Paying for attention attracts exactly the people you don't want. What I learned building (and pivoting) Demofi
A few weeks ago I started building Demofi: indie builders discover AI tools by watching a short demo and earn rewards for it; vendors pay per session. Felt clean — two-sided, everyone wins.
The flaw I missed: paying people for attention selects for people who want the reward, not the tool. I was manufacturing interest that converted into nothing — and any vendor paying for it would've had unworkable CAC.
So I flipped it. Instead of paying builders to watch, the demo now unlocks a real perk for a tool they were already curious about. Same 2-minute demo, opposite selection — the people who finish actually want the thing. Vendors pay per real trial-start, not per view.
The lesson I keep turning over: what you ask someone to do is your real filter, not what they say. A cheap action pulls in the cheapest motivation; gating something they genuinely want flips who shows up.
I'm pre-product on purpose — proving the loop by hand with a few founders before building more. Just put up an early-access page for both sides: builders who want first dibs on perks, and founders with an AI tool.
One thing I'd love this crowd to pressure-test: does gating a wanted perk actually select for real intent, or does it just attract a classier deal-hunter who takes the perk and never sticks? Tear into it.
The insight about action as filter is the cleanest way I've seen this problem framed. The action someone takes reveals what they actually want, not what they say. To your question about the classier deal-hunter: I think the real test is what happens after the perk. If the perk is genuinely aligned with a problem they have, the drop-off should be lower than the reward-seeker version. But if the perk is just attractive in isolation, you've just moved up the quality of freeloader. The only honest filter is removing the free option entirely - but that kills top-of-funnel. Going through the same tension right now trying to validate a consumer app: how do you get genuine signal from strangers without either bribing them or shouting into a void?
Moved up the quality of freeloader" — exactly, and that's the trap I'm designing around.
I think there's a third option past bribe-or-void: make the incentive itself the filter. Cash or a generic discount is universally valuable, so it always attracts freeloaders — you can only move their quality up. But a reward that's worthless to someone without the problem has no freeloader version: a perk for a specific tool is useless to someone who doesn't want that tool. You keep top-of-funnel without bribing, because only the right person values it. And you're right that post-perk retention is the real proof — if drop-off matches the reward-seeker version, the perk was attractive in isolation and I've failed.
For a consumer app it's genuinely harder, since consumer rewards tend to be generic — the "attractive in isolation" problem baked right in. So I'd lean on the other lever you named: the action, not the reward. Make the entry require something only someone with the real problem would bother doing, and skip the universal incentive. The right people still do a costly action; the wrong ones just don't — honest filter, top-of-funnel intact.
Same trench — happy to keep comparing notes as we both find the line.