<?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 Jeff Garrett on Medium]]></title>
        <description><![CDATA[Stories by Jeff Garrett on Medium]]></description>
        <link>https://medium.com/@jeff2invest?source=rss-2f130afc9b93------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*bvEZoLv6_DvVKg9gUV6BtQ.jpeg</url>
            <title>Stories by Jeff Garrett on Medium</title>
            <link>https://medium.com/@jeff2invest?source=rss-2f130afc9b93------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 23:56:24 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jeff2invest/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[Pivoting Product at AI Speed]]></title>
            <link>https://medium.com/@jeff2invest/pivoting-product-at-ai-speed-815955d7a2e6?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/815955d7a2e6</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[engineering-leadership]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Mon, 04 May 2026 20:48:49 GMT</pubDate>
            <atom:updated>2026-05-04T21:07:06.033Z</atom:updated>
            <content:encoded><![CDATA[<h4>Pivot product faster than transformation ever could.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b7hJcrAjaz3PyKOYx-pw6w.png" /></figure><p>Talk to your customers! Drive adoption! Own the metrics! Be the voice of the paying user! Statements every PM has heard a thousand times. Meanwhile they’re writing PRDs, updating personas, and scoping features from a team brainstorm. None of which involved a paying customer.</p><p>Your engineers shipped four experiments and talked to two customers this sprint. They have a sharper picture of the user than your product team does. The gap is real, and you can feel it widening every sprint.</p><p>You don’t run product. You can’t make them change. So what do you do?</p><p>Most engineering leaders reach for the same move: transformation. Decks. Re-orgs. Offsites. New operating models. RACI charts.</p><p>Don’t. Here’s why.</p><h3>Transformation almost always fails</h3><p>The data has been the same for fifteen years:</p><ul><li>70% of transformation efforts fail (McKinsey)</li><li>88% of business transformations miss their original ambition (Bain, 2024)</li><li>96% of organizations are in some phase of change at any given time (KPMG)</li></ul><p>That last number is the kicker. Most companies are constantly transforming. Almost none of them succeed.</p><p>When you tell people the way they’ve been working is wrong, they get defensive. Defensive people produce the same artifacts as before with new vocabulary on top. The roadmap gets re-titled “outcomes-based.” The persona doc gets a “willingness to pay” field. The same six-week discovery cycle gets renamed “lean validation.” Nothing changes. The decks just get prettier.</p><p>You don’t need an announcement. You need a different intervention entirely.</p><h3>Behavior changes through three things only</h3><p>I’ve come to the conclusion that team behavior, engineering or product or anywhere else, is downstream of three inputs. That’s it.</p><ul><li>The artifacts they produce</li><li>The questions they get asked</li><li>The metrics they own</li></ul><p>Change those three quietly and behavior shifts within a sprint. Try memos, training, or all-hands speeches and you’ve entered a coin flip you’ll lose 70% of the time.</p><p>Your product team isn’t slow because the people are bad. They’re slow because the inputs they’re working from are five years old.</p><h3>What the swap looks like</h3><p>Not a transformation. A substitution, done at room temperature.</p><h3>Artifacts</h3><p>Hand them v2 templates.</p><ul><li>Persona becomes “Buyer &amp; User Profile” with willingness to pay, churn triggers, expansion triggers, top 3 jobs-to-be-done</li><li>Feature spec becomes “Revenue Hypothesis” with adoption metric, monetization path, leading indicator</li><li>Roadmap becomes outcomes-based (revenue, adoption, retention), not feature lists</li></ul><p>You’re not killing the old work. You’re handing them a v2 that quietly makes v1 stop getting requested.</p><h3>Questions</h3><p>Insert four standing questions into the meetings that already exist:</p><ul><li>What’s the revenue or retention hypothesis?</li><li>Who pays for this and why now?</li><li>What did three customers tell you this week?</li><li>What does the adoption dashboard say?</li></ul><p>Don’t accept “let me get back to you” three weeks running. By sprint two, prep changes on its own.</p><h3>Metrics</h3><p>Pick one PM. Hand them ownership of one number. Activation, retention, expansion revenue, anything. Weekly presentation, one insight, one action. The moment a person owns a metric, they stop thinking features and start thinking business. It requires no permission to spread once one person is doing it.</p><p>Three substitutions. No announcement. No deck. No offsite.</p><h3>Doing this without authority</h3><p>You don’t run product. You don’t get to issue templates or assign metrics. So how does this work?</p><p>You build the artifacts yourself. You walk into a 20-minute meeting with whoever runs product and frame it as: “here’s where I think modern PM teams are heading. Would it help if I drafted a v2 for us to look at?” Not as a critique. As a contribution.</p><p>Most heads of product are busier than they want to admit. They will gratefully take pre-built artifacts off your hands. Their political capital. Your strategy.</p><h3>Why this can’t wait</h3><p>The standard advice (wait, build alignment, get a mandate) is exactly why 70% of transformations fail. Mandates take quarters. Alignment is what you’re trying to manufacture. Waiting is the most expensive thing you can do right now.</p><p>The pace data tells you why:</p><ul><li>AI fluency demand in job postings up 7x in two years (McKinsey)</li><li>75% of workers already using AI on the job (Microsoft Work Trend Index)</li><li>80% are bringing their own AI tools in when their company hasn’t formally adopted any</li><li>67% of business leaders say they wouldn’t hire someone today without AI skills</li></ul><p>The product talent market has already repriced itself. AI-fluent PMs are getting hired. Legacy PMs are getting filtered out. If your team isn’t pivoting, the gap compounds every quarter, until it’s uncloseable without firing the whole team and starting over. Which, unfortunately, is how most companies actually solve this. They wait too long, decide it’s hopeless, and clear-cut the org.