87% of streaks die 2 taps before the progress bar ends

There is a moment in every loyalty program, cashback tracker, or savings streak where the reward is close enough to taste but not yet banked. And according to a pattern that product teams quietly trade anecdotes about, roughly 87% of streaks die exactly two taps before the progress bar fills. The question worth asking isn't whether that number is precise — it isn't, and anyone who claims otherwise is selling something — but why human beings are so reliably willing to walk away from value they've almost earned.

The strange economics of almost

Behavioral economists have a name for the mental accounting that happens here, and it traces back to Kahneman and Tversky's work on prospect theory. People don't evaluate outcomes in absolute terms. They evaluate them relative to a reference point, and that reference point moves. A user who has completed eight of ten required transactions isn't thinking "I've earned 80% of a reward." They're thinking "I'm two transactions short," which feels like a debt rather than an asset.

That reframing matters enormously. Loss aversion — the well-documented tendency for losses to feel roughly twice as painful as equivalent gains feel pleasurable — kicks in on unfinished progress. The closer you get to completion, the more the unfinished portion feels like something being taken from you rather than something you simply haven't received yet. You'd expect this to drive people forward. Often it does the opposite: it produces a flinch.

What variable reinforcement actually does to persistence

B.F. Skinner's variable-ratio reinforcement schedules are the most cited finding in this territory, and for good reason. When a reward arrives after an unpredictable number of actions, behavior becomes remarkably persistent — more persistent, in fact, than when rewards arrive predictably. Slot machines are the famous example, but the mechanism shows up everywhere: email refresh, social feeds, and yes, the randomized "bonus points" that some card issuers attach to spending categories.

Here's the twist that payment product managers keep rediscovering. Variable reinforcement builds persistence in the action. It does very little for persistence in the streak. A user who doesn't know whether this transaction will trigger a bonus will keep transacting. A user who knows they need exactly two more qualifying purchases to hit a milestone has converted an open-ended habit into a closed-ended task — and closed-ended tasks are things people schedule, defer, and abandon.

The unpredictability that makes habits sticky is the same unpredictability that makes progress bars feel like homework.

The two-tap problem is a design problem, not a willpower problem

Consider a concrete case that circulated among fintech product teams a few years ago. A mid-sized issuer ran a "spend on five separate days, get a statement credit" promotion. Internal data showed a striking cliff: enormous drop-off between day three and day four of five. Not day one to day two. Day three to day four.

When researchers interviewed lapsed participants, the answer was mundane. Days one through three fit inside a normal week. Day four required a deliberate, out-of-routine action — a small purchase scheduled specifically to satisfy a program. The reward, meanwhile, was a statement credit that wouldn't appear for weeks. The mental transaction was: do something unnatural now, receive something abstract later, for a benefit I've already mentally spent.

That's the two-tap gap. It isn't laziness. It's a rational response to a badly structured incentive.

Dan Ariely's work on the "what-the-hell effect" adds another layer. When people perceive themselves as having already broken a goal — dieting, saving, exercising — they often abandon the goal entirely rather than partially salvage it. Streaks are uniquely vulnerable to this because a streak is binary. You either have it or you don't. A user who misses a day doesn't think "I completed four of five." They think "I failed," and failure licenses quitting.

Where payments intersect with risk-taking

There's a second psychological current running underneath all of this, and it's about risk tolerance. Payment products increasingly ask users to take small, bounded risks: try a new merchant, link a new account, authorize a recurring charge, accept a variable rate. Each of these is a miniature bet, and people's willingness to place miniature bets is not stable. It fluctuates with how much cognitive budget they have left, how recently they've been burned, and whether the last similar decision worked out.

Competitive play — leaderboards, tiered status, "you're in the top 12% of savers this month" — taps a different circuit. Social comparison can drive engagement, but it can also produce what researchers call ego depletion in the users who are losing. A progress bar that shows you're behind is not motivating to most people. It's a reason to close the app.

The practical implication is uncomfortable for anyone who designs these systems: the mechanics that maximize short-term completion often corrode long-term trust. A program that nudges you with artificial urgency two taps from the finish line may get the transaction. It won't get the relationship.

What actually closes the gap

The interventions that work tend to share a few properties, and none of them are exotic.

Shrink the last mile. If the final step of a streak is the hardest, make it the easiest. Some issuers now auto-apply the final qualifying transaction when a user has completed all but one — not as a giveaway, but as a recognition that the marginal cost of that last action exceeds its marginal value to the user.

Make progress recoverable. Streak freezes, grace days, and "earn it back" mechanics all attack the what-the-hell effect directly. They reframe a miss as a pause rather than a failure. Duolingo's streak freeze is the canonical example, and the retention data behind it is not subtle.

Pay out in the same channel as the effort. A reward delivered through the same app, in the same session, within seconds of the qualifying action, converts an abstract future benefit into a concrete present one. Statement credits arriving three weeks later are economically identical and psychologically invisible.

Stop showing people they're losing. Percentile rankings motivate the top decile and demoralize everyone else. If you must use social proof, use it to show that people like them are succeeding, not that they personally are behind.

The forward question

The next generation of payment loyalty won't be built on streaks at all, I suspect. It'll be built on something closer to ambient recognition — rewards that accrue without requiring the user to track a counter, and that never present a cliff to fall off. The 87% figure, wherever it came from, is really a measurement of how much friction we've been willing to tolerate in exchange for a completion metric that flatters the dashboard more than it serves the person staring at it.

The interesting work ahead isn't figuring out how to push more people over the final two taps. It's designing systems where the final two taps don't exist.