The most important number in my funnel was hiding: users who scan twice convert at 72%

I run Underpriced AI — an AI pricing app for resellers (snap a photo, get what an item's worth from real sold comps, then list it). ~4,600 signups, 259 paying subscribers, and I'm solo and bootstrapped.

I've been optimizing the wrong end of the funnel for months. This week I finally cut my 90-day cohort by number of scans, and the picture was stark:

• Did exactly 1 scan: convert to paid at 3.1%
• Did 2–4 scans: 71.6%
• Did 5+: 99.2%

The whole game is the jump from one scan to two. And here's the gut-punch: 54% of my signups do exactly one scan and never come back. I was pouring effort into activation (getting people to that first scan — which is actually fine, 67% do it) when the real leak was one step later.

Worse, I was paywalling the exact moment of maximum persuasion. My only path to a free second scan was "connect eBay and publish a listing" — a mountain a browsing user won't climb. 0.4% of the one-and-done crowd ever cleared it.

So the fix I shipped: a lower rung. Record what you paid for an item (one tap) → +2 free scans. List it → +5. The "record what you paid" gate is deliberate — it's a real reseller-commitment signal, not just "scanned again," so it shouldn't cannibalize people who'd have paid anyway. I'm watching the exactly-1 vs exactly-2 split daily, with paid-conversion-at-scan-2 as the trip-wire.

Too early for results, but the lesson landed hard: activation isn't the same as your activation cliff. Cut your funnel by depth of use, find the step where conversion jumps, and check whether you're accidentally charging admission right before it.

Curious if others have found their real leak one step past where they were looking. Where was yours?