</p><p>That’s the failure mode you’re trying to avoid.</p><h3>The 48-hour move</h3><ul><li>Tomorrow morning. Draft a one-pager. Just the four standing questions and a v2 persona template.</li><li>Tomorrow afternoon. 20-minute meeting with whoever runs product. Walk in with the one-pager. Frame as modernization, not correction. Hand it over.</li><li>End of week. Find the PM quietly asking better questions than the rest. Ask them, casually, if they’d want to pilot weekly ownership of a metric. Make their work visible.</li></ul><p>Real behavior change takes 30 days. Visible motion takes 48 hours. That’s the game.</p><h3>Conclusion</h3><p>The transformation industry has spent fifteen years selling executives that the bigger the intervention, the higher the chance of success. The data says the opposite.</p><p>The PMs who pivot will be doing the most leveraged work in the company within a year. The ones who don’t will be the bottleneck. You can’t choose for them. But you can set the conditions under which they choose. That’s what an engineering leader’s actual leverage looks like in 2026.</p><p>Companion to <a href="https://medium.com/@jeff2invest/ai-didnt-kill-product-management-it-exposed-it-7c2c3ad60ea7">AI Didn’t Kill Product Management. It Exposed It.</a> Same shift. Same leverage. Just from the other side of the wall.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=815955d7a2e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Didn’t Kill Product Management. It Exposed It.]]></title>
            <link>https://medium.com/@jeff2invest/ai-didnt-kill-product-management-it-exposed-it-7c2c3ad60ea7?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/7c2c3ad60ea7</guid>
            <category><![CDATA[organizational-change]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[product-management]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Mon, 04 May 2026 20:43:17 GMT</pubDate>
            <atom:updated>2026-05-04T21:05:18.418Z</atom:updated>
            <content:encoded><![CDATA[<h4>Refocusing product: Business, Customer, User</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z_qV-uYA-U4txYZRmRkj6A.png" /></figure><p>Let me say something that’s going to upset some people.</p><p>If your product manager can’t tell you how the company makes money on the thing they’re shipping this quarter, they’re not a product manager. They’re a translator. And translation is the cheapest job in software right now.</p><p>For two decades, the PM role has been a strange hybrid. Half strategy, half stenography. The work that actually mattered kept getting squeezed by the work that was easier to measure. Writing specs. Owning roadmaps. Running ceremonies. Shipping personas with hobbies and pet names. Meanwhile the real work, figuring out what people would pay for, watching how they used the thing, learning what made them stay, kept getting pushed aside.</p><p>Guess which half AI just ate?</p><h3>The Spec Factory Was Always a Lie</h3><p>Here’s the part nobody likes to look at directly.</p><p>Pendo studied feature adoption across hundreds of cloud products and found that 80% of features in the average product are rarely or never used. The Standish Group put it slightly differently and got the same answer: 64% of what teams ship is hardly used or never touched at all.</p><p>Translate that into a real engineering team. Five engineers, fully loaded, costs about a million dollars a year. If 64% of their output goes unused, that’s roughly $640,000 a year you’re paying to build things nobody wanted. Pendo extrapolated to public cloud companies and got to $29.5 billion a year, industry-wide, vaporized.</p><p>Whose job was it to figure out what to build? It was the PM’s. And the PM, in most companies, spent that time writing specs.</p><p>That’s the joke we’ve all been living inside.</p><h3>AI Just Took the Easy Half</h3><p>Drafting a PRD. Writing user stories. Synthesizing customer feedback. Building a competitive analysis. Mocking up a wireframe. Writing release notes.</p><p>Every one of those is now a ten-minute job.</p><p>McKinsey estimates AI could technically handle around 44% of work hours in the US economy, and PM grunt work is sitting right in the bullseye. A Microsoft Work Trend Index report this year found that 75% of workers are already using AI on the job, and 80% are bringing their own tools in even when their company hasn’t formally adopted any. Two-thirds of business leaders say they wouldn’t hire someone without AI skills now. Most of them mean PMs when they say it.</p><p>The PM who built their career on producing the most thorough spec is the PM whose career is in trouble. Not because AI does it better. Because AI does it in a tenth of the time, and the value of the work was never in the spec anyway. The value was in the decision underneath the spec. AI doesn’t make that decision. A person does.</p><p>That’s where the actual job is now. And it’s the job product management was always supposed to be.</p><h3>What’s Actually Left</h3><p>Three things. That’s it.</p><h3>Business focus</h3><p>How does this product make money? Who pays for it, why now, and what would make them pay more? A PM who can’t sketch the monetization path on a whiteboard in two minutes is not bringing their full value. They should be living inside expansion revenue, churn analysis, and willingness-to-pay research. They should know which customers are profitable and which ones are leaking margin. If they don’t, they’re a translator with a fancier title.</p><h3>Customer focus</h3><p>Five customer conversations a week. Written up. Not personas. Conversations. Real ones, with real users who have real problems and real wallets. Bain &amp; Company has been running the same number for years and it hasn’t moved: a 5% increase in retention drives a 25% to 95% increase in profit, depending on industry. The lever is enormous. The work to pull it is unsexy. Most PMs aren’t doing it. Older industry surveys found that more than 1 in 5 PMs “rarely leave the building,” and I see no evidence that’s gotten better. AI didn’t change that math. It just made it embarrassing to keep ignoring.