depth. Session length averaged 9.4 minutes. Users explored three to four sections per visit. During the rewards window, session length dropped to 3.1 minutes. Click path narrowed. People came in, completed the rewarded action, and left. Clean in. Clean out.
At first I blamed the design. Maybe the call to action was too strong. Maybe we over-optimized the funnel. But after replaying user sessions, it became obvious. They were rational. We had taught them the optimal behavior.
This is the tradeoff no one talks about when they praise incentive layers. Incentives increase activity. They can decrease curiosity.
Fabric didn’t cause that. It exposed it.
One small example: we tested a 50 point bonus for users who created a configuration template inside our app. Before rewards, only 12 percent of users created one. With the bonus, it jumped to 58 percent. Sounds great. Except when we looked 14 days later, only 9 percent of those templates were ever reused.
Templates became artifacts. Not tools.
That’s when the uncomfortable question surfaced. If we removed all rewards tomorrow, would anyone stay?
We ran a small cohort test to find out. No announcement. No drama. For 25 percent of new signups, we removed visible incentives entirely. Same product. Same features. No points. No leaderboard.
Their Day 7 retention was 23 percent. Lower than the incentivized cohort during the campaign, but higher than post-campaign decay. More importantly, their average session time was 11 minutes. They explored more. They broke things. They sent better feedback.
It felt slower. But it felt real.
Fabric’s infrastructure made it almost too easy to boost metrics. That’s not a flaw in the protocol. It’s a temptation in how we use it. When reward logic can be adjusted in hours, you start chasing response curves instead of underlying behavior change.
There were genuine improvements too. Our activation time dropped from 48 hours to 6 hours once we added small completion nudges. Users understood the initial steps faster. The friction we thought was onboarding complexity turned out to be motivation ambiguity. A 20 point reward clarified what mattered. After watching that pattern, we redesigned the onboarding even for non-incentivized users. Cleaner copy. Clearer milestones.
So incentives helped shape the product. But they also distorted the signal.
I still believe Fabric is powerful infrastructure. The coordination layer works. The accounting is precise. Distribution friction is low enough that micro-rewards are viable. That changes what experiments are possible.
But product-market fit is quieter than a spike.
It shows up when users return without being nudged. When they tolerate friction because the core action matters. When support tickets ask for deeper functionality instead of payout clarification.
Right now, we are somewhere in between. Incentives accelerate learning. They inflate vanity metrics. They can hide weak value propositions for longer than you expect.
I catch myself wanting to relaunch a bigger campaign to push DAUs past 10,000. It would probably work. For a month.
What I am less certain about is whether that would bring us any closer to the thing we actually want. #ROBO $ROBO
