<?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 Gary Thomas George on Medium]]></title>
        <description><![CDATA[Stories by Gary Thomas George on Medium]]></description>
        <link>https://medium.com/@garythomasgeorge?source=rss-9314c115d166------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*8AXu8xrGaqPJfs-fpBDUmw.jpeg</url>
            <title>Stories by Gary Thomas George on Medium</title>
            <link>https://medium.com/@garythomasgeorge?source=rss-9314c115d166------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 10:57:15 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@garythomasgeorge/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[CX Agent Studio: The Evolution of Google’s Conversational AI Stack — A first look]]></title>
            <link>https://medium.com/google-cloud/cx-agent-studio-the-evolution-of-googles-conversational-ai-stack-a-first-look-09d87a95a093?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/09d87a95a093</guid>
            <category><![CDATA[google-cloud-platform]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Sat, 11 Apr 2026 02:17:39 GMT</pubDate>
            <atom:updated>2026-05-14T02:56:21.342Z</atom:updated>
            <cc:license>http://creativecommons.org/licenses/by/4.0/</cc:license>
            <content:encoded><![CDATA[<p><em>An early product breakdown for those evaluating Google’s latest conversational AI tooling — written by someone who’s tried the three generations.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dugSWke1ifEVgIgq3kszAQ.png" /><figcaption>The evolution of Google’s Conversational Platforms</figcaption></figure><h3>The CX Bot Build Dilemma</h3><p>Your CX team wants AI agents. Your engineers want flexibility. Your leadership wants something shipped. And somewhere in Google’s sprawling product portfolio, there’s a tool that promises to square that circle.</p><p>That tool is CX Agent Studio — Google’s latest entry in its conversational AI platform stack. But if you’re the decision maker trying to evaluate it, the marketing copy won’t tell you what you actually need to know: where it sits relative to what came before, what it genuinely unlocks, and where it will quietly frustrate you six months into a build.</p><p>I’ve built on Dialogflow, worked through Conversational Agents, and recently built a working demo on CX Agent Studio for a fictional D2C brand called April Threads. This is an early first look at what is good and what could be improved.</p><h3>Where It Came From: The Lineage That Matters</h3><p>Understanding CX Agent Studio requires understanding the product decisions that preceded it. I first tried out Dialogflow (then called <em>Api.ai</em>) around 2018, by then it had already had some history. Google’s conversational AI tooling has gone through three meaningful generations, each reflecting a shift in how the industry thought about intelligent automation.</p><h4>Dialogflow: Structured, Predictable, Limited</h4><p>Dialogflow was the workhorse. Intent-based, flow-driven, and deeply deterministic. You defined what users could say, what the bot would respond, and exactly what would happen at each node. It was rigid, but it was predictable — a quality that enterprise CX teams valued highly. When I built on Dialogflow CX, the mental model was a flowchart. Every deviation felt like a bug.</p><h4>Conversational Agents: More Power, Same Paradigm</h4><p>Conversational Agents (the successor to Dialogflow CX) brought better NLU, more sophisticated session management, and improved tooling. But the underlying paradigm remained flow-based. You were still, fundamentally, building a decision tree, with the ability to add some smarter agents in between. The ceiling was clear: complex multi-turn interactions that required genuine reasoning, not just pattern matching, would always hit it.</p><h4>CX Agent Studio: The Agentic Shift</h4><p>CX Agent Studio represents a genuine paradigm shift, not just a version bump. It’s built on top of Google’s Agent Development Kit (ADK) and is oriented around agents, subagents, tools, and orchestration — not intents and flows. The product is asking a different question: not “what did the user say?” but “what does the user need, and which agent should handle it?”</p><p>That’s a meaningful evolution. And it creates a meaningfully different set of tradeoffs.</p><h3>What’s New: Features That Stand Out</h3><p>Rather than cataloguing every feature, I want to call out the ones that stood out in the 2 weeks I have tried it — and why they matter from a product perspective.</p><h4>1. Visual Subagent Orchestration</h4><p>CX Agent Studio gives you a visual UI to structure, connect, and order subagents within a parent agent. This is more significant than it sounds. In Dialogflow, as a product scaled, you’d end up managing a web of Flows and Pages and the flows became increasingly hard to organize. In my last implementation, hit the 250 pages limit for many flows. 200–300 pages would like like a literal web.</p><p>The new UI lets you think in terms of responsibilities — which agent handles returns, which handles order tracking, which escalates to a human — and see that structure clearly. And the UI improvement that seems small — the ability to move around the sub agents logically you want, a big improvement for someone who has built flows running into 100s of pages.</p><h4>2. Global Variables Management Page</h4><p>There’s a dedicated page for managing global variables across your agent — a small feature with outsized implications. State management has historically been one of the messiest parts of building conversational products. Variables would get set in one flow, referenced in another, and debugging a mismatch was painful.</p><p>Having a single place to define, view, and manage variables signals that Google is treating state as a first-class product concern, not an afterthought. I wrote in more detail about how I would have built this<a href="https://garythomasgeorge.medium.com/prd-to-prototype-with-gemini-and-figma-make-the-future-of-ai-product-management-7a93bd09a6a7"> in a previous post </a>— it’s worth reading if you’re planning to build anything with a meaningful session state.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zSFGJwQF6tvo51QtO4TJ3g.png" /><figcaption>The variable management page is a huge addition</figcaption></figure><h4>3. Conversation Preview / Test Mode</h4><p>Anyone who’s suffered through Dialogflow’s simulator will appreciate this one. The test and preview experience in CX Agent Studio is noticeably cleaner — faster to load, more readable, and better at surfacing what actually happened during a conversation turn. The visibility into agent reasoning and tool calls during testing makes debugging significantly less frustrating.</p><p>For teams iterating quickly, this isn’t a minor UX nicety — it’s a real velocity difference.</p><h4>4. Widget Support</h4><p>CX Agent Studio includes built-in widget support, which means you can deploy a conversational interface to an end user without needing a separate front-end build. For teams that want to ship fast — or run a proof of concept with real users — this removes a layer of dependency that often slows down CX projects.</p><p>It’s not a fully custom UI solution, but for many enterprise CX use cases it’s more than sufficient.</p><h4>5. Eval and Testing</h4><p>This I feel is a biggest feature change from the previous iterations of Dialogflow because of how this has incorporated AI into the process of testing. I am still exploring this feature but in a generative world, this assumes a lot of significance.</p><p><strong>Scenario-based testing: </strong>Scenario tests are AI-powered. You describe a user’s goal — for example, “a customer wants to track their order and escalate if it’s delayed” — and the platform automatically simulates users, generates conversations, and tests your agent’s ability to handle that goal robustly. It surfaces edge cases you wouldn’t have thought to write manually. This is particularly valuable in the early stages of building, when you’re still discovering where your agent breaks.</p><p><strong>Golden tests: </strong>Golden tests are ideal conversations you define as the correct path — including expected tool calls and agent responses. Once you’ve established a conversation that works well, you can save it directly from the simulator as a golden test case. These become your regression suite: every time you change the agent, you can run the goldens to verify that your critical paths haven’t broken.</p><h3>The Build vs. Buy Lens</h3><p>Every PM who’s evaluated a platform tool has asked some version of this question: do we build the scaffolding ourselves and get full control, or do we buy an abstraction layer and get speed with constraints?</p><p>CX Agent Studio sits clearly on the “buy” side of the spectrum — but understanding what it abstracts (and what it doesn’t) is the real evaluation question.</p><p>Raw ADK gives you full programmatic control over your agent’s behavior, tools, memory, and orchestration logic. You’re writing Python, managing your own infrastructure, and owning every design decision. The ceiling is very high. So is the overhead.</p><p>CX Agent Studio wraps ADK in a managed UI and opinionated structure. It handles agent orchestration, session management, evaluations, even deployment and monitoring for you. You move faster, but you’re operating within Google’s model of what an agent product should look like.</p><p>A rough heuristic: if your use case maps well to CX — customer support, order management, FAQ handling, escalation routing — CX Agent Studio will likely get you to production faster than raw ADK. If you have highly specialized agent behaviors, deep integration requirements, or need full observability into every layer, you’ll feel the ceiling.</p><h3>What It’s Good For — And Where It Falls Short</h3><h3>Where it shines</h3><ul><li>Enterprise CX teams that need to ship multi-turn conversational products without building AI infrastructure from scratch.</li><li>Products with multiple distinct intents or domains that benefit from clear subagent separation.</li><li>Teams that need a deployable, testable product quickly — the ability to add widgets and test mode make early validation accessible.