.</p><p>But what are the reasons generally? It is all divided into 3 categories of: <strong>Speed</strong>, <strong>Cost </strong>and<strong> Privacy</strong>.</p><p><strong>Speed: </strong>Since you don’t need to rely on network connectivity in order to use this method and all the requests are sent locally on the device itself, then <strong>speed is the main advantage</strong> in this method. It is much faster and more reliable compare to inference in cloud and you can use it without any worries. It also allows you to do things like realtime predictions on the big data types since you won’t need to send the data over the network which is impossible when you use inference in the cloud.</p><p><strong>Cost:</strong> You use the user’s device itself for doing predictions and don’t need to pay for the cloud and all the other small things that comes with it. It will have a <strong>huge impact on big applications</strong> and the number of requests increase and you can keep the cost low and have control over it.</p><p><strong>Privacy:</strong> This is more about the privacy of the user’s data as it is all being handled inside the device and no data will be upload to anywhere. In one sentence, <strong>data will never leave the device</strong>.</p><p><strong>Advantages:</strong></p><ul><li>There is no need for any type of network connectivity for it to work</li><li><strong>Speed is much faster</strong> and in some cases allows you to do things which are impossible when you use the cloud</li><li>You don’t need to deal with all the server side problems. For instance when your app becomes more popular there is no need to you to scale up.</li><li>The data of the users is safe.</li><li>Since you don’t use the cloud, then you don’t need to pay for anything.</li></ul><p><strong>Disadvantages:</strong></p><ul><li>There is always hardware limitations and if your trained model is big it is almost impossible to do inference on the device due to the performance limitation of the device.</li><li>Including the model in the app bundle will <strong>increase the size of the app</strong> download signficantly, often by many megabytes.</li><li>It is usually really <strong>difficult to update the model</strong>. Users need to update their app in order to benefit from the new model or the app needs to download it automatically.</li><li><strong>Hard to use in other platforms</strong>. For other platforms, you need to do the inference for each individually.</li><li>Other developers can dig around inside your app bundle. It’s easy to copy over the learned parameters, and if you’re including a TensorFlow graph definition or caffemodel file it’s very simple for unscrupulous people to steal your entire model.</li></ul><h4><strong>In the cloud (online)</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EB5O3LGJ_sOsmMc6TUbH6g.png" /><figcaption>architecture of how inference in the cloud works</figcaption></figure><p>Once you have the training model you can setup a server either on your own computer or the one you have rented from a cloud service provider. After that you need to upload your training model into it and the server gives you API endpoints that your app can call through a network connection.</p><p>When you are choosing this option, most of the complexity of the app will be on the server side and the app itself would be simple. The model can continuously be updated or also you can re-train and improve a model with new features and deploy it on the server. With this option you won’t be needing for update your actual app every time that you have a new model and the app can automatically use it through the server.</p><p><em>You should note that instead of rolling your own API from scratch you can also use existing tools, such as </em><a href="https://tensorflow.github.io/serving/"><em>TensorFlow Serving</em></a><em>.</em></p><p>What are the main reasons to choose this method? there are 2 main categories that why you may be considering to choose this over inference on device: <strong>Convenience</strong> and <strong>Performance</strong>.</p><p><strong>Convenience:</strong> there are few reasons why it would be more convenient to use a server. The first one would be that since your inference is in the cloud you can easily get to use it in any platform easily and you won’t be needing to implement the inference individually for each, over and over. The second reason is that you don’t need to obfuscate the data since normally, nobody can steal your training model which means it can be more secure for you as a developer.</p><p><strong>Performance:</strong> Since the model is in the cloud, it can continuously improve and get better without the need to be re-trained and it can be much smarter and be more accurate. Also since there will be no hardware limitations, you can do any type of predictions and your hands would be open to do anything.</p><p><strong>Advantages:</strong></p><ul><li>You can <strong>update the model anytime</strong> you want.</li><li>Since machine learning logic would be on the server, you can easily <strong>port the app to different platforms</strong> such as iOS, Android, Web…etc.</li><li>It will be <strong>more secure</strong> for you as developer since normally, no one can steal your trained model and parameters.</li><li>The app will be much simpler as all the logic and complexity will be on the server and it will be really nice specially if you already use a back-end for your app as you can integrate with it.</li></ul><p><strong>Disadvantages:</strong></p><ul><li>User always<strong> has to be connected to a network</strong> in order to use the service</li><li>You need to <strong>deal with all the typical server headaches</strong> such as protecting against hackers, denial-of-service attacks, preventing downtime…etc.</li><li><strong>You need to pay for it</strong>. As users are constantly sending files and data to the server you need to pay for the bandwidth and it will be hard to have control over the cost.</li><li>If your app becomes very popular, you may need to <strong>scale up</strong> to many servers. It’s not good for business if your app goes down because the servers are overloaded.</li></ul><h3>Conclusion</h3><p>So there are many options that you can choose from and many ways to start implementing and using a machine learning. Each of these ways has their own advantages and disadvantages and at the end it all depends on you and your usage.</p><p>It is really hard to recommend any of these ways for all cases as one can be the best option for one person but it won’t be for another person. But based on the conclusions and comparison between each of these parts separately we can say:</p><ul><li>If you just want to get started and you don’t know anything about the machine learning and how you can train a model and the costs isn’t your main concern then it would be better for you to consider fully managed machine learning.</li><li>If you have exported your trained model and having the best speed and availability without the network connection is crucial, and you don’t plan to implement really huge machine learning and also you want to just deploy the app for maximumly one or two platforms then your best bet would be to choose to do the predictions locally on device (offline).</li><li>And if you are planing to do something performance heavy or you want to deploy the app for multiple platform then you should take a closer look at inference on the cloud (online).</li></ul><p>I hope this article has been useful for you and helped you to have a better idea on how to start implementing machine learning into your new app.</p><p>Good luck with your new app and have a nice coding time ;)</p><h3>Some of the references:</h3><ul><li>influenced by a post on <a href="http://machinethink.net/blog/machine-learning-device-or-cloud/">machinethink.net</a></li><li><a href="https://www.quora.com/What-is-an-easy-way-to-explain-online-vs-offline-machine-learning-to-a-machine-learning-novice">quora.com</a></li><li><a href="https://developer.apple.com/documentation/coreml">developer.apple.com</a></li><li><a href="https://en.wikipedia.org/wiki/Machine_learning">wikipedia.org</a></li><li><a href="https://medium.com/@jason.io/wtf-is-offline-ai-293d328f8a8f">medium.com</a></li></ul><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2F7b0181%3Fas_embed%3Dtrue&amp;dntp=1&amp;url=https%3A%2F%2Fupscri.be%2F7b0181%2F&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/748a06e9019942f2677bba917ef55804/href">https://medium.com/media/748a06e9019942f2677bba917ef55804/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=934801e0a418" width="1" height="1" alt=""><hr><p><a href="https://medium.com/good-audience/online-vs-offline-machine-learning-which-one-to-use-934801e0a418">Online vs. Offline Machine learning: Which one to use?</a> was originally published in <a href="https://medium.com/good-audience">Good Audience</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>