Skill score rises 12% per win—losses teach your onboarding