</p><h3>User focus</h3><p>Owning the adoption dashboard. Tuning it. Finding the signal in it. Knowing why activation dropped 4% last week and what’s being done about it. That data was always there. The PM who looks at it daily is rare. The PM who can’t tell you the activation rate of their last release should be replaced by an AI agent. At least an AI agent will look at the dashboard.</p><p>That’s the job. Smaller than what most PMs have been doing. Harder. Worth dramatically more.</p><h3>The New Scorecard</h3><p>If I were running a product org tomorrow, this is what I’d ask each PM to walk me through, weekly. No more, no less.</p><ul><li>What’s the revenue or retention hypothesis you’re testing right now?</li><li>Who pays for this and why now?</li><li>What did three customers tell you this week that you didn’t already know?</li><li>What does the adoption dashboard say, and what are you doing about it?</li></ul><p>If the answer to any of those is “let me get back to you” three weeks in a row, that’s not a coaching problem. That’s a “we hired the wrong shape of person” problem.</p><h3>The Connection Nobody’s Making Yet</h3><p><a href="https://medium.com/@jeff2invest/the-best-thing-that-ever-happened-to-software-engineering-is-already-here-ef758532169e">I wrote a piece a few months ago</a> about how engineering is splitting in two. Harder technical work at the frontier on one side, and something new called the product engineer on the other. Someone who can hold the whole arc of a problem from conversation to shipped solution.</p><p>Product is doing the same thing, in the same direction, at the same time.</p><p>The product engineer is going to absorb a huge amount of what feature-factory PM looked like, because they’ll just build the thing once they understand the user. Specs as a handoff artifact start to dissolve. What does that leave for product management as a profession?</p><p>It leaves the work that always actually mattered. Knowing what to charge. Knowing who to keep. Knowing why people stay. Knowing what the next dollar of revenue looks like and what it costs to get.</p><p>If your PMs are doing that work, they’re more valuable now than they’ve ever been. If they’re not, the job is changing whether they want it to or not. The people redefining themselves around business, customer, and user are going to look up in eighteen months and realize they’ve been doing the most leveraged work in the building the whole time.</p><p>The other PMs will be writing LinkedIn posts about how AI took their jobs.</p><p>It didn’t. AI took their busywork. The job was already there underneath it.</p><p>The next piece is for engineering leaders watching their product team run a 2019 playbook on a 2026 problem. <a href="https://medium.com/@jeff2invest/pivoting-product-at-ai-speed-815955d7a2e6">Read it here.</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7c2c3ad60ea7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What If the Best Version of Your Engineering Career Starts Today?]]></title>
            <link>https://medium.com/@jeff2invest/the-best-thing-that-ever-happened-to-software-engineering-is-already-here-ef758532169e?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/ef758532169e</guid>
            <category><![CDATA[product-engineering]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-coding]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Wed, 11 Mar 2026 08:17:58 GMT</pubDate>
            <atom:updated>2026-03-11T08:23:27.276Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Qir1UmZLbtAfogc958Rptw.png" /></figure><p>Something good is happening to your profession. It doesn’t look like it from the outside. The headlines are noisy, the job market feels weird, and everyone has an opinion about whether AI is going to take your job. But underneath all of that noise, something genuinely exciting is taking shape.</p><p>Engineering is splitting into two. And both halves are better than what most of us have been doing.</p><p>Let me explain what I mean.</p><h3>The Thing Nobody Is Saying Out Loud</h3><p>For the last twenty years, most software engineering has quietly settled into a dominant pattern: someone decides what to build, writes it down in a ticket, and an engineer builds it. Sprint after sprint. Feature after feature. The engineer’s job was to translate requirements into working code, and to do it fast, cleanly, and with few questions.</p><p>That pattern worked. It scaled. It builds the products you use every day.</p><p>It also left many engineers feeling like they were only ever seeing one piece of a puzzle; they weren’t allowed to look at it as a whole.</p><p>AI is breaking that pattern, not by replacing engineers, but by dissolving the conditions that created it. When the mechanical parts of building software get faster and cheaper, the assembly line stops making sense. And when the assembly line goes away, something more interesting takes its place.</p><p>Two kinds of engineers are emerging. Not better or worse, just different. And understanding which one you are might be the most useful thing you do this year.</p><h3>The First Half: Engineering Gets More Engineering</h3><p>Here’s what the job market is actually doing, because the numbers tell a story the headlines don’t.</p><p>General software development job postings have dropped 34% since 2020. At the same time, AI and ML engineering roles are up 143% since last year alone, growing three times faster than the average job across any industry. In just the first four months of 2025, AI-related job postings more than doubled, from 66,000 to nearly 139,000 open roles. Right now, there are over 500,000 AI and ML engineering positions open globally, and companies are struggling to fill them. Not because engineers don’t exist, but because the work requires something rare: a deep technical foundation combined with the ability to build systems that have genuinely never been built before.</p><p>That’s the split the market is making visible. The routine implementation work is contracting. The frontier work is exploding. And the frontier isn’t just a Silicon Valley phenomenon anymore. In early 2025, engineering hiring in finance was up 91% and in industrial automation up 73%. Industries that relied on spreadsheets and legacy systems for decades are now urgently seeking engineers to build intelligent infrastructure from the ground up. Do not integrate a tool. Build the thing itself.