</li><li>Organizations already invested in the Google Cloud ecosystem, where native integrations reduce friction.</li></ul><h3>The one feature I would have added: Deterministic Flows</h3><p>As a product person, if I had to add one feature that would be determinstic flows in some capacity. This is the most significant structural gap, and decision makers should go in with eyes open about it.</p><p>CX Agent Studio is fully committed to the agentic paradigm. Every conversation is handled by agents making decisions — there is no native ability to create new deterministic flows. You can import existing Dialogflow flows, but you cannot build new ones within the platform.</p><p>For many products, this is fine. But consider the use cases where certain paths must be predictable: regulatory disclosures, step-by-step troubleshooting scripts, compliance-driven conversation sequences. In these scenarios, the absence of deterministic flow creation is a real constraint, not a minor limitation.</p><p>This is an architectural bet by Google that the future is agentic. They may be right. But bet-hedging is a reasonable product strategy, and right now CX Agent Studio doesn’t give you that option natively.</p><p>Customization more broadly also remains a step behind raw ADK. The platform is maturing, and Google is clearly investing in it, but teams with highly specific requirements around agent behavior, memory, or tooling may find themselves working around the abstraction layer rather than with it.</p><h3>The Honest Verdict</h3><p>CX Agent Studio is a genuinely meaningful step forward from its predecessors. The shift from flow-based to agent-based is the right architectural direction. The UI improvements — subagent orchestration, global variables, better testing — reflect real product thinking, not just feature additions.</p><p>For most decision makers evaluating a CX AI platform today, it is the right starting point. It abstracts the complexity that would otherwise require a dedicated ML engineering team, while still giving you enough structure to build products that actually work.</p><p>But it is a foundation, not a finished product. Go in knowing the customization ceiling. Keep an eye on how the ADK integration deepens over time — that’s the key indicator of how far the platform can grow. And if your product has deterministic flow requirements, either plan to manage that within your agent’s prompt logic or keep a Dialogflow component in your architecture for those specific paths.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*thAT9E0WGC8tsuE82wCLhQ.png" /><figcaption>The April Threads CX Agent Demo</figcaption></figure><p>In Part 2 of this series, I’ll walk through exactly how I built the April Threads CX agent from scratch . If you want to see how CX Agent Studio works in practice rather than in theory, that’s the post for you.</p><p>If you resonate with the post and have worked on any of these products, how has your experience been, leave a comment. You can also <a href="http://linkedin.com/in/garythomasgeorge">connect with me on LinkedIn</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=09d87a95a093" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/cx-agent-studio-the-evolution-of-googles-conversational-ai-stack-a-first-look-09d87a95a093">CX Agent Studio: The Evolution of Google’s Conversational AI Stack — A first look</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Google ADK’s Human-in-the-Loop Story Has a Production Gap — And one way it could be fixed]]></title>
            <link>https://garythomasgeorge.medium.com/why-google-adks-human-in-the-loop-story-has-a-production-gap-and-one-way-it-could-be-fixed-66aabef33a32?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/66aabef33a32</guid>
            <category><![CDATA[google-adk]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[google]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Tue, 31 Mar 2026 21:28:20 GMT</pubDate>
            <atom:updated>2026-04-01T16:07:39.916Z</atom:updated>
            <content:encoded><![CDATA[<h3>Why Google ADK’s Human-in-the-Loop Story Has a Production Gap — And one way it could be fixed</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Af-KKUAJWCsqWMgm" /><figcaption>Photo by <a href="https://unsplash.com/@taopaodao?utm_source=medium&amp;utm_medium=referral">taopaodao</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>I’m a product manager. I don’t write production Python for a living. But over the past few months I found myself doing something I hadn’t planned — shipping an open source contribution to Google’s Agent Development Kit ecosystem, because I stumbled into a problem worth solving.</p><p>My interest in chatbots started a few years ago. In 2018, I built a chatbot using Google Dialogflow and the Zomato API to help people discover places to eat. Back then it took me about three months of part-time weekend work to get a basic PoC running. A couple of months ago, when Advent of Agents Season 1 dropped, I started building a chatbot with a live agent handoff PoC. With AI-assisted building, what would have taken months took a couple of weeks. I even put it on GitHub but never wrote about it. It felt incomplete — and I eventually figured out why.</p><p>Working on chat products, one thing I’ve learned is that production systems don’t work seamlessly. They’re messy, often carrying years of accumulated technical debt across many teams and reorgs. When I thought about my chatbot and human agent PoC, those same problems came to mind.</p><h3>The Problem I Stumbled Into</h3><p>One of the core concepts in my PoC is Human-in-the-Loop — HITL in developer shorthand. The idea is simple: before an AI agent takes a high-stakes action, a human reviewer should be able to approve it. Think of an agent that processes expense reimbursements, issues credits to customers, or modifies customer records. You want a human checkpoint before those actions execute.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1010/1*rmc5OBXHsPyoxJtjUIcweQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3EbnW0YOmdp1EmGEsvv2Bg.png" /><figcaption>The PoC for a human-in-the-loop chatbot and the agent dashboard app</figcaption></figure><p>While the demo looked good as a PoC — a customer-facing chatbot, a human agent interface that gets notified when a HITL event occurs, an option to “Barge-in” and take over the chat — I realised that real-world products don’t work that way. In a real setup like the environment I work in, the chatbot and human agent app are two different products maintained by different teams, communicating through a third-party tool. What I’d built was tightly coupled. What was actually needed was something frontend-agnostic.</p><p>Looking at the ADK documentation, there’s a feature for this called Tool Confirmation, introduced in version 1.14.0. I had used a version of this in my PoC. But when I went to understand it properly, I found something buried in the official docs that stopped me.</p><p>The docs say — explicitly — that Tool Confirmation does not support DatabaseSessionService or VertexAiSessionService.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ccfHyKpXs7VoerNXXlczwg.png" /><figcaption>Screenshot of the Action Confirmations Documentation</figcaption></figure><p>Those are the two session backends that any real production deployment of ADK uses. In-memory sessions don’t survive server restarts. They don’t scale horizontally. They’re for development, not production. So ADK’s built-in HITL story, as of today, simply doesn’t work in production.</p><h3>Validating That This Was Real Pain</h3><p>So before building anything new, I did what product managers do — I went looking for evidence that other people were hitting the same wall. Afterall, I was not as close to the problem like a dev building this as their day job.</p><p>What I found on the google/adk-python GitHub confirmed it wasn&#39;t just me. A feature request filed in August 2025 explicitly compared ADK unfavorably to LangGraph&#39;s interrupt_before — the mechanism that lets you pause any workflow node for human approval with full state persistence. It had 8 upvotes from real developers. A July 2025 issue described needing event-driven HITL for a banking transaction confirmation use case. A November 2025 bug report showed a developer hitting a concrete production error: <em>&quot;invocation_id provided but app is not resumable.&quot;</em> Discussions showed HITL confirmation breaking entirely inside multi-agent flows using ADK&#39;s A2A protocol.</p><p>The only existing community example — a repo showing HITL for an expense approval workflow — used a 30-second polling loop and stored everything in memory. No database persistence. No audit log. No multi-agent support. A demo, not a solution.</p><p>This was a real gap, not an imagined one. The community was asking for it, hitting walls trying to build it, and working around it with fragile polling hacks.</p><h3>The Decision to Build It as a Generic Solution</h3><p>I could have patched my own PoC and moved on. Instead I decided to build something reusable — a community contribution rather than a private fix.</p><p>That decision came directly from my background. When I mapped out the problem, it was clear this wasn’t specific to my use case. Any team deploying ADK agents in production would hit the same wall the moment they tried to use a real session backend. The right answer wasn’t another custom implementation bolted onto another demo. It was a shared layer that any ADK developer could drop into their existing agent with minimal code changes.</p><p>That’s the product thinking part of this. Not the code — the decision about what to build and where to put it.</p><h3>What I Built (With AI)</h3><p>I built the <strong>ADK HITL Approval Gateway</strong> — a production-ready middleware layer that sits between ADK agents and human reviewers</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TxAr3IxOMy0NIg0Y1C4xQA.png" /><figcaption>The ADK Webview example of a chat</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MKnv3hQp9gqdUp8hjEaw4A.png" /><figcaption>A Streamlit app dashboard showing the example of the ADK HITL Approval</figcaption></figure><p>The core design decision was to make it session-agnostic. Instead of relying on ADK’s session to hold approval state, the gateway manages its own state in a SQLite database via a FastAPI backend. This means it works regardless of which session service your production app uses — InMemorySessionService, DatabaseSessionService, or VertexAiSessionService. The agent resumes via ADK&#39;s standard REST API after a human decision is submitted.