Launching With a Measurement Window: Decide What You'll Decide Before the Numbers Arrive

The most dangerous week in a product's life is the first one after launch. Not because things break, although they do, but because every number is new and every number looks like a signal. Three people drop out at the same step and it feels like a pattern. One good day feels like traction. If you're the founder, you're also the person most desperate to read something into it.

I spent nine years watching this happen to experiments. A test goes live, the dashboard updates, and by lunchtime someone has an opinion. The defence is the same one I wrote about in pre-registering an A/B test: decide how you'll read the result before you have one.

So when I launched Keniri on 27 October, I launched it with a measurement window.

What Keniri is, briefly

Keniri is a conversation with an AI that helps you understand yourself and your relationships. You call it, like a phone call, talk through real moments in your life for fifteen or twenty minutes, and get a plain-language report on how you think and what you need. It's free up to a Report Preview, and the full report costs £15.

The window

The window closes after four weeks or 300 Report Previews, whichever comes first. Until then there is one offer on the table, the £15 full report, and the biggest open product question stays parked.

That question is whether to build the two-person version: inviting a partner or friend to have their own conversation and reading how the two of you fit together. It's the part of Keniri I'm most excited about, which is exactly why it shouldn't be decided on launch-week enthusiasm. If people don't finish a conversation on their own, a second conversation is a solution to the wrong problem.

Three numbers, and nothing inside the call

The funnel has three steps, and each one is counted as an anonymous event on the server:

1. A conversation starts. Someone has their link and begins talking.
2. A Report Preview is reached. They finished enough of the conversation to get something back.
3. A full report is bought.

That's all. The conversation page itself has no analytics on it at all. People are talking about their lives in there, and I don't need to know where their cursor went to make this decision. Three counts are enough to answer the only questions that matter at this stage: do people finish, and do the ones who finish want the rest?

Why write it down first

Writing the window down before launch does two things.

It stops me moving the goalposts. Without a stopping rule, a weak first week becomes "it's early", and a strong first week becomes "let's build the next thing now". Both are the same mistake: deciding first and finding the numbers afterwards.

It also makes the decision smaller. When the window closes I'm not asking "is Keniri working?", which no four-week dataset can answer. I'm asking whether enough people finish a conversation, and buy the full report, to justify building the two-person reading next, or whether the single-person experience needs work first.

I fully expect the honest answer at the end of November to include the word "inconclusive" somewhere. That's fine. An inconclusive result you planned for is still more useful than a confident one you went looking for.

If you want to see what the numbers are measuring, the quickest way is to have a conversation.