I ran my own 21-day sprint and missed one of four goals

I build a product that's supposed to get developers to finish their side projects. In July I went through it myself, as a participant, for 21 days.

I missed one of four goal criteria.

The obvious move is not to write about that. An accountability product whose founder falls short of their own goal is not a good sales story. But the miss is the part that demonstrates something — and the success report would have been the part that demonstrates nothing.

What has to be said first

This was a self-test, not a customer outcome. I was operator and participant at the same time. Nobody paid for it. There is no testimonial, and as of today no paying customer has completed a full sprint.

Reading this as proof that the product works overstates it. What it does establish is narrower: that the mechanism runs end to end under real use rather than merely passing in staging — and that running it surfaces things no test had found.

The goal was fixed before day one

The sprint goal gets written during the intro call and stored at acceptance. After that it can only be corrected downward, never upward. Every milestone review measures against that stored wording — not against a later summary of it, and not against whatever has since become convenient.

My goal had four checkable criteria. On day 21 I assessed each one separately against the stored text:

1. Public landing page with intake in both directions, two languages — met.
2. At least 15 candidates contacted with an internal scoring sheet — met, 16.
3. At least 3 completed reciprocal feedback exchanges, each within 48 hours — missed: 2 of 3.
4. Willingness-to-pay question asked for real — met, with a disclosed limitation.

Why the third exchange never happened

Two completed fully. In both cases I went first — delivered before receiving — and in both cases the return came back without me chasing it. Once the same day, once three days later.

The third contact didn't respond to the initial approach, and didn't respond to a follow-up either. That closed it under my own rule: one follow-up per contact, then never again. A second nudge might have saved the criterion and would have broken the rule.

I kept the rule and missed the criterion. That's a description, not an excuse: the miss follows from a rule deliberately honoured, and it's still a miss. That's exactly how it appears in the proof I submitted — not as "substantially met."

Which is the only reason this piece exists:

A frozen goal is worth precisely what it costs to miss it.

Had I rounded two of three up to a checkmark, I'd have proved that the enforcement is decoration.

The fourth criterion was the more uncomfortable one

Criterion 4 required the willingness-to-pay question to be asked for real — explicitly accepting "a paying yes or a documented honest no."

What came out was a no by null signal: the question was never put to a person, because nobody appeared to whom my own rules allowed it to be put. The option sat passively in the intake form from day 16. Response: zero.

By the wording of the goal, that's a pass. Honestly assessed, it's weaker evidence than an offer made and declined. Willingness to pay is therefore neither established nor refuted — it wasn't tested. That's what the proof says, instead of recording a clean checkmark.

What using it found that testing did not

The unexpected return wasn't in the sprint result but in the product itself. As a participant inside your own system you see things an operator never sees:

• Three faults in the system that generates and sends the daily prompts. They surfaced at the end of a sprint — a place nobody had simply ever reached before.
• A data-protection defect that had been live for about three and a half months. Found on the first real form run by an outsider, not by an audit.
• A qualification requirement only the wrong people got to see. It appeared at exactly one point in the funnel — visible only to those who had already disqualified themselves. Anyone who needed to read it never passed it.

That last one is the one I keep coming back to. A filter that only reaches people who have already identified themselves isn't filtering. While you're building it, though, it feels exactly like one that does.

None of the three would have been caught by a test. All three came from use, and from someone else's eyes.

What I take from it

The most useful finding has nothing to do with the goal: both exchanges that happened came inbound, from a single public post, and both through going first. Cold outreach to more than a dozen candidates stayed essentially silent.

Going first converts existing interest. It doesn't create any. That's a less comfortable result than "more outreach would have helped," and it's what the data says.

Where this actually stands

One completed sprint, run by the founder. No paying customer who has completed one. No testimonial. Willingness to pay untested.

That's the honest position. The full write-up, including the day-by-day log with the finding of each day, is here — including the days where the finding was that waiting looks like work.