</p><p>Adding HITL to an existing ADK tool takes about five lines of code:</p><pre># Before — no human oversight<br>tools=[FunctionTool(process_credit)]<br><br># After — production HITL in one line<br>tools=[HITLGateway(FunctionTool(process_credit), risk_level=&quot;high&quot;)]# After — production HITL in one line<br>tools=[HITLGateway(FunctionTool(process_credit), risk_level=&quot;high&quot;)]</pre><p>The contribution has three layers. The HITLGateway is a drop-in Python wrapper for any ADK FunctionTool. A FastAPI backend manages the approval queue, persists decisions to SQLite, and maintains a full audit log. A reference Streamlit dashboard gives reviewers a visual inbox to approve, reject, or escalate pending actions. The full implementation is a draft PR on adk-python-community, with 11 passing tests and a working end-to-end demo.</p><p>I also built it with ADK 2.0 in mind. ADK 2.0 Alpha introduces a new RequestInput pattern for graph-based workflows — a cleaner approach to HITL that doesn&#39;t carry the session service limitations. By building an adapter layer, the same approval dashboard and backend work with both ADK 1.x today and ADK 2.0 when it reaches stable. Teams won&#39;t need to rebuild their approval layer when they upgrade.</p><h3>What This Is Really About</h3><p>I’ll be direct about something. I built this partly because I believe it’s useful to the ADK developer community. And partly because as a PM, I’ve seen firsthand how real-world production systems are assembled from different solutions built and maintained by different teams — and how often the gaps between those systems are where things break.</p><p>The ADK multi-agent patterns guide — <a href="https://adventofagents.com/2026/03/14">specifically Day 14 of Google’s Advent of Agents Season 2 </a>— covers the HITL design pattern conceptually. It shows the shape of the solution. What it doesn’t provide is a production implementation. That’s what this contribution is.</p><p>If you’re building ADK agents and hitting the same wall I described — or thinking about what HITL should look like in your production agent workflows — the code is at github.com/google/adk-python-community and the <a href="https://github.com/google/adk-python-community/pull/110">PR is open</a>.</p><p>And if you’re a PM wondering whether you can build things like this without being a full-time developer — you can. You just need to be honest with yourself about what you’re actually doing. I didn’t write this code from scratch. I used Google Antigravity and Claude Code extensively. What I contributed was the product thinking: finding the right problem, validating it properly, scoping a generic solution, and understanding the ecosystem well enough to put it in the right place.</p><p>That combination — product judgment plus AI-assisted building — is, I’d argue, exactly what the next generation of builder PMs needs to develop. This was my attempt at it. If you’re building in this space and have solved similar problems, I’d love to compare notes.</p><p>Thank you for reading. If you need to connect with me, <a href="http://linkedin.com/in/garythomasgeorge">I am on LinkedIn</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=66aabef33a32" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[PRD to Prototype with Gemini and Figma Make: The future of AI Product Management]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://garythomasgeorge.medium.com/prd-to-prototype-with-gemini-and-figma-make-the-future-of-ai-product-management-7a93bd09a6a7?source=rss-9314c115d166------2"><img src="https://cdn-images-1.medium.com/max/2600/0*vdhaqczhHDr8WD55" width="5418"></a></p><p class="medium-feed-snippet">Building a feature in Google Conversational Agents in an AI first approach</p><p class="medium-feed-link"><a href="https://garythomasgeorge.medium.com/prd-to-prototype-with-gemini-and-figma-make-the-future-of-ai-product-management-7a93bd09a6a7?source=rss-9314c115d166------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://garythomasgeorge.medium.com/prd-to-prototype-with-gemini-and-figma-make-the-future-of-ai-product-management-7a93bd09a6a7?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/7a93bd09a6a7</guid>
            <category><![CDATA[gemini]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[figma]]></category>
            <category><![CDATA[google-dialogflow]]></category>
            <category><![CDATA[conversational-ai]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Fri, 01 Aug 2025 21:43:03 GMT</pubDate>
            <atom:updated>2025-08-03T17:14:06.052Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Digital Payments in an increasingly contactless world]]></title>
            <link>https://garythomasgeorge.medium.com/digital-payments-in-an-increasingly-contactless-world-90f2f1226f90?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/90f2f1226f90</guid>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[india]]></category>
            <category><![CDATA[payments]]></category>
            <category><![CDATA[product]]></category>
            <category><![CDATA[google]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Sun, 06 Sep 2020 15:04:51 GMT</pubDate>
            <atom:updated>2020-09-10T04:40:34.561Z</atom:updated>
            <content:encoded><![CDATA[<h3>The future of Digital Payments in an increasingly contactless world</h3><p>Can the introduction contactless cards on payment apps like Google Pay replace the QR code from India’s payment landscape?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dhH_xEX7YSpJrNHNC-zBFQ.jpeg" /><figcaption>A grocery and bakery store in Bengaluru</figcaption></figure><p>Last year when conferences were still ‘live and in-person’, I got the chance to attend <a href="https://razorpay.com/ftx/">FTX 2019</a>, RazorPay’s annual fintech expo, and conference. With a mix of people heading payments at pioneers like Mastercard and PayPal to the new players on the black Amazon Pay and Netflix, the panel discussion on frictionless payments was an interesting one. The discussion moved to around how the payments landscape in India offered so many options when my question, incidentally on similar lines, was taken up —</p><blockquote>“<em>Does India have too many payment options?”</em></blockquote><h3>The Digital Payments scene in India</h3><p>I have been observing the Indian fintech payments scene over the past few years, as a matter of interest. Digital Payments got me interested at work by working with institutions on automating their fee payments via payment gateways. This was around 2016 when demonetization had suddenly pushed India to go digital. Although having existed for more than one and a half-decade in India, card payments were yet not a casual payment mode outside metros and cash was still king. Post-2016 also produced new modes of payments like the Unified Payments Interface (UPI), Adhaar enabled Payment System (AePS), Bharath Bill Pay System (BBPS) to name a few. Before that India had wallets for some years and it was the face of digital small ticket payments in metros. UPI as we know now changed all of that by first completely owning Peer to Peer (P2P) payments space. <em>When was the last time you used net banking to pay your share of that movie share to your friend?</em> And then also capturing Peer to Merchant (P2M) Point of Sale (PoS) payments effectively overshadowing wallet and at times even card payments. Fast forward to 2020 and we now have a host of new and old payment modes in India existing together— Cards, UPI, NetBanking, NEFT/RTGS/IMPS, UPI, AePS, BBPS, Wallets.</p><h3>QR Codes and India</h3><p>The growth of small-ticket P2M payments first by wallets and then by UPI has been one of the biggest contributors for the spurt in digital payments in India. This was enabled in no small part by the ubiquitous QR code, a 2D barcode that was first designed in Japan in 1994.</p><p>Each country has taken different routes in adopting digital payments. Payments in the western countries were however built on top of an already existing network of credit cards and later improved upon on. Card payments there are simpler and frictionless for the end-user. While China led the world in adopting QR codes for payments, the rise of QR codes over card-based POS terminals in India can be attributed to factors like:</p><ol><li>QR codes are easy to use for a country with smartphone users crossing 500 million in 2019. A largely young population that skipped the desktop revolution and jumped directly to the mobile certainly helped.</li><li>They are easy and cheap to place as well as maintain. With just a printout over a plastic required, QR codes can be generated and regenerated as many times with close to zero expense on the establishment. A set of 5 plastic QR code stands costs just under INR500 (~7USD). Print and stick a Black and white copy on a paper, the cost becomes even more negligible.</li></ol><p>Now it might sound counter-intuitive but does the easiest and cheapest always mean the best user experience?</p><p>Now consider this scenario. You are at the nearby convenience or fast-food chain, for a small ticket purchase, say for Rs 500 (~7USD) in India. You find the QR code scanner, you scan the code and the scanner keeps trying to read the code. Then you wait…and wait… and wait for some more. No response. The store clerk tells you to step outside and try because the network reception may be poor. You take the QR code stand and go outside and try again. If you have had this experience, you are not alone. We have all faced it at some time. Easy to use and cheap to produce but not necessarily the best experience. Let us also step back to just after the billing was done. The clerk announced the total bill amount, you take out your phone, open the app, look for the code stand, pick up or adjust it, and then go to scan. If the scan was successful, you ask the store clerk the amount again and enter the value. So in spite of being hugely popular, QR codes not always offer the best experience. Attempts like Bharath QR have reduced the problem of multiple QR codes but still, the other problems remain.