Account first, then three taps, then the map wakes up. The onboarding questions and the app's first act of usefulness are the same taps — we stop asking people to tick a checklist to teach the app who they are, and learn it from what they chose instead. A new user should finish onboarding knowing something about their own water that they didn't know when they opened the app.
Where the variability comes from: real data variance, not a die roll. Switch Water near them between Dense, Typical and Thin — three genuine situations a new user can be in. Everyone gets something true; not everyone gets the same thing.
| Measure | Today |
|---|---|
| Finish today's onboarding screen | 59% |
| Open the assistant within 7 days | 8.9% |
| Send anything once they open it | 10.5% |
| Prompts from the 3 canned starters | 21% |
| Location permission granted | 94.8% |
The 21% is the load-bearing number. Three generic starter prompts, aimed at one kind of angler, already carry a fifth of everything anyone asks. People will use prewritten questions — we just haven't given them ones that fit.
The 94.8% killed a screen. v1.0 of this pitch moved the location ask into context and called it the biggest win. The data says grants already run at 94.8%, so there's about four points of headroom and a working flow to break. The permission screen is gone from this prototype and location is now a guardrail. Flip the toolbar to Denied 5.2% to see the honest degraded answer.
Three taps, then the wow. Four is the ceiling. Every tap earns itself by making the answer better — the radius tap does, an "experience level" tap wouldn't.
The radius tap pays for itself twice. Choose I'll make the drive and both the count and the amount of water that lights up scale with it.
Cold start is the risk. A brand-new user has no journals and no bookmarks. Location plus season is all we get, so the answer has to be good on that alone. Only questions we're confident answering cold should be entry points.
How the answer gets produced isn't specified. It could come from the assistant, it could be a query against water and conditions data we already hold. What matters is that it's true, specific to this user, different for different people, and instant.
Signup stays firstaccepted cost The account wall — our biggest drop-off — doesn't move this cycle.
Variable reward, not a slot machine. The draw varies because the water varies. No randomised prizes, no streaks, no confetti, no manufactured scarcity, and nothing withheld to make the next tap feel better.
The drama is allowed because the data is real. Every line that lights up is a water body we hold. The reveal is theatrical about true things — that's the whole licence for it.
Never invent a find. If the data is thin, the reveal says so and gives the best true thing instead — see the Thin draw. An onboarding that overpromises to an angler is worse than one that underwhelms, because they'll drive there.
The reward is never behind the paywall. The first reveal is free, whole, and not a teaser.
Rubik 600 for display and CTA labels, Figtree 500 for body, Figtree 800 uppercase for eyebrows. Full-screen geometry: 24px sides, 32px top, 12px CTA radius, 48px minimum tap target, 44px circular close. Logo is the production asset.
One gap: there are no Paddleways tokens in the shared system — the paddle accent here is a placeholder and needs real values before any paddle work starts.