The code is the thing every analytics tool is blind to

UserDrift is not built yet, so this is a thesis rather than a postmortem. But the thesis is the entire company, so it is worth stating plainly and betting on in the open.

The people shipping apps now often did not write the code. They built on Lovable or Bolt or Replit, it works, and then they are staring at silence: users arrive, users leave, and they have no idea where. The tools they would reach for to find out are all blind in the same specific way.

PostHog, Hotjar, FullStory, even Celonis: they only have the event log. So the most any of them can tell you is that button#a3f2 was clicked 1,240 times. That is not an answer. It is a puzzle with the labels filed off, and a non-technical founder cannot solve it.

The thing every one of them structurally cannot see is the code, and the code is where the meaning lives. Routes are screens. Components are the buttons. The repository is the map of what the product is supposed to do. And on these platforms the code is right there, because Lovable syncs to GitHub.

So the plan is to read it. Pull the repo, turn routes into screen names and elements into human labels, and reconcile three sources of truth nobody else holds together at once: what the product was designed to do (the code), what it is set up to measure (the instrumentation), and what people actually did (the events). Where those three disagree is exactly where the dead features, the untracked buttons, and the silent drop-offs are hiding. It is process mining, pointed at a product instead of a factory floor.

Reading the code pays a second time. The suggested fix can be rendered from the user's own components and styles, so it looks like their real app, not a generic mockup. The diagnosis and the repair both come from the same place the incumbents refuse to look.

One honest constraint, because the product stands or falls on it: the target user's app has forty users, a scale at which no A/B test will ever reach significance. So the rule is to stop pretending to be statistics and start being observation. Six people opened your signup this week, none finished, here is a replay of one giving up. And never invent a number, not once, because the whole thing is trust, and trust does not partially survive.