</p><h3>A better User Experience?</h3><p>Continuing with the previous example, imagine if you were paying by a credit or debit card. The store clerk asks for your card, enters the amount, and then asks for the PIN. One touchpoint for the end-user, entering the PIN. A notification by the country’s central bank RBI in 2015 aimed to make this process even smoother by allowing Tap and Pay using NFC, for payments less than INR 2000. In many of the developed countries, where cards already had a high acceptance, made the shift to contactless payments a few years ago. India took some time to find ground as well as get the infrastructure ready but that is changing fast. Contactless payments accounted for about 15% face to face transactions in India in 2019. According to a <a href="https://www.financialexpress.com/money/india-making-rapid-strides-in-bridging-digital-payments-divide-mastercard/2050456/">report by Mastercard</a>, 40% of transactions in the first quarter of 2020, were contactless.</p><p>Card-based payments offer many other advantages apart from reducing touchpoints. It requires only one network, that of the retailer, to be stable, unlike the QR code method which depends on the network of the payee (your mobile phone network). It is also much safer in the post COVID world where the impetus is on avoiding unnecessary contacts. In fact, the RBI also recently allowed all transactions to be contactless, the PIN will be however required for transactions above Rs. 2000. The biggest advantage is how it drastically reduces the payment cycle. With POS terminals connected to billing platforms, the process involves just a tap of the card on the machine placed conveniently near the payee.</p><h3>The Best of both worlds</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0FpzQTr3EobWi46Lg8nHvQ.jpeg" /><figcaption>Phone-based Tap and Pay — Representational Image</figcaption></figure><p>What if the convenience of mobile and card could be given together. That is where tokenized cards come in. By adding your debit or credit card to an app, you are able to make use of the NFC based tap and pay system in POS terminals to make the payment. The technology itself is not new, Samsung phone users have had Samsung Pay since 2017 at least. With a high degree of phone penetration, this technology was waiting to take off. In fact, the National Payments Council of India (NPCI) which manages UPI had initiated the process a few months ago.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/986/1*4RNoKWt3fgugjQFqP3Q8Zw.png" /><figcaption>Google Play add card flow</figcaption></figure><p>Google Pay which announced tokenization service on Google Pay <a href="https://india.googleblog.com/2019/09/powering-digital-economy-building.html">last year</a> rolled out this service a few days back. I tried the setup and it was very easy to do. At the time of writing this, the service is available for Axis Bank and State Bank of India cards only. Other cards will hopefully be added soon. What is more, this is the technology that could even find acceptance for ATM transactions. Imagine walking into an ATM, taping your phone on the machine, authenticate from your phone and then complete the cash withdrawal. In fact, the NPCI and RBI are exploring proposals for the same.</p><p>Therefore, for a population that knows its phones and apps, the benefits can be many. To list down a few:</p><ol><li>It offers the convenience of a phone and the superior user experience of card payments. The best of both worlds in one.</li><li>In the post Covid world, it also offers a safer transaction experience. The lesser contact points for the customer, the safer for the customer and the store workers.</li></ol><p>I started with the question asked at the event, whether India had too many payment methods. The answer that came out of the discussion was something that is true for most product decisions in India. It is too large a market for only one to exist. Unlike other homogenous markets in the US, Europe or even China, India is diverse and hence comes with many use cases. Probably the most important being that POS machines are expensive to buy and maintain. The QR code scanner is still the best bet for many corner side or fast-food vendors. Similarly, many have niche use cases like the Aadhaar enabled Payment System made for where ATMs and POS machines cannot be set up or serviced.</p><p>But card payments have grown constantly over the last 4 years (debit cards with a CAGR 53% and credit cards with 30%)and with a large network of POS terminals, card-based payment on UPI apps like Google Pay will be a game-changer that provides a far more intuitive user experience. What do you think?</p><p><strong>Additional reading and references:</strong></p><ol><li>Adding cards on Google Pay: <a href="https://support.google.com/pay/india/answer/9673903?hl=en-GB">https://support.google.com/pay/india/answer/9673903?hl=en-GB</a></li><li>Tokenization: <a href="https://en.wikipedia.org/wiki/Tokenization_(data_security)">https://en.wikipedia.org/wiki/Tokenization_(data_security)</a></li><li>RBI Report on deepening Digital Payments (May 2019): <a href="https://rbidocs.rbi.org.in/rdocs/PublicationReport/Pdfs/CDDP03062019634B0EEF3F7144C3B65360B280E420AC.PDF">https://rbidocs.rbi.org.in/rdocs/PublicationReport/Pdfs/CDDP03062019634B0EEF3F7144C3B65360B280E420AC.PDF</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=90f2f1226f90" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How I built Wheretoeat — A Facebook Chatbot: Part 1]]></title>
            <link>https://blog.chatbotslife.com/wheretoeat-a-facebook-chatbot-to-find-places-to-eat-part-1-6a3f5e537880?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/6a3f5e537880</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[bots]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[chatbots]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Mon, 02 Apr 2018 13:56:08 GMT</pubDate>
            <atom:updated>2018-04-03T13:08:06.877Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MdWVvVEI6qmBhyYfa5p4_A.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/photos/T9pyYmue1x4?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">rawpixel.com</a> on <a href="https://unsplash.com/photos/XlYmpmNJhf8?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><h3>How I built Wheretoeat — A Facebook Chatbot: Part 1</h3><p>Chatbots became popular tech sometime in 2016. They are present in many apps and websites, and stats say they are only <a href="http://www.convinceandconvert.com/digital-marketing/6-critical-chatbot-statistics-for-2018/">expected to grow</a>. <strong>Wheretoeat </strong>is a simple Facebook Messenger Bot that I started as a side project to understand how they work. It is built on Rails, Dialogflow and the Zomato API and suggests restaurants in a given location. In this two part series, I write how to build a such a bot.</p><h4>Getting the tools ready</h4><p>I built the bot on Rails because I have previously worked on it, though you could select almost any <a href="http://valosohub.com/blog/2017/08/21/programming-languages-chatbots/">language</a>. The next step is to select an NLP platform. At the core, a bot just answers questions. What makes the responses meaningful is <strong>NLP or Natural Language Processing</strong>. These platforms can be trained to understand the input and respond in a way you want them to. Google and Facebook have popular and free platforms for this. For my project, I went with the Google one — <a href="https://dialogflow.com/">Dialogflow</a> (formerly called Api.Ai). If found that it was well documented and had a rails gem for it. After that is sorted, you need a select the messaging app and Facebook Messenger was the obvious choice here due to its reach. (Ok after Cambridge Analytica, some of you may disagree! 😨) Finally, the most important part: What should the bot do? Finding a good problem to solve will make it interesting to build.</p><figure><a href="https://businessofbots.com/san-francisco-2018/"><img alt="" src="https://cdn-images-1.medium.com/proxy/1*LeQDzsUCGnHVAO8LohatOg.jpeg" /></a></figure><h4>A problem to solve</h4><p>Every other day at work, we spent a good 5–10 minutes deciding the place to go for lunch. More often than not, we wished if this decision could be simplified. It was over one such lunch discussion, the idea for this bot came up — Get a list of popular restaurants for a given location. For the restaurant data, I used Zomato, a popular restaurant aggregator that also has an API.</p><p>Part 1 of this post shows the basics to build a conversational bot that responds to your messages. In Part 2, we integrate Dialogflow and the Zomato API to find restaurants for a given location.</p><h3>Trending Articles on Chatbot Marketing:</h3><blockquote><a href="https://chatbotslife.com/how-my-chatbot-got-100k-users-in-1-week-f66a5f4f2212">1. How my Chatbot got 100K users in 1 week</a></blockquote><blockquote><a href="https://chatbotslife.com/botxperts-asked-with-veronica-belmont-of-growbot-io-5cba22ae6027">2. BotXperts asked — with Veronica Belmont of Growbot.io</a></blockquote><blockquote><a href="https://chatbotslife.com/messenger-user-acquisition-how-to-build-facebook-comment-auto-reply-private-reply-using-php-a084bdf0834b">3. Messenger User Acquisition: How to build Facebook Comment Auto Reply</a></blockquote><h3>Part 1 — Getting the Chatbot to reply</h3><h4>Setup the Rails app</h4><p>If you have worked with rails previously, this should be known to you. If you haven’t, you could start with a basic app to know how rails works before trying this out. <a href="http://guides.rubyonrails.org/">Rails guides</a> is a good place to start. We will use the following custom gems for the app. Add them to the Gemfile.</p><p><strong>The Gems</strong></p><p><strong>a. Facebook Messenger Gem:</strong> This is undoubtedly the best for messenger. The documentation itself is enough to get you started with the bot. More details <a href="https://github.com/jgorset/facebook-messenger">here</a>.