ScreenshotOne

This one is different. Most case studies here reverse-engineer a success. This one is a founder opening his dashboard on camera and letting a two-time unicorn founder tell him what's wrong with it. Dmytro Krasun posted publicly that ScreenshotOne's growth had stalled; Jason Cohen — WP Engine founder and author of A Smart Bear — offered to walk through every number live. It's the most useful hour of plateau diagnosis on the internet, and it doesn't end with a growth hack.

What ScreenshotOne Does

ScreenshotOne is a screenshot API: send it a URL, get back a rendered image. Full-page captures, ad and cookie-banner blocking, multiple formats and video capture — all behind one endpoint, with SDKs, a CLI and an MCP server. Dmytro launched it solo in May 2022 after leaving a server-side engineering job and has run it bootstrapped ever since. Today it does $32K MRR across 1,000+ paying customers, with 100M+ API requests served and no employees. It's a genuinely excellent small business — which is exactly what makes the problem interesting.

The Problem They Solve

Rendering a web page reliably is deceptively brutal. Headless browsers leak memory, fonts fail to load, lazy images never fire, cookie banners cover the content, and scaling to millions of captures is a full engineering project on its own. ScreenshotOne's pitch: never build that yourself — it's infrastructure, not your product.

The trouble is who needs it. Some customers embed screenshots permanently: hosting providers capturing hundreds of sites before and after deploys to catch visual regressions, sales tools dropping a prospect's homepage into a cold email, SaaS onboarding flows pulling brand colours from a customer's site. Those people stay for years. But a large share of signups need screenshots for one project, for one month, and then leave. Both groups are averaged into the same metrics — and that average is what quietly caps the company.

The Growth Story

The curve looks healthy. Launch in May 2022. $10K MRR by September 2024 — the goal Dmytro had chased for years, which on arrival felt strangely empty. $20K MRR by June 2025. Then $25K, then $32K. Five of the last six months went up.

So what's the complaint? Cohen's first move is to correct the framing. Dmytro isn't upset that revenue is falling — it isn't. He's upset that the growth of growth is flat: roughly $1K of net new MRR every month, a straight line rather than a curve. Cohen's response is worth internalising: straight lines are normal. Almost every bootstrapped SaaS graph is a line, occasionally bending into a new, steeper line. Constant acceleration is what Facebook did, not what you should expect. The real problem isn't that the line is straight — it's where it stops.

And it does stop, because of one number: 9% monthly revenue churn. Cohen's Max MRR formula makes it concrete — Max MRR = new MRR ÷ cancellation rate. Marketing puts in a roughly fixed $3K of new MRR each month; churn takes 9% of an ever-larger base. $3,000 ÷ 0.09 ≈ $33K, almost exactly where the business now sits. At that point the dollars walking out equal the dollars walking in, and growth is mathematically over — regardless of effort or ambition.

So the conversation moves to the churn. Why 9%? Because most cancellations say the same thing: the project ended, the need was temporary. Which leads to the question that turns the whole session — is there a use case a large number of companies need permanently?

Cohen tests each candidate with arithmetic instead of enthusiasm. Hosting providers screenshotting sites around deploys — maybe 10,000 such firms worldwide, and convincing even 1% to retrofit their pipeline gets you perhaps $10K MRR. Cold email with the prospect's homepage embedded — Dmytro's own estimate of the entire market is about a thousand companies, smaller than his current customer base. Onboarding flows that extract brand colours — delightful, and vanishingly rare.

That's the crux, and it's harder than a marketing problem: in four years and thousands of customers, no repeatable, high-retention market has materialised. Not none — just none big enough to build a tripling on. People want screenshots occasionally, for scattered reasons. As Cohen puts it, if it were obvious and easy, Dmytro would already be doing it.