Skill score climbs 12% per win—the tutorial ignores the other 88%
A 12% skill score gain is measurable but misses the 88% of user behavior that reveals hesitation and lost confidence
Picture a website that tells you, after every session, that your skill score has gone up 12%. Nice feeling. Now picture the same site never mentioning what happened in the other 88% of the time you were on it — the hesitation before a click, the abandoned form, the moment a user's confidence quietly collapsed. The 12% is real, measurable, and almost certainly the least interesting number on the dashboard. What would your product look like if you designed for the 88%?
The number you can see is the number you'll chase
This isn't a hypothetical. It's a well-documented pattern in product design and behavioural science: whatever gets instrumented gets optimised, and whatever gets optimised starts distorting behaviour around it. Call it the dashboard's gravitational pull.
Consider the classic 2010 study by behavioural economists at Duke and the University of Florida on "goal loathing" — when people are given a numeric target, they'll bend their behaviour to hit the number rather than improve the underlying thing the number was meant to represent. Participants in the study didn't just work harder; they changed how they worked, often in ways that undermined the original goal. The metric became the mission.
Website teams do this constantly. A win-rate metric — sessions where a user completed the intended action — climbs 12%, and everyone high-fives. Meanwhile the other 88% of user behaviour is invisible because it wasn't instrumented, wasn't named, wasn't given a colour on the chart. The 12% isn't lying to you. It's just a very small window with very good lighting.
What's actually in the 88%
The 88% isn't a single thing. It's a stew of:
- Hesitation — the cursor hovering over a button, the tab-switching, the scroll-back-up-to-re-read.
- Abandonment — forms half-filled, carts left, sign-ups started and dropped.
- Recovery — users who got confused, found their own workaround, and completed the task anyway. (These are your gold: they're telling you where your design failed and how users patched it.)
- Silent failure — users who gave up and left without a trace, because the site never gave them a reason to complain.
Most analytics setups treat all four as one grey mass called "bounce" or "exit." That's like a cricket coach only counting boundaries and ignoring the 88% of balls the batter leaves alone. The leaves tell you as much as the sixes.
Loss aversion is doing more work on your site than you think
Kahneman and Tversky's loss aversion — the finding that losses feel roughly twice as painful as equivalent gains feel good — is usually wheeled out to explain consumer pricing. But it's just as active in interface design, and it's almost entirely absent from the 12%-style dashboards.
A user who has spent four minutes filling out a form has invested. The moment they hit a validation error that wipes their progress, they don't experience it as a neutral setback. They experience it as a loss — and losses trigger abandonment far more reliably than gains trigger completion. The "win" metric (form completed) goes up 12%; the "loss" event (form reset by a bug) never gets a name.
Practical upshot: instrument the near-misses. Track how often users get within one field of completion and then leave. Track how often a validation error fires. Track time-to-recovery after an error. Those are the numbers that tell you whether your site respects the people using it, and they're the numbers your 12% is quietly hiding.
The variable-ratio trap in your own notifications
Here's where it gets uncomfortable. Variable-ratio reinforcement — the intermittent reward schedule that B.F. Skinner showed produces the most persistent behaviour — is a design pattern, not just a psychology term. Progress bars that fill unpredictably. Streaks that reset. Badges that appear at random. "Your skill score went up 12%!" notifications that arrive on a schedule the user can't predict.
These work. That's the problem. They work on your users the same way they'd work on you, and they work because they're intermittent, not because they're meaningful. If your site's reward system is teaching users to chase a number rather than achieve a goal, you've built a slot machine with a login screen. The 12% is real; the learning it represents might not be.
The design question worth asking: does your reward loop reinforce the behaviour you actually want, or does it reinforce checking the reward? Those are different behaviours, and only one of them makes your product better.
Decision-making under uncertainty is the actual user experience
Here's the thing about the 88%: it's mostly made of moments where users had to make a decision with incomplete information. Which plan? Which button? Is this the right form? Do I trust this site with my card details?
Website development has spent twenty years optimising the certain parts of the experience — page load, layout, contrast — and comparatively little on the uncertain parts. But uncertainty is where users hesitate, and hesitation is where conversions die.
A concrete example from the research: the "paradox of choice" work by Iyengar and Lepper (the jam study, 2000) found that shoppers presented with 24 jam options were less likely to buy than those presented with six — even though the larger display attracted more initial interest. More options, more uncertainty, more abandonment. Your site's product grid, pricing page, or navigation menu is a jam display. The 12% metric counts the buyers. The 88% counts the people who were interested enough to look and then paralysed by the looking.
The fix isn't fewer options. It's decision architecture: defaulting, recommending, sequencing, and — critically — measuring the moments where users pause. A site that helps people decide under uncertainty will out-convert a site that merely presents options and hopes.
Designing for the 88%
So what does it look like, practically?
Name your near-misses. Pick three events that happen just before a user abandons — error fired, back-button used, field left empty — and put them on the dashboard next to your win metric. If the win rate is 12%, the near-miss rate is the other 88%, and it deserves equal billing.
Instrument recovery, not just completion. Users who hit a problem and solved it themselves are your best research subjects. Track them. Interview them. They'll tell you exactly where your design failed and how to fix it.
Audit your reward loops for meaning. For every notification, badge, or progress indicator, ask: does this teach the user something true about their progress, or does it just feel good? Skinner's pigeons pecked for food; your users shouldn't be pecking for a percentage.
Reduce decision uncertainty, not decision count. Fewer options is one strategy. Better defaults, clearer recommendations, and progressive disclosure are usually better. The goal is confidence, not simplicity.
Watch the pause. If you can measure time-to-decision, you can find the moments where users are stuck. A 4-second pause on a pricing page is a data point; a 40-second pause is a design brief.
The 12% will keep climbing, because it's easy to make climbable. The question worth carrying into your next sprint is what the other 88% would look like if you gave it a name, a chart, and a designer's attention. My bet: it would change what you build next more than any win-rate ever has.