</p><pre>gem &#39;facebook-messenger&#39;</pre><p><strong>b. Api.Ai Gem:</strong> Dialogflow was earlier called <strong>api.ai</strong> and we will use the ruby SDK available as a gem. More details about the gem <a href="https://github.com/dialogflow/dialogflow-ruby-client">here</a>.</p><pre>gem &#39;api-ai-ruby&#39;</pre><p><strong>c. Figaro: </strong>This is a gem for Rails config values. There are many alternatives to this and you can search and use one that works best for you. Details about this gem is available <a href="https://github.com/laserlemon/figaro">here</a>.</p><pre>gem &#39;figaro&#39;</pre><p><strong>d. Romato: </strong>This is a Zomato API Wrapper in Ruby by Kyle Annen. Details <a href="https://github.com/kyle-annen/romato">here</a>. You can search for other gems or use the API directly also.</p><pre>gem &#39;romato&#39;</pre><h4>Configure Dialogflow</h4><p>Go to the <a href="https://console.dialogflow.com/api-client/#/login">dialogflow console</a> and sign up with your Google Account. <strong>Create a new agent</strong> for your chatbot as shown below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*po-tggxCwnrYfEMu21Hp_w.png" /><figcaption>Create a new agent for your chatbot</figcaption></figure><p>Under the ‘General’ tab of <strong>Settings, </strong>you will find the <strong>Client Access Token</strong>. We will use this key in a bit. Dialogflow uses Intents to identify what the human is saying and what action should be taken on it by the AI. This will be covered in detail in Part 2 when we implement the food search logic. There is a Default Welcome Intent in each agent. Click on that and add some training phrases and the response you want the bot to return.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TXSlP0zvWkCjOKjaTmXYxQ.png" /><figcaption>Modify the default welcome intent to match your requirement</figcaption></figure><h4>Get Server Running</h4><p>Run your rails app as usual. In the upcoming steps you need to connect your app to Facebook. To do that we need an public url. Since we are running it locally, we will create a tunnel for the local host using ngrok a free tool available at <a href="https://ngrok.com/">https://ngrok.com/</a>. Read the instructions and get it running. You will get 2 public IPs for your app like the sample shown below. Note the SSL secured one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/997/1*IDxT-pHf9tSpZnzE6HJmwA.png" /><figcaption>Ok judge me, I use Bash on Windows. 😛</figcaption></figure><h4>Setup Environment Variables</h4><p>We use three ENV tokens for Facebook and one for the Zomato API key. As these variables should not be committed to production code, the gem Figaro can be used. Usage details here — <a href="https://github.com/laserlemon/figaro">https://github.com/laserlemon/figaro</a> When running locally, you can add these variables by using the export command in your shell.</p><h4>The basics</h4><p>Next we have to setup the app to enable it to communicate with the Facebook and Dialogflow. As given in the messenger gem Readme, we need to create a listen.rb file inside a app/bot folder for it to listen to our chats from messenger. For that we need to tell Rails to load the app/bot folder in production. But first we start with the routes so that requests are received at /bot.</p><pre># config/routes.rb</pre><pre>Rails.application.routes.draw do<br>  mount Facebook::Messenger::Server, at: &quot;bot&quot;<br>end</pre><p>Start by creating a facebook_messenger.rb file inside the config/initializers folder. In the next steps we will create a folder bot</p><pre># config/initializers/facebook_messenger.rb</pre><pre>unless Rails.env.production?<br>bot_files = Dir[Rails.root.join(&#39;app&#39;, &#39;bot&#39;, &#39;**&#39;, &#39;*.rb&#39;)]<br>  bots_reloader = ActiveSupport::FileUpdateChecker.new(bot_files) do<br>    bot_files.each { |file| require_dependency file}<br>  end<br>  ActiveSupport::Reloader.to_prepare do<br>    bots_reloader.execute_if_updated<br>  end<br> bot_files.each { |file| require_dependency file }<br>end</pre><p>After that is done, create a file listen.rb inside the bot folder. This will contain the action to perform when the bot listens to a message.</p><pre># app/bot/listen.rb</pre><pre>require &#39;facebook/messenger&#39;<br>include Facebook::Messenger</pre><pre>Facebook::Messenger::Subscriptions.subscribe(access_token: ENV[&#39;VERIFY_TOKEN&#39;])</pre><p>The listen code does a couple of things here. First is the <strong>Facebook:Messenger::Subscriptions</strong> portion. This is an authentication step with Facebook to verify that our app is authorized to talk to Facebook and vice versa. To get this done, we need to set up a Facebook Developer account and a page for your app.</p><h4>Setting up Facebook</h4><ul><li><strong>Create a Facebook Page:</strong> This is fairly straightforward and well explained. From your Facebook home page, click the drop down arrow on the top bar and select ‘Create Page’. Enter details like name, address, category to setup the page.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/340/1*Pqf5y1L5vih-U9_qOVEE6A.png" /><figcaption>Create a Facebook Page</figcaption></figure><ul><li><strong>Create the App:</strong> Next, go to the <a href="https://developers.facebook.com/apps">Facebook Developer Console</a> and click on ‘<strong>Add a New App</strong>’. Enter the basic details and click on ‘Add Product’ in the app home page as shown below.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xjLK8lfCBdf9_yERTlgftg.png" /><figcaption>Setup Messenger from the New App page</figcaption></figure><p>After creating the app, add the product ‘Messenger’. Click on Set Up. This will take you to the messenger app page. First generate a token for the page by selecting the new page you have created in the step above. Keep this code.</p><p>Next, to setup Webhooks click on ‘Setup Webhooks’ . Enter the https URL we generated using ngork followed by /bot in the validation URL text box. Example: <a href="https://xyz.com/bot">https://xyz.com/bot</a>. Enter the Verify Token (a text you will use in both your app and Facebook to validate the app. ) and select <strong>messages, messaging_postbacks and messaging_optins. </strong>Click on Verify and Save to complete.</p><h4>Coming back to the App</h4><p>We have previously added the subscription part in the listen file. Now we write the code for the app to listen to all incoming messages. This is the prescribed syntax for the gem. More details in the gem page.</p><pre># app/bot/listen.rb</pre><pre>require &#39;facebook/messenger&#39;<br>include Facebook::Messenger</pre><pre>Facebook::Messenger::Subscriptions.subscribe(access_token: ENV[&#39;ACCESS_TOKEN&#39;])</pre><pre>Bot.on :message do |message|<br>  request_message = message.text</pre><pre>unless request_message.nil?<br>    chat_service = ChatService.new(uid)<br>    chat_service.execute(request_message)<br>    FacebookMessengerService.deliver(uid, chat_service.response_message, chat_service.quick_replies, chat_service.response_template)<br>  end<br>end</pre><p>As you can see here, we are creating two new service classes inside the services folder in the app directory.</p><p><strong>ChatService</strong></p><pre># app/services/chat_service.rb</pre><pre>class ChatService<br>  attr_reader :response_message</pre><pre>def initialize(uid)<br>    @client = ApiAiRuby::Client.new(<br>      client_access_token: ENV[&#39;API_AI_CLIENT_ACCESS_TOKEN&#39;],<br>      api_session_id: uid<br>    )<br>  end</pre><pre>def execute(message)<br>    api_ai_response=<a href="http://twitter.com/client">@client</a>.text_request(message)  <br>    api_ai_response_message=api_ai_response[:result][:fulfillment]        <br>    [:speech]   <br>    @response_message = api_ai_response_message<br>  end<br>end</pre><p>Here we initialize the Class using the client access token generated in Dialog Flow. The output of Dialogflow is a JSON from which the [:result][:fulfillment][:speech] hash has the text response from it. We use the response from it and send to FacebookMessengerService</p><p><strong>FacebookMessengerService</strong></p><pre>class FacebookMessengerService<br> class &lt;&lt; self<br>  def deliver(uid, message, quick_replies=nil, template=nil)<br>    if quick_replies.present?<br>     quick_replies = quick_replies.map {|replies| replies ==  <br>      &#39;Location&#39; ? {content_type: &#39;location&#39;, payload: &#39;location&#39;} : <br>      {title: replies, content_type: &#39;text&#39;, payload: <br>      &#39;test_payload&#39;}}<br>    end<br>    message_content={<br>     text: message,<br>     quick_replies: quick_replies ? quick_replies : nil<br>    }<br>   end<br>   begin<br>    Bot.deliver(<br>    {<br>     recipient: {<br>      id: uid<br>     },<br>     message: message_content<br>     },<br>     access_token: ENV[&#39;ACCESS_TOKEN&#39;]<br>     )<br>   rescue Exception =&gt; e<br>    puts &#39;Error&#39; + e.message<br>   end<br>  end<br> end<br>end</pre><p>This service returns the reply to Messenger. We are using a feature of Messenger called <a href="https://developers.facebook.com/docs/messenger-platform/reference/send-api/quick-replies">Quick Replies</a> here that allows to send short clickable buttons as response to some text. It will be used to get the location to search for places to eat if the user has not input it upfront. and will be explained in detail in the Part 2.</p><p>After this is done, go to messenger, search for the Facebook Page (Where To Eat in my case) and enter any of the welcome intent phrases you created. You should get something like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/759/0*uWEPn7ugC57NddJ_.gif" /></figure><h3>What will we do in Part 2</h3><p>In the next part, we will use the Dialogflow feature of locations and cities to get the area the user is searching for. In case a user does not provide a location, we will use the send location feature of messenger to get a location. Then, we integrate Zomato API to search in that location and return restaurants in that area. Here is a preview:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3uMudoOS9SZS7v4LaYi3eA.png" /></figure><p><strong>P.S:</strong> I don’t code for my day job, atleast not anymore. Comments welcome on what could have been done better.