</p><p>What’s significant about this demand is what kind of engineer it’s calling for. Of all ML engineering job postings analyzed in 2025, 57% specifically sought domain experts, engineers with depth in one or two technical areas, over generalists. The market isn’t asking for someone who can do a bit of everything. It’s asking for engineers who have gone genuinely deep.</p><p>This is what real engineering always was, underneath the feature factory. The assembly line obscured it for a generation. As AI absorbs the routine work, what’s left at the technical frontier is purer, harder, and more interesting than what most people spent the last decade calling “senior engineering.”</p><p>The engineers doing this work aren’t being replaced. They’re being handed more leverage than any generation of engineers has had before, with an open field of problems that didn’t exist five years ago.</p><p>If you’re the kind of engineer who gets restless when the problem is solved, and someone else’s harder problem is what pulls you forward, this is your moment.</p><h3>The Second Half: A Role That Didn’t Exist Before</h3><p>Then there’s the other half. And this is the part I find myself thinking about most.</p><p>Something new is being born right now, and it doesn’t have a clean name yet, but “product engineer” is the closest we’ve got. It’s a role that couldn’t exist before AI, because it would have been too much for one person to hold. Now it’s not.</p><p>Picture this: you start from a real conversation with a real person who has a real problem. You understand it deeply, not from a spec, but from listening. You can see what the solution should feel like before you’ve written a line of code. You build it, test it with your own hands, watch it work in someone else’s hands, and iterate until the gap between what you made and what they needed disappears.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TlsLPLpFrG-BbE4rVd32dw.png" /></figure><p>That’s the whole arc. Start to finish. No handoffs. No telephone game between the PM, the designer, the engineer, and the QA. Just a person who can hold the whole problem and see it through.</p><p>That person is the product engineer. They’re not a lesser version of an engineer or a technical version of a product manager. They’re something genuinely new. Someone who brings engineering rigor to the question of what should exist, not just how to build what someone else decided.</p><p>LinkedIn’s Chief Economic Opportunity Officer described it well: the tools we have now take what used to be a days-long relay race between design, product, and engineering and put it in the hands of a single person with the right instincts. The bottleneck was never the talent. It was the handoffs. Now those handoffs can disappear.</p><h3>Why Engineers Are the Right People for This</h3><p>Here’s something that doesn’t get said enough: engineers are uniquely suited to the product engineer role.</p><p>Not despite their technical background. Because of it.</p><p>The ability to look at a messy, ambiguous problem and find the structure inside it is engineering thinking. The instinct to ask “but what’s actually causing this?” instead of accepting the surface complaint is debugging applied to human problems. The drive to build something that holds up under real conditions, not just demo conditions, is the same discipline pointed at a different target.</p><p>What product engineers need to grow into isn’t the technical part. It’s the permission to care about the human part. To spend time with users not as a formality but as the most important research available. To have a vision for where something should go and trust it enough to build toward it. To understand the business well enough that their technical decisions serve the actual goal.</p><p>Forrester found that engineers spend only about 24% of their time writing code. The rest is already there: designs, tests, bugs, conversations, decisions. The product engineer just makes all of that intentional and points it at the user.</p><h3>Who’s Already Living This</h3><p>Look at the engineers in your organization who seem slightly restless in a good way. Who always has a take on the product that isn’t in their job description. Who comes back from a conversation with a user lit up with ideas?</p><p>They’re not ignoring their engineering responsibilities. They’re doing engineering the way it could be done when someone refuses to stay in their lane.</p><p>Engineering managers tend to recognize this feeling immediately because the skills of a great EM look almost identical to what a product engineer needs: holding the technical picture and the human picture at the same time, thinking about where the architecture needs to go while understanding what the user actually needs, making judgment calls that require all of it at once.</p><p>The product engineer isn’t a new idea. It’s those instincts, finally given a lane.</p><p>What AI unlocks is scale. One product engineer, with the right tools and the right curiosity, can do what used to require a cross-functional team. Not because they’re superhuman, but because the overhead that used to require a team has been absorbed. What’s left is the part that always required a person who actually cared.</p><h3>So Which One Are You?</h3><p>You don’t have to decide today. But it’s worth sitting with.</p><p>Are you most alive when the problem has no existing answer, when you’re operating at the edge of what’s technically possible, and the challenge is figuring out what’s even buildable? That’s the engineering path. It’s getting clearer, harder, and more important.</p><p>Or are you most alive when you can feel the gap between what someone needs and what currently exists, when you can see the solution before you’ve built it, and you just want to close that gap as fast and cleanly as possible? That’s the product engineer path. It’s emerging in real time, and the people who step into it early will define what it looks like.</p><p>Both paths are worth wanting. Both are where the industry is heading. Both are more interesting than the assembly line most of us have been running.</p><p>The feature factory was never the destination. It was what we built while we were waiting for the tools to catch up to the ambition.</p><p>The tools are here.</p><p><em>Which half of this split do you feel pulling you? I’d genuinely love to hear it in the comments.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ef758532169e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A DevOps Solution to Increased Customer and Product success]]></title>
