We Mentioned Replit in 214 Articles Last Year. For Free. Most Vendors Have No Plan For Customers Like That

extract

Here is what SaaStr AI did for Replit over the past 12 months, without being asked, without being an investor at SaaStr Fund (oh well): - Mentioned them in 214 articles - Mentioned them in 40+ podcasts reaching hundreds of thousands - Generated 5,901,900+ page views and impressions for them, across an audience of roughly 450,000 of the top B2B executives in the world Replit paid $0 for all of it. They did later sponsor SaaStr AI 2026 (thank you), and they built our packed vibe coding cafe, which was one of the best activations at the event. They’ve also given us S-tier forward-deployed engineering support. All of that came after, and none of it was the reason for the 214 articles. The 214 articles happened because Amelia and I use Replit every day to build real production software with no engineering background, and writing about what I’m actually doing is the job for us. I’ve shipped 10+ production apps on it. They’ve been used close to a million times. SaaStr.ai hit 500,000 users in its first 45 days. I cannot write a line of Python from scratch. There was no deal, no brief, no approved messaging, no “partner content” tag. That’s the part worth thinking about if you run a B2B company. Somewhere in your customer base there is probably someone doing a smaller version of this for you right now. And there is a very good chance nobody at your company knows their name. The six things that matter here: - Advocacy inside your exact ICP is worth multiples of the same reach outside it, and it’s the one channel you cannot buy at any price. - Count surfaces, not mentions. One podcast segment becomes a YouTube video, X clips, LinkedIn articles, a newsletter placement, and several articles. And a real advocate also carries you onto platforms they don’t own, which you could never buy your way onto. - It gets created by the product working on a hard job, not by your partner marketing team. - Most companies never detect it, because nothing in the CRM fires when someone praises you unprompted. - When they do detect it, the standard corporate response is to route the person to a rep and try to turn it into a contract. That kills it. - The correct response is to make that person absurdly successful before you ask them for anything, starting with one named engineer who never gets reassigned. #1. Run the media math, then apply a credibility multiplier 5,901,900 impressions. Pick any reasonable B2B CPM you like. At $50 per thousand, that’s roughly $295,000 of media. At $100, it’s $590,000. Those are rough numbers and the CPM assumption is doing all the work, so don’t over-index on the precision. But the reality is given the audience, it was worth much more than $590,000. The more important point is that the arithmetic understates it badly, because CPM prices reach and this wasn’t reach. It was audience. Those 5.9 million impressions landed on an audience of roughly 450,000 of the top B2B executives in the world. Founders, CEOs, CROs, CTOs, VPs of product and engineering. For a company like Replit, that isn’t an audience that contains some buyers. Every one of those people is a builder, which means every one of them is a customer or controls a budget that buys for builders. The ICP overlap is close to total. Compare that to what the same $295,000 buys on a general technology channel, where you’re paying for millions of impressions and maybe two percent of them hit someone who could ever purchase your product. Same CPM, fiftieth of the value. The number that matters is cost per qualified impression, and almost nobody computes it because it requires you to be honest about how little of your paid reach is actually your buyer. Then apply the credibility multiplier on top of it. Sponsored content is read as sponsored content. A vendor logo on a slide is read as a vendor logo on a slide. What happened here was a practitioner describing, in detail, over and over, what he built that week and what broke. That includes the parts that didn’t work. I’ve written publicly about an AI agent deleting 2,400+ production records and reporting no changes detected, and about fabricated test results showing an 88% pass rate when the real number was 48%. That’s not flattering coverage. It’s why the flattering parts land. You cannot purchase this. There is no line item. There is no agency that sells it. The only known method of production is: build something that works well enough on a hard enough problem that a credible person in your ICP can’t stop talking about it. #2. Those numbers only count our own surfaces The 214 articles number is easy to picture. The 40+ podcasts number is the one people misread, because they hear it as 40 mentions. The Agents is the weekly show I co-host with Amelia Lerutte, our Chief AI Officer, about running our agent stack in production. Episode 1 did about 40,000 views in 72 hours, roughly 7x our channel average. The audience is people currently deciding what to deploy, which is the only audience an infrastructure company actually wants. Here’s the part that matters for the math. A single mention on that show does not stay on that show. It fans out: - The YouTube episode - The audio episode - Clips cut for X - A LinkedIn article on my feed and Amelia’s and the SaaStr page - The newsletter - Show notes - And two to five SaaStr blog posts derived from the episode, each of which is one of the 214 One clip about localizing an app into Chinese and Spanish in 20 minutes using Replit did about 24,000 impressions on X by itself. That’s one clip, from one segment, from one episode. And none of these were ad reads. They were segments about what we shipped that week and what broke, which is why people watch them. An ad read gets skipped. A story about a thing that actually worked at 11pm on a Tuesday does not. Then there’s everything on somebody else’s platform Every count above is a SaaStr surface. None of it includes the weekly 20VC x SaaStr episodes I do with Harry Stebbings and Rory O’Driscoll, where Replit comes up constantly. It’s been the headline topic more than once. The $9B round was a full segment. When we picked founder of the year, the argument was that without Claude there is no vibe coding, and Replit was the example. 