</p><p>Loved it so far? Click the 👏button below to show your appreciation. Share if you found it useful. Connect with me on <a href="http://garythomasgeorge.com/"><strong>garythomasgeorge.com</strong></a></p><h4><strong>Part 2 coming soon!</strong></h4><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2F3e59ef%3Fas_embed%3Dtrue&amp;dntp=1&amp;url=https%3A%2F%2Fupscri.be%2F3e59ef&amp;image=http%3A%2F%2Fapi.screenshotlayer.com%2Fapi%2Fcapture%3Faccess_key%3Dfe59908dad3baab69ffab249a2224b03%26viewport%3D1024x612%26width%3D1000%26url%3Dhttps%253A%252F%252Fupscri.be%252F3e59ef%253Fscreenshot&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/7078d8ad19192c4c53d3bf199468e4ab/href">https://medium.com/media/7078d8ad19192c4c53d3bf199468e4ab/href</a></iframe><figure><a href="https://chatbotslife.com/bot-communities-mastermind-group-d2dae9876709#.53x0py6ou"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*6XUspT9JOSq0w0Fi35HIaA.png" /></a></figure><figure><a href="https://m.me/ChatbotsLife"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*c1LDMH5vbnIz9rmAka8Hwg.png" /></a></figure><figure><a href="https://chatbotslife.com/how-to-get-a-free-chatbot-b1fb9dfe109#.z9dtp2sy0"><img alt="" src="https://cdn-images-1.medium.com/max/255/1*D0Jf3dI6ZThtqcfwDYY7mg.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6a3f5e537880" width="1" height="1" alt=""><hr><p><a href="https://blog.chatbotslife.com/wheretoeat-a-facebook-chatbot-to-find-places-to-eat-part-1-6a3f5e537880">How I built Wheretoeat — A Facebook Chatbot: Part 1</a> was originally published in <a href="https://blog.chatbotslife.com/">Chatbots Life</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Kitchen to the Stomach: Why last mile delivery matters]]></title>
            <link>https://garythomasgeorge.medium.com/kitchen-to-the-stomach-why-last-mile-delivery-matters-d0d33655747?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0d33655747</guid>
            <category><![CDATA[zomato]]></category>
            <category><![CDATA[food-delivery]]></category>
            <category><![CDATA[product]]></category>
            <category><![CDATA[logistics]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Sun, 17 Sep 2017 17:00:22 GMT</pubDate>
            <atom:updated>2017-09-17T17:00:22.164Z</atom:updated>
            <content:encoded><![CDATA[<p>Earlier this week, <a href="http://www.livemint.com/Companies/W8wu28THxovZpVYufEvrnL/Zomato-acquires-delivery-startup-Runnr.html">Zomato announced</a> it had acquired delivery startup — Runnr. With this acquisition, Zomato has added a much needed missing link in its food delivery solution— locating where your food is. Why is this important?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6yxDEI_i1aNOvSigscMqZg.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/photos/KSQgzzn3dW0?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Maarten van den Heuvel</a> on <a href="https://unsplash.com/?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><h3>The advantage of owning Logistics</h3><p>The time taken for a delivery boy to leave with your order and reach your hands can be upto 50% of the total time you wait for your food. Assuming you can wait about 30–45 minutes for your food, it leaves about 15–20 minutes for delivery. If it takes any more than that, the restaurant should be placed closer to you or not deliver to you in the first place. In most cases, it will be less than that for you to get your food delivered hot (or cold) and tasty just the way it was meant to be. There are two problems solved when the company has its own logistics:</p><h4>1. Order Tracking</h4><p>Over the past 2 years, I have been ordering on major FoodTech apps in Bangalore. Zomato was (and is still) my goto app for food delivery from the beginning. One of the major problems with Zomato orders though was my inability to know where the food was after leaving the restaurant and their inability to know if I had received my food. Order tracking becomes possible when the company does not depend on the restaurant or a third party for food delivery</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KyOERE4LY0-dwwWHM8y0SA.jpeg" /><figcaption>Order Tracking in Zomato vs Swiggy</figcaption></figure><p>This is why I said the Runnr acquisition is a major addition to the Zomato armory against market leader Swiggy. But that is not the only reason. If you really think of it, then is it that important to know the location of your food unless it is not delivered on time? That brings me to the other reason.</p><h4>2. Faster Delivery Times</h4><p>When there is control over the last mile of the supply chain, it is able to ensure the order reaches faster without having the delivery boy call you repeatedly for instructions to reach your home. In a place like India where lanes are not marked properly and address numbers have no particular order in most places, this is even more important. Owning the logistics ensures that a company can use its own tech stack to map out locations and ensures delivery times are reduced. For an industry running on ‘wafer’ thin margins, more orders delivered in the same time means more commission. Faster delivery translates to repeat orders and loyalty.</p><blockquote>Won’t you think twice before ordering from an app or restaurant that kept you hungry late into the night?</blockquote><h3>The Economics of it</h3><p>Food Delivery is a costly business. There are a lot of variables in terms of individual tastes and preferences, low margins, volumes etc. In 2016, we saw how owing to low entry costs, many companies started in food delivery and soon shut shop also. For each order, companies shell out on an average of Rs. 50–60. The low <strong>Average Order Value</strong> <strong>(AoV)</strong> in India of about Rs. 300 compared to 5 times the amount in the US market means there is very less to make in terms of commission. CAC as discounts and promotions take up a large part of the commission after mandatory costs. Add delivery costs to that and you are left with nothing. In order to make money Zomato will probably charge customers for lower value orders or for peak times like Swiggy. They have <a href="http://blog.zomato.com/post/142067240706/two-million-meals-on-heres-what-weve-learned">written about this</a> in the past also and probably have taken this into account. Runnr also aims to be also be in other delivery businesses while servicing Zomato. How that plays out only time will tell.</p><p>For me though, I am happy at least there will be no more of those annoying ‘<em>Have you received your order</em>’ notifications!!!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*39NSmFOC973515iHA6szdg.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d0d33655747" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6 months as a first time Product Manager: My 5 lessons]]></title>
            <link>https://garythomasgeorge.medium.com/6-months-as-a-first-time-product-manager-5-lessons-gary-434ab35b2879?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/434ab35b2879</guid>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[education-technology]]></category>
            <category><![CDATA[learning]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Mon, 20 Feb 2017 18:31:18 GMT</pubDate>
            <atom:updated>2017-02-21T07:56:56.292Z</atom:updated>
            <content:encoded><![CDATA[<h3>6 months as a first time Product Manager: 5 lessons</h3><p>About 6 months ago I moved into Product Management in the <a href="http://www.foradian.com/">EdTech company</a> I have been working for the past 4 years. For a role that has no ‘one definition’ and where job profiles vary from company to company, the transition was like entering uncharted territory, specially when as the first formal Product Manager.</p><p>Before moving into the role, one of the major sources of learning was reading the blogs on Product Managers — from <em>how they work</em> to <a href="https://www.productmanagerhq.com/2014/06/day-in-the-life-of-a-product-manager/"><em>what they do on a typical day</em></a>. 6 months down the line, each day has been a different experience for me, knowing fully well that there is so much more to learn and improve. It is in this context that I thought to share my lessons from these past 6 months.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5-q3ijsR7d3k47-UU8vkKQ.jpeg" /></figure><h3>My 5 Lessons</h3><h4>1. Look beyond Roadmaps</h4><p>Creating and managing roadmaps is probably the first thing you will find in any PM job description. While it is good to plan ahead for your product, be ready to drop everything or some part of it and start afresh from day one. Roadmaps are good but not sacrosanct. There are lots of things that can bring you back to the drawing board. As long as you realize that the change is for the better, there is nothing wrong in having features that remain in your google drive as a document.</p><p>One feature that I had been involved in perfecting over the year and particularly in the last 6 months may have to be redone due to statutory changes. This means dropping some features and accommodating the immediate change required. It doesn’t matter.</p><h4>2. Keep reading</h4><p>In my almost 1 year before moving to this role, I helped with roadmap planning, understanding what our users want and working with Design and Engineering to make it ready in a way that solves their problem without the official title. What I also did was a lot of reading. Reading up how other companies work, how to solve user problems etc. <strong>DO NOT stop that</strong>. Even after you have timelines to meet and decisions to make, let the reading be a part of what you have do regularly. It gives you ideas and opens up your mind to a whole lot of things that you may not come across in your work. There are a lot of <a href="http://www.cleverpm.com/2015/09/30/a-product-management-reading-list/">resources online</a>, <a href="https://medium.com/@sicross/the-product-manager-s-essential-reading-list-for-2016-c7fd4c0491bf#.xnyu1y9rj">Medium articles</a> and <a href="https://www.quora.com/What-are-the-best-books-for-Product-Managers">Quora answers</a> on what product managers should read.