            <link>https://medium.com/@jeff2invest/a-devops-solution-to-increased-customer-and-product-success-3684ea8e2f38?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/3684ea8e2f38</guid>
            <category><![CDATA[customer-service]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[software-bugs]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Tue, 19 Mar 2019 02:31:39 GMT</pubDate>
            <atom:updated>2019-03-22T22:59:10.074Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vLtFVPTHJGDfw3XOl4C1Sw.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/photos/bq31L0jQAjU?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">“My Life Through A Lens”</a> on <a href="https://unsplash.com/search/photos/teamwork?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>Ship code faster! Build more features! Create better experiences! Add more value for the customers! Statements a DevOps Product team are all too familiar with. To do this, we need great efficiencies, take risks and enjoy our work! There isn’t anything more destructive to this culture than bugs.</p><p>Clarification, when I say DevOps, I mean the culture that dev and ops work together. They share operational and product goals focused on customers. I don’t mean the position of a DevOps engineer who works on Kubernetes, Docker and the cloud. That’s a miss-justification of DevOps which we can debate later.</p><p>I have come to the conclusion that bugs are destructive to a business and it’s product development. I fix the same bugs. Customer service gets tired of fielding calls and the product isn’t selling. Many hours are spent handling attrition, customer perception, and low product adoption. So I theorized the following:</p><ol><li>With a lower volume of calls, customer service can have more valuable conversations.</li><li>Developers understand the problem better if they troubleshot in real time.</li><li>Ticketing and non-technical employees cause the <a href="https://en.wikipedia.org/wiki/Chinese_whispers">“telephone game effect”</a></li><li>Developers troubleshoot bugs better when they “feel” issues</li><li>Fewer bugs would increase the time to market for features</li><li>Developers being a part of the entire bug process would give them better insight to reduce bugs in future releases</li><li>Increased morale would increase product adoption</li></ol><p>Aiming to solve these problems I have come up with the following plan.</p><h4>A developer is always “on-call”</h4><p>The biggest change is to put developers on call. During their work hours, they are available for questions and a resource for support and ops. This isn’t the whole team, but a single developer at a time. Devs are able to document the root cause and create workarounds. When not needed they work on bugs that are prioritized. This process gains efficiencies leading to a reduced volume of non-bugs and increased resolution of bugs.</p><ul><li>Common questions were documented by developers</li><li>Developers created troubleshooting tools on their own to speed up troubleshooting and offload simple issues to customer service</li><li>Hard to understand performance bugs were fixed</li><li>Support was more self-reliant</li><li>Reduced Bugs introduced into production</li><li>Tickets had better technical detail reducing time to fix a bug</li></ul><h4>Ops and devs own the prioritization</h4><p>When it’s time to review and prioritize the bugs I have dev and ops do it. Because they have been on call they should have a good understanding of what the bug is.</p><ul><li>Bugs missing data are kicked back immediately reducing cycle times</li><li>Low hanging fruit can be fixed quickly appeasing more customers</li><li>Repeated bugs are recognized</li><li>Correct monitors can be put in place to address misunderstood issues</li><li>Work on top priority bug til fixed, no matter the time</li><li>Feedback needs to be taken from the Product Manager, Sales and Customer support</li></ul><h4>Bug Prioritization</h4><p>This is a guideline, and there are exceptions. Every once in a while a bug for a big customer will need to take priority. If the exception becomes the rule, you’ll need to circle yourself back to this priority list.</p><ul><li>Bugs that take the most time organizationally</li><li>Low hanging fruit</li><li>Biggest client bug</li><li>Everything else</li></ul><p>This reduces the time everyone in the org is dealing with issues. Happier employees, increased velocity and overall better product adoption are the outcomes.</p><h4>Attitude/Aptitude and Everything else:</h4><p>The attitude needed here is that everything is fixable. All too often you’ll hear it’s not possible or there is some sort of feature that fixes it in the future. Have some grit and determination to solve the problem at hand.</p><ul><li>Understand a bug before building new (<a href="https://www.joelonsoftware.com/2000/08/09/the-joel-test-12-steps-to-better-code/">5 on Joel test</a>)</li><li>Have attention to detail</li><li>Test the assumption — can you reproduce the abnormal behavior</li><li>Have grit: everything is fixable</li><li>If you don’t have enough data, go get it. Put monitors in place</li><li>Timebox bug prioritization. Better to fix the wrong bug than fixing none</li><li>Identify product issues and fix those too. (bugs and product experiences are perceived the same)</li></ul><h4><strong>Conclusion</strong></h4><p>This process increases product adoption, morale, sales, feature development, stability and more. This process isn’t easy, it rubs every department a little. It’s a disruptive process but necessary. You need to be committed and resilient when others say no. But in the end, when the tides are turning, you’ll be happy you did.</p><figure><a href="https://usejournal.com/?utm_source=medium.com&amp;utm_medium=noteworthy_blog&amp;utm_campaign=guest_post_image"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f2IVAl0TbsfES9cFGYr40g.png" /></a></figure><p>📝 Read this story later in <a href="https://usejournal.com/?utm_source=medium.com&amp;utm_medium=noteworthy_blog&amp;utm_campaign=guest_post_read_later_text">Journal</a>.</p><p>🗞 Wake up every Sunday morning to the week’s most noteworthy Tech stories, opinions, and news waiting in your inbox: <a href="https://usejournal.com/newsletter/?utm_source=medium.com&amp;utm_medium=noteworthy_blog&amp;utm_campaign=guest_post_text">Get the noteworthy newsletter &gt;</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3684ea8e2f38" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Advanced ARM Templates: Microservice Pattern for Functions]]></title>