20VC is not our show. Different host, different audience, mostly investors rather than operators. We can’t sell a placement on it and neither can Replit. That’s the property nobody prices correctly. An advocate doesn’t just talk about you on their own channels. They carry you onto other people’s channels, into rooms where you have no relationship and no way in. Then each of those episodes turns into a SaaStr recap post and a Sunday follow-up, so it loops back into the 214 anyway. If you’re trying to value an advocate, don’t count their mentions. Count the surfaces each mention lands on, including the ones they don’t own, and how long each one keeps working. #3. Nothing in your stack detects this Here’s the operational failure, and it’s near-universal. Your CRM fires on demo requests, pricing page visits, trial signups, and support tickets. It does not fire when someone with 100,000 followers in your ICP mentions you favorably for the fortieth time. There’s no object for it. No owner. No alert. So the typical sequence at a B2B company is: - A customer becomes a genuine public advocate. - Nobody notices for months. - Eventually a marketer notices and pings the AE. - The AE looks at the account, sees a small contract, and moves on. - The advocate’s support ticket goes into the normal queue behind 400 others. Every step of that is defensible in isolation. The aggregate outcome is that your single highest-leverage marketing asset gets the same treatment as a $99/month self-serve account with a billing question. If you want to fix one thing this quarter, make it detection. Somebody, or some agent, needs to be watching for unprompted mentions from people who matter in your category, and there needs to be a named human who owns what happens next. We run 20+ AI agents in production at SaaStr with a team of three humans, and this is exactly the kind of always-on, low-glamour monitoring work agents are good at. There’s no excuse anymore for not knowing who is talking about you. #4. The three ways companies kill it Once a company does notice, the usual responses do real damage. - Turning it into a deal too early. Every sales rep does this, and kills the relationship. The “we should formalize this into a partnership” email. The advocate was talking about you because they wanted to. The moment there’s an agreement, their audience discounts everything they say, and so do they. You converted an asset with a 10x credibility multiplier into a media buy at 1x. - Routing them to normal support. This one is the most common and the most expensive. The advocate hits a real bug, files a ticket, and gets a template response and a 48-hour SLA. What Replit did instead was give us forward-deployed engineering support, which means when something is broken, an actual engineer who knows the system looks at it. That is not a favor. It’s the cheapest insurance policy in the company. - Changing the deal underneath them. I watched Adobe kill EchoSign’s free edition within weeks of my departure. The next owner of a product always cares less about the specific promises the last one made. Advocates are exposed to this in a way regular customers are not, because they told their audience you were good. When you break it, they have to explain the reversal in public, and the honest explanation is a much better piece of content than the original praise was. We were one of Marketo’s first ten customers. We stayed roughly 20 years. We left. Twenty years of goodwill does not make anyone permanent. 5. What Replit actually got right, in order The sequence is the lesson. First, the product did the job. I’m not an engineer. I build production applications. That gap is the entire story, and no amount of marketing spend creates it. Second, the CEO assigned their best forward-deployed engineer, early. Amjad gave us Kody near the start of all this. Not a rotating pool, not a shared Slack channel, not a CSM who escalates. One named engineer who is genuinely good, and who still works with us to this day. Here’s us together: Two things about that are worth separating out. It came from the CEO. Nobody below the founder at a fast-growing company voluntarily takes their best forward-deployed engineer off revenue-generating enterprise work and puts them on a small account. That decision only gets made at the top, and it only gets made by someone who is measuring something other than contract value. Second, and I see this mistake way too often, it never got taken away. Continuity is the part almost everyone gets wrong. Companies will assign a great person to an important customer for the first 90 days and then quietly reassign them once the relationship looks stable. What you lose when you do that isn’t the headcount. It’s every piece of context that person accumulated about how the customer actually builds, which bugs they’ve already hit, and what they’re trying to do next. Kody knows our stack. That knowledge took a year to build and would take a year to rebuild. Third, they gave real engineering support to a customer who was small on paper. By seat count and contract value, SaaStr is not a large Replit account. By what SaaStr does for Replit inside Replit’s exact ICP, it isn’t close. Somebody there was measuring the second thing. Fourth, they showed up where the audience already was. They sponsored SaaStr AI 2026 and built the vibe coding cafe rather than buying a banner. Replit engineers sat in the room all three days. They ran 10+ hands-on classes organized by role, so a CS leader and a RevOps leader each got a session about their own job rather than a generic demo. Amjad did a live build on the main stage. Founders got to sit down and actually make something with a Replit engineer next to them. That’s a demo with a floor plan, and it converts at a rate no booth does. And they got the most leads of ANY sponsor. Fifth, none of that was conditioned on the coverage continuing. The support didn’t arrive with an expectation attached. If it had, I’d have written about that instead. #6. What to do on Monday Concretely: - Find them. Pull every unprompted public mention of your company in the last 90 days. Filter to people who are actually in or adjacent to your ICP. The list will be shorter than you expect and more valuable than you expect. - Rank by audience-in-ICP, not by ARR. A 3,000-follower account where all 3,000 are your buyer beats a 300,000-follower account where 500 are. - Assign a named owner to the top 10. Not a segment. A person. - Give them one named FDE, and don’t take that person back. Not a queue, not a channel, not a CSM. Your best forward-deployed engineer, assigned by the founder, permanently. The reassignment 90 days in is the mistake, because the value is the accumulated context, not the headcount. Do this before asking for anything, and don’t mention the coverage when you do it. - Fix their problems in public. The advocate who writes about a bug and then writes about you fixing it in six hours is worth more than the one who only ever writes praise. - Only later, if they want it, make it commercial. And structure it so their independence is obvious. Sponsoring the event is clean. Paying for the article is not. The asset you already have and aren’t counting Most B2B companies spend an enormous amount of energy trying to manufacture credibility they don’t have, and almost no energy protecting the credibility somebody already handed them for free. 214 articles. 40+ podcasts. 5.9 million impressions against 450,000 of the best B2B executives on earth, every one of them a builder. Plus every 20VC episode where they came up, which isn’t in any of those numbers. Zero dollars. If someone with standing in your ICP is doing any version of that for you, treat it as one of the most important accounts you have, regardless of what the contract says. Because the one thing more powerful than an advocate with an audience is that same person explaining why they left.