</p><p>Some regular subscriptions to my inbox every week are <a href="http://firstround.com/review/">First Round Review</a>, <a href="https://www.productmanagerhq.com/">Product Manager HQ</a>, <a href="https://blog.intercom.com/">Inside Intercom</a> and a whole lot more.</p><h4>3. Acknowledge mistakes and correct them</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zDNIoiebyTWt-YE7n6WyVg.jpeg" /></figure><p>Mistakes are fine. I have made many small and some not so small ones in these 6 months, but make sure you admit them and take corrective measures. Make sure you don’t repeat them because each mistake is a lesson. With practice they will come down and I know six months are not enough for that.</p><p>One incident that comes to mind is how I agreed to a request from some users to have a particular transaction page auto populated with the amount to be paid. It seemed logical since it would mean time saved in entering values. After the feature went live, many other users started complaining that they did not pay the complete amount so it meant they had to first remove the values and enter again. Soon I realized that I had not considered this use case and therefore had it corrected.</p><h4>4. Remember you are not doing it for yourself</h4><p>Eventually, everything you do is for the user. It is essential to keep that in mind. As the voice of the customer you are there to ensure that it is build the way the user wants it. But, as a first time PM, you can be carried away to do all that you find missing in the product. Some of it may be correct but many of them may have not been done for a reason. When in doubt — ask(your manager, the product head etc.)</p><p>It is important you understand what can be done and what needs to be done and draw a clear line. You may want to have a lot of new features but does the user want it? Your engineering team may want to build a lot of things but does the user want it?</p><h4>5. Understand the business</h4><p>I am fortunate to have been in a position to understand the business of the company I work in. This might be easier in a start-up where one is exposed to the business daily than a bigger firm where your function maybe just a small component of the whole business. Whatever the size of the firm, you can make an effort to understand what brings contributes to the top-line. Why is this important? Because — cash runs the business. How do you do that?</p><p>You can start by becoming friends with the sales and marketing teams and try then to understand what brings in the revenue. Going further, it will also help to learn up some financial and business basics. This will greatly help you understand why certain decisions are taken at the C-Level.</p><h3>And finally</h3><p>These lessons are based on my experience as a first time PM at a startup building <a href="http://www.fedena.com/">educational web apps</a>. Your experiences may have been very different in the first 6 months. Though different, each one matters because every experience is a new learning so write about your experience.</p><p>Liked what you read? Hit the green like button below. Share you comments.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=434ab35b2879" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Instagram more like Snapchat: The Levi’s Effect in Social Media]]></title>
            <link>https://garythomasgeorge.medium.com/instagram-more-like-snapchat-the-levis-effect-in-social-media-de6691fa51e4?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/de6691fa51e4</guid>
            <category><![CDATA[instagram]]></category>
            <category><![CDATA[branding]]></category>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[snapchat]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Thu, 11 Aug 2016 16:01:01 GMT</pubDate>
            <atom:updated>2016-08-11T16:01:01.313Z</atom:updated>
            <content:encoded><![CDATA[<h3>Instagram Stories and Snapchat: The Levi’s Effect in Social Media</h3><p>For sometime now I was seeing that my Facebook feed was having less of stories about ‘people’ and more news stories. There were stories about Donald Trump saying something than by friend Donald visiting India for the first time. Okay, I don’t have a friend Donald on Facebook but you get the point!. Instagram launched stories recently and this brought me to this headline.</p><p>There was an article on similar lines about how Facebook was worried that people were sharing less on Facebook. This was ironic for a social network that became popular connecting people through likes and shares. People are still sharing no doubt, probably more than before, but what is concerning is the lack of original content. There are more <em>cat videos</em> (and now GIF’s) shared than ‘<em>Birthday and Anniversary</em>’ Photos.</p><p>This has Facebook worried because personal connection was one of the main reasons for its hyper growth. Without original content there could emerge another media app that will take away a chunk of their users. It might sound alarmist but in the ever changing world of tech and business you never know.</p><h3>The rise of Snapchat</h3><p>When Snapchat came up 4 years ago, it instantly captured the attention of a large population of millennials — or maybe the early Generation Z. In part, it was due to their focus on photos (and later videos) that disappeared after a single view but also in part due to the fact that this new medium had all their friends but not their family. This meant that Snaps provided a medium for tweens to share content with their friends without getting that awkward comment by their uncle or aunt who just got introduced to social media. Facebook which <em>snapped up</em> (pun intended) Instagram and Whatsapp for record valuations tried to buy Snapchat also in 2013, because it understood this early on. In retrospect, this spurn seems to have paid off well since Snapchat has not only shown huge prospects for advertisers but also grown its user base to late millennials.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gIkHI1o5aSliPIwGdtqMIA.png" /></figure><h3>Instagram Stories — Avoiding the Levi’s Effect</h3><p>When Instagram debuted its brand new update — Stories, it was another attempt by the parent company Facebook to retain its appeal to a large population of youngsters that seems disillusioned by a social network which is itself entering teenage next year. This new update allows users to upload photos or short videos that last just 24 hours and do not appear on the timeline. Wait a minute! Isn’t that what Snapchat does? Yes, they have copied the entire playbook off Snapchat. And by the first looks, seem to have done it better.</p><p>This brings me to the reason for the title of this Many years ago Levi’s which was a hip brand among youngsters for a long time faced a similar problem. A generation that wore these jeans kept wearing them after growing older. This made their children to look for newer, more younger brands that appealed to them and even jeans itself. Remember the baggy pants from the 90’s? Levi’s tried hard to shed this image and bring back their youth appeal. In fact marketeers have faced this problems in many brands.</p><p>Facebook is also at a similar situation but from the other side. When parents started joining Facebook to connect with their old friends the children started looking for alternatives. Many brands have faced and some have come out it successfully also. Levi’s again is a good example. For a brand that is over 160 years, by re-inventing itself and staying on top of the latest trends Levi’s has mastered ‘staying power’ better than anyone else. And the best part it has not compromised on core brand values.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KP2NNz__cShm5McKIDnOhw.jpeg" /></figure><p>With <strong>Stories</strong> and many other similar changes, Facebook is trying to do just that. Remain relevant. With the large user base it already has, Instagram Stories might end up beating Snapchat at its game in the long run. They also seem to have come up with a more polished offering that attracts to its current user base. After all there is only so much that one can get through the First Mover Advantage. Or maybe not! The jury is still out.</p><p>Whatever the case, these kids who are uploading morning selfies with cat filters are soon going to join the workforce and bring with them with big purchasing power. And there will be brands vying to spend ad dollars on catching them. Facebook has to be there to get a share of this money and knows that this is a game it cannot afford to loose. <strong>Snapchat … your move</strong>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=de6691fa51e4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building software with Empathy]]></title>
            <link>https://blog.prototypr.io/building-software-with-empathy-62307c0f5ecd?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/62307c0f5ecd</guid>
            <category><![CDATA[product-development]]></category>
            <category><![CDATA[edtech]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[empathy]]></category>
            <category><![CDATA[user-experience]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Fri, 01 Jul 2016 07:46:01 GMT</pubDate>
            <atom:updated>2016-07-01T09:34:09.386Z</atom:updated>