            <link>https://medium.com/@jeff2invest/advanced-arm-templates-microservice-pattern-for-functions-d4e7ca9c9078?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/d4e7ca9c9078</guid>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[azure-resource-manager]]></category>
            <category><![CDATA[microservices]]></category>
            <category><![CDATA[azure-functions]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Fri, 08 Mar 2019 17:30:17 GMT</pubDate>
            <atom:updated>2019-03-08T17:30:17.419Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AvcN13NdywJR4bl_sRihYw.png" /></figure><p>There is no doubt that Azure is catching up in cloud wars. With more developers flocking to Azure Services, ARM is getting is getting more traction.<br>However, ARM with all its goodies does fail short in certain cases. I attempt to shed some light on the issues, hiccups, and nuances in deploying services and how we overcame them. I want to thank my buddy <a href="https://github.com/thatinfrastructureguy">Ashish</a> for his help in getting this up and shared!</p><p>What I’ll cover:</p><ul><li>General file structure for readability and clean modularization of code using linked templates</li><li>Create clean function/web app settings input files for toggles</li><li>local and CI/CD deployments without a storage account</li><li>Things you didn’t think you can do (Comments!!)</li></ul><p>Who’s this for:</p><ul><li>Someone using ARM IaC solution for all Azure services (declarative code practices)</li></ul><h3>Building your code structure</h3><p>Create your folder structure with 3 levels of abstraction: Deployment, service, &amp; resource. This allows for easy troubleshooting, testing and configuring at each layer.</p><pre>├── deployments <br>│ ├── region<br>│ │ └── microservice1      (service in the region your deploying)<br>│ │   ├── deploy.sh<br>│ │   └── parameters.json<br>└── public <br>  ├── resources            (storage account, cosmosdb, webapp etc.)<br>  │ └── global             (resources with location ‘global’)<br>  │   └── policy.json<br>  └── services             (your microservices.)<br>    └── microservice.json</pre><p>For in-depth technical details and deployable examples reference the my <a href="https://github.com/JeffreyGarrett/ARM-Microservice-Design-pattern">GitHub repo</a>.</p><h3>Application Settings: Function/Web Apps</h3><p>Templates for functions need connection strings, keys and app settings. In the previous section, I stated you need to use outputs and parameters to do this. I’ll explain this more.</p><h4>Connection strings:</h4><p>Always output connection strings from resource template as in this CosmosDB template output. Then pass as a variable into your subscription level template</p><pre>&quot;outputs&quot;: {<br>  &quot;cosmosConnection&quot;: {<br>      &quot;type&quot;: &quot;string&quot;,<br>      &quot;value&quot;: &quot;[concat(&#39;AccountEndpoint=https://&#39;,toLower(parameters(&#39;rgName&#39;)), &#39;.documents.azure.com:443/;AccountKey=&#39;,listKeys(variables(&#39;resourceId&#39;), variables(&#39;apiVersion&#39;)).primaryMasterKey, &#39;;&#39;)]&quot;<br>}</pre><h4>App Settings Config:</h4><p>Developers have app settings that need configuration in their app. DevOps Engineers will have connection strings, keys, and other secrets and references. Also, web apps settings are overwritten with every deployment. So we’ll need a way to incorporate all these together.</p><p>App Settings: incorporate your developer defined app settings in the parameters.json file as an object.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/860/1*Wn7rC0iJwUNKwM5r31c4iQ.png" /></figure><p>in here we have two app settings and an API key. For more details on the API key see here: <a href="https://azure.microsoft.com/en-us/updates/azure-functions-key-vault-integration/">https://azure.microsoft.com/en-us/updates/azure-functions-key-vault-integration/</a></p><p>Note that we have an app settings object with a customSettings array. This is important for combining with variables you generate in the template.</p><h4><strong>Combining your settings</strong></h4><p>Below we have the microservice deployment. Notice the appSettings object we pass in from previous section.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/971/1*487xap2z2KcnfJvs899hBg.png" /></figure><p>Pass the setting array into the function app deployment as an extra parameter. Pass in the storageAccount and cosmos connection strings as well. By doing this we can keep our resources loosely coupled.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lmTcIgkMKM2pkPL3H-52tQ.png" /></figure><p>In the functionApp resource, we start to build our app settings. Combine function app variables and the connection strings. Create an array of name/value pairs with the variables and parameters.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SMI8JQLe4pZISdcRgDsjcA.png" /></figure><p>For the final app settings, union the developer-defined app settings and the DevOps defined settings into a single array.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p_0WpRL2I2GXpWjKvQD03g.png" /></figure><p>And now you should have the developer, DevOps and system defined app settings combined!</p><h3>Linked templates deployed from a local machine</h3><p>Microsoft suggests deploying linked templates via a storage account with SAS token. Instead, we can use local web servers to deploy our templates from the file system. This is helpful to speed up the iteration and testing processes.