            <content:encoded><![CDATA[<h4>Notes from an EdTech product team</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1laejc11GP81Zi83LWvsWg.jpeg" /></figure><blockquote>Don’t make a better [x], make a better [user of x] — Anonymous</blockquote><p>This quote sums up why this word empathy, has gained enormous relevance in software development in the past few years. For every team in a software company, be it product, design, development or customer support, empathy for the user should be a fundamental guideline on which the product is built.</p><p>Building a product according to the needs of the customer has its origins in the late days of the mass production. In those days what the factory produced was ‘consumed’ by the users and all products were without variation. The initially successful Ford Model T is an infamous example of this. When General Motors introduced newer models with better features, they were, in fact, responding to the needs of the customers who wanted more than just an ‘engine on 4 wheels’ that was faster than a horse.</p><h4>Building products that users want to use</h4><p>Though software products have existed for many decades now, over the past few years there has been a change in how they are built. Companies have moved from <a href="http://itlaw.wikia.com/wiki/Feature_bloat">feature bloat</a> to building things that users want to use and not forced to use. Responding to the needs of the users or risk facing obsolescence (to use the Ford analogy) has been a major reason for this change.</p><p>This change put the focus on the need for user empathy in software development. Empathy, which has been for a long time associated with design-led thinking and user-centered product management, can infact be used by all teams that are in the process of building software.</p><h4>What is building with Empathy?</h4><p>Apps of the Internet Age are now increasingly giving importance to solving real problems faced by users. Technology which was playing the role of a differentiator is becoming more of a facilitator in this journey. Empathy is not very different from words like <em>Experience</em> or <em>Engagement</em> that you have often heard associated with software development. Put simply, it means to bring the user to the center of any product decision you are having. Like I previously mentioned, they can hence be applied to any team that is building it and not only design and product management as traditionally known. Whether it is a small validation script that the developer has written to avoid unwanted input or a new test case by the QA member, every team can contribute to ensuring that the product is built with the user in mind.</p><p>For a product that is used in over <a href="http://www.fedena.com/">40,000 schools and colleges</a>, we have building apps with emphasis on the user. From feature specs to code and test cases, <em>our philosophy at Fedena has always been to simplify tasks for educators so that they can focus on what they do best — educate our future generations</em>. This approach has helped us with making many of the decisions for Fedena in recent times. However, empathy is much more than that and when we first decided to take this approach we realized easier when starting from scratch. Something that is working just fine is in a state of inertia and it remains in that state — ‘If it ain’t broke, don’t fix it<em>’</em>. Then we made a conscious decision to take up the empathy-driven development approach for all feature enhancements one at a time.</p><blockquote>A student management system has the first admission enquiry from a student to his leaving certificate and everything else in between.</blockquote><p>The last release — version 3.5 was the first time we applied this approach comprehensively. It took the efforts of all departments — from UX to Development and QA to get started. If you have used the new version, you will be able to spot the difference in many areas. Implementing this was not easy and we are not there yet completely but here<em> are 3 takeaways from this exercise so far</em>:</p><ul><li><strong>Put the user in the center of your product decisions:</strong> Often the decision that product teams have to take is how much is enough for a feature. Have we considered all scenarios? Is this required? An easy as well as difficult answer to this is to think as the user who will be using this feature. Easy because it gives you a canvas on which you have to take decisions, difficult because the canvas will often have many unknowns. However, segmenting and identifying your core user base will help give you the right direction to start with.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N-6vdg-eP1w_5Femejg5pA.jpeg" /><figcaption>Put the user at the ‘center’ of your product decisions</figcaption></figure><ul><li><strong>Empathy starts and ends with the user:</strong> This is a given. Once you have this canvas you can start planning user stories around this. Be prepared to go further because there will be always something you could do more. If you have not thought through all the ways the user will be using the feature or product then you have not justice. An example that comes to my mind is from the CCE and ICSE examination system that we enhanced in v3.5. In this, the final result of an exam was calculated by schools in different ways. We brought in formulas like summation, difference and average as options. Then a teacher who saw the prototype said <em>‘Hey, the formulas are fine but we add up the final results is different’</em>. This made us bring in two modes to handle the two different types of calculations.</li><li><strong>Be empathetic but not too empathetic:</strong> It is a corollary of the first two points but is equally important in your journey. You cannot build something that satisfies everyone. There will always be some features and some users where you have to say ‘no’. If you are able to define what you are as a product and what you are ‘not’ then this problem can be avoided to a great extent. You can read more about this in <a href="https://blog.intercom.io/product-strategy-means-saying-no/">Product Strategy Means Saying No</a>. It will also help keep feature creep away and over time only help your users by providing the features that they want to use.</li></ul><p>Finally, you need to keep in mind that as with any other activity you undertake, there will be some quick results and there will be some that take time. The key here is to make this a habit and carry on. Once your teams start approaching your users with empathy, you will not only build better products but also make better users of your products!</p><p>As for us, we plan to continue with this approach in <a href="http://www.fedena.com/">Fedena</a> and keep learning. If you are interested, you can <a href="http://www.fedena.com/blog/2016/05/8-reasons-to-be-excited-about-fedena-erp-in-2016-rails-5-0-analytics-and-more.html">read about what are we working on</a>. Have you tried empathy driven product development? What are your experiences?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=62307c0f5ecd" width="1" height="1" alt=""><hr><p><a href="https://blog.prototypr.io/building-software-with-empathy-62307c0f5ecd">Building software with Empathy</a> was originally published in <a href="https://blog.prototypr.io/">Prototypr</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Netflix in India and why the Hotstar model makes more sense for Indians.]]></title>
            <link>https://garythomasgeorge.medium.com/netflix-in-india-and-why-the-hotstar-model-makes-more-sense-for-indians-3ebd97d54d57?source=rss-9314c115d166------2</link>
            <guid isPermaLink="false">https://medium.com/p/3ebd97d54d57</guid>
            <category><![CDATA[advertising]]></category>
            <category><![CDATA[netflix]]></category>
            <category><![CDATA[movie-streaming]]></category>
            <dc:creator><![CDATA[Gary Thomas George]]></dc:creator>
            <pubDate>Mon, 15 Feb 2016 18:45:12 GMT</pubDate>
            <atom:updated>2016-02-16T12:43:55.728Z</atom:updated>
            <content:encoded><![CDATA[<p>So it’s been more than a month since Netflix launched in India and signed us all up with their free one-month subscription. By this time, most of you must have cancelled it. If you haven’t, don’t worry, you are in company, you can do it next month.</p><p>Netflix was supposed to be a godsend for urban Indians when it launched last month but soon there were complaints of its high data usage (<a href="https://help.netflix.com/en/node/87?ba=GSButtonClick&amp;">upto 3 Gb for an hour of HD</a>) something that was not suitable for our limited data plans. Then were the issues of speed itself where the buffering makes up for the time you would spend on TV watching ads. So on one side there is the Rs. 500 subscription and on the other the free albeit illegal torrent that we so dearly love. So it begs the question why should one pay for a data connection and Netflix. This is where I feel that the Hotstar model of showing content with ads is more a better match for India.</p><p>Hotstar is a streaming service by the Rupert Murdoch-owned News Corp that streams content by its Star Network. With a plethora of channels (48 as per Wikipedia) producing in 8 languages, they do not have a problem of content or pan India reach. With the app, they are just replicating what they used to do in YouTube only better with the revenue going directly to them. They still continue to upload content on YouTube thoAugh. Hotstar is not the only one doing it, there is the Eros app, TVFPlay from The Viral Factory and many others. The ad model ensures there are no subscription charges also. Music streaming services like Gaana from the times network have a similar ad-supported version along with a premium ad-free one. In fact, Gaana and its likes have brought down illegal music downloads significantly among the Tier 1 city smartphone carrying youth who access to significantly cheap and high-speed data throughout most of the day.</p><p>This is the same age group that will be Netflix’s target audience and unless they opt for a lower price or a different model this target audience would still prefer torrents to streaming. Ads on Netflix might be a <a href="http://qz.com/416907/people-are-freaking-out-about-ads-coming-to-netflix/">huge no go area</a> for most loyal users of the service but for India where they are just starting it could be a way out. The other could be to continue it as a <a href="http://timesofindia.indiatimes.com/tech/tech-news/Netflix-is-for-premium-customers-in-India-says-company/articleshow/50809626.cms">premium service</a> for a group of English speaking credit card holders from India. This group, after all, is a growing one with more money to spend and one that prefers Orange is the New Black to a <a href="https://en.wikipedia.org/wiki/Prem_Ratan_Dhan_Payo"><em>Receive a Treasure Called Love</em></a><em> </em>movie.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3ebd97d54d57" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>