</p><p>Linked templates need the templates to be public. To easily solve for, locally broadcast the templates using http-server and ngrok. This allows for quick and iterative deployments by omitting the storage account process.</p><ul><li>install http-server npm install -g http-server</li><li>install ngrok npm install -g ngrok</li><li>from public folder deploy http server http-server ./ -p 8080</li><li>deploy ngrok ngrok http 8080</li><li>update your deployment script (in deployments folder)</li><li>run your deployment.sh script (see examples)</li></ul><p>By using HTTP<a href="https://www.npmjs.com/package/http-server">-Server</a> and<a href="https://ngrok.com/"> Ngrok</a> you can deploy linked templates from your machine or your ci/cd agent.</p><h4>BONUS Content!</h4><p>Lastly, know you can do comments in ARM templates. There are plenty of articles that say you can’t. It’s been confirmed by azure that you can.</p><h3>Conclusion</h3><p>Even with all the issues of ARM templates, we can see there is an idiomatic solution. Templating, structure and good coding practices reduce these concerns. Following these guidelines, we can quickly build stable, repeatable and testable infrastructure.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d4e7ca9c9078" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Devops, Serverless, and Azure Durable Functions — Part 2 Costs]]></title>
            <link>https://medium.com/@jeff2invest/devops-serverless-and-azure-durable-functions-part-2-costs-d71e1fc39df?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/d71e1fc39df</guid>
            <category><![CDATA[azure-functions-cost]]></category>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[azure-functions]]></category>
            <category><![CDATA[durable-functions]]></category>
            <category><![CDATA[serverless]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Wed, 24 Oct 2018 02:28:50 GMT</pubDate>
            <atom:updated>2018-10-24T02:28:50.466Z</atom:updated>
            <content:encoded><![CDATA[<p>In my <a href="https://medium.com/@jeff2invest/devops-serverless-and-azure-durable-functions-part-1-azure-wins-2a725bcd03d8">first post</a>, I explained why to choose Azure and Azure Durable Functions. In this article, I will provide real-life cost and POC with supporting data. Remember costs change over time so to confirm the costs I’ll share the tools and method to run your own tests.</p><h3>Create the Azure Function App</h3><blockquote>I set up a function app on Azure from the Azure portal. I used c# .net core and windows boxes because Linux is still in preview. No need to fight bugs from preview products.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4TLk-0b954EGnZvSbjVgOQ.png" /><figcaption>Search for Function App in the marketplace</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/522/1*IZgHkDKmMl-WarbbpyK4fg.png" /><figcaption>Setup your function app</figcaption></figure><h3>Create a durable function</h3><p>My Colleague Kevin Jones, created a quick durable function that is ready to deploy in Azure.</p><p><a href="https://github.com/KevinDJones/DFDevOpsTest">https://github.com/KevinDJones/DFDevOpsTest</a></p><p>It’s an HTTP event orchestrator that accepts a JSON object. The request payload is editable to change the length of execution time or delay the job processing. If you’re more interested in Azure functions in general, then see Kevin’s <a href="https://hackernoon.com/durable-functions-backend-development-made-easy-and-cheap-cbab7acc7050">article</a>.</p><h3>Deploy your function</h3><p>I hooked up a test CI/CD pipeline to make deployments and changes easier. You can see how it’s done in the next couple of articles. If you’re doing the cost analysis, connect Github and auto deploys your master branch. That works fine for POC’s and testing</p><p>Once deployed you should have these functions and we can run our load test.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/241/1*wukV5yXTlYO_brfYS8nLIg.png" /></figure><h3>Running the test</h3><p>The test needed to mimic the following use cases and requirements:</p><ul><li>Burstability — Automatically scale up and down</li><li>Error rate — increased load doesn’t increase errors</li><li>Response Time — is constant and predictable</li><li>Cost — was cheaper than running your own Kubernetes/docker cluster</li></ul><p>To set up the test I used <a href="https://www.blazemeter.com/">Blazemeter</a>. There are many tools, this one happened to be an evaluation stage and easy to implement. The test configuration included 1 engine, 1000 threads for a 5 min period. I ran the test a few times ending in ~450k requests.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4o3bKkuXhsVmvb0Ied0_0w.png" /><figcaption>Load Test configuration</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LRMrh14aP65KTSsT3zJRvQ.png" /><figcaption>Performance results</figcaption></figure><p>You can see from the load test that we had a solid performance with a steady increase of response time. It’s worth noting that response time is slower than typical web apps. If you need faster response times you may need to do some further tuning or run separate tests.</p><p>Results were:</p><ul><li>Function Executions = 500k</li><li>Function Exec time = 17894.1 gbs</li><li>Total costs $.085 (remove azure devops from image below)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/426/1*UiJF9Q4IfBlyJBGot5J4cQ.png" /><figcaption>Cost Results</figcaption></figure><p>I was not charged for idle time and only for storage resources. Since Azure gives you a free tier we never hit costs for the function apps. Go to <a href="https://azure.microsoft.com/en-us/pricing/details/functions/">azure pricing page</a> and compare the executions against the sheet to determine what your load might look like.</p><h3>Serverless is cost effective!!</h3><p>This model is great for burstable loads or event based process’s. The cost is much cheaper than a Kubernetes, Docker or VM implementation</p><p>Docker/Kubernetes is still built on the foundation of pre-allocated resources. In comparison, a stack similar to this you need at least 6 vms.</p><ul><li>Two for application high availability</li><li>Two for data persistence and high availability</li><li>Two to replicate across regions</li></ul><p>If these vm’s were even $50 a piece you would be spending 300 for the vm’s plus storage. With Azure Durable Functions we paid $.085 and really just for the storage.</p><h3><strong>How does Azure make money if it’s so cheap!</strong></h3><p>Make development easy, cheap and business effective and charge for the plumbing. Database’s, messaging bus, region availability are all areas Azure provides great value. It’s plumbing you’re business doesn’t want you to focus towards (unless you work for a DaaS company or azure competitor). Azure is utilizing it’s idle resources to give you a good way to get business going. In return, you’ll use the savings to buy infinitely scalable database’s like cosmosdb. And I suggest you do.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d71e1fc39df" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[DevOps, Serverless, and Azure Durable Functions — Part 1 Azure wins!]]></title>
            <link>https://medium.com/@jeff2invest/devops-serverless-and-azure-durable-functions-part-1-azure-wins-2a725bcd03d8?source=rss-2f130afc9b93------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a725bcd03d8</guid>
            <category><![CDATA[azure]]></category>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[azure-durable-functions]]></category>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[serverless]]></category>
            <dc:creator><![CDATA[Jeff Garrett]]></dc:creator>
            <pubDate>Fri, 19 Oct 2018 00:26:10 GMT</pubDate>
            <atom:updated>2018-10-19T00:26:10.617Z</atom:updated>
            <content:encoded><![CDATA[<h3>DevOps, Serverless, and Azure Durable Functions — Part 1 Azure wins!</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RgYAxwRNTxdamppQ" /></figure><p>If your diving into Serverless you need to pick a cloud provider you align with. And the best for the future is Azure. The capabilities and future of Azure look bright and exciting. But why not AWS and GCP they are more developer friendly? Good question and I got an answer… because they don’t. Azure and serverless are going to fit together well and here’s why.</p><h3>Serverless and what it means</h3><p>Since code and infrastructure are tightly coupled in serverless, vendor lock-in is inevitable. Third-party tools love to tell you they are agnostic, but your code is still tied to its underlying OS. In the case of serverless, it’s your cloud provider.</p><p>For any successful project, your cloud provider needs to align their long-term strategy with your business strategy. Microsoft aligns with your business more than you think.</p><ul><li>AWS — Main focus: Grow Amazon.com</li><li>GCP — Main Focus: create marketing and ads with free technology</li><li>Azure — Main Focus: Give tools and API’s to developers to grow their business</li></ul><h3>The business of Microsoft is Developers</h3><p>Microsoft was the original software developer bff. They pioneered the concept of developer communities and tooling to bolster their platform. If you aren’t familiar with the history of Microsoft, read this article.</p><p><a href="https://arstechnica.com/information-technology/2017/05/microsofts-renewed-embrace-of-developers-developers-developers-developers/">Microsoft&#39;s renewed embrace of developers, developers, developers, developers</a></p><h3><strong>Microsoft’s strength is API’s</strong></h3><p>Joel Spolsky, in his <a href="https://www.joelonsoftware.com/2004/06/13/how-microsoft-lost-the-api-war/">article </a>“How Microsoft Lost the API War, ” stats Microsoft became successful from opening up its windows API’s. What led to their downfall is the internet boom and the irrelevance of those API’s.</p><blockquote>Microsoft’s crown strategic jewel, the Windows API, is lost. The cornerstone of Microsoft’s monopoly power and incredibly profitable Windows and Office franchises, which account for virtually all of Microsoft’s income and covers up a huge array of unprofitable or marginally profitable product lines, the Windows API is no longer of much interest to developers.</blockquote><p>As the internet grew, OS API’s took the form of REST API’s and consumable services. Open source helped others build faster. Companies like Twilio branded themselves as communication platforms for developers. In order for Microsoft to get back into the game, they must reinvent what their API is… along comes Azure.</p><h3>What’s Great About Azure</h3><p>Since features change so often, I’m not going into a feature list comparison. What is something to take note is that Azure/Microsoft have been doing software for a long time, and it shows! If the following things matter to you, you should definitely move to Azure.</p><ul><li>Compliance: <a href="https://azure.microsoft.com/en-us/overview/trusted-cloud/">More compliance than anyone</a></li><li>Data: SQL Server is consistently a leader. <a href="https://cloudblogs.microsoft.com/sqlserver/2017/11/03/three-years-in-a-row-microsoft-is-a-leader-in-the-odbms-magic-quadrant/">gartner quadrant</a></li><li>Data: <a href="https://azure.microsoft.com/en-us/services/cosmos-db/">Cosmosdb</a> first globally distributed database service with guaranteed latency</li><li>Data: <a href="https://powerbi.microsoft.com/en-us/blog/gartner-recognizes-microsoft-as-a-leader-in-analytics-and-bi-platforms-for-11-consecutive-years/">Gartner recognizes PowerBI leader in analytics and BI</a></li><li><a href="https://www.cbtnuggets.com/blog/2018/02/how-ai-is-taking-over-azure/">AI is everywhere with Azure</a></li><li>Visual Studio still best IDE according to <a href="https://www.g2crowd.com/categories/integrated-development-environment-ide">G2crowd</a></li><li>Azure Durable Functions: <a href="https://docs.microsoft.com/en-us/azure/azure-functions/durable-functions-overview">First stateful serverless </a>function</li></ul><h3>Buy in</h3><p>Embrace the technology that is coming. Azure is going to be a powerhouse for any developer in this century. For all the reason I already stated, partnering more with Azure will benefit greatly in the long run.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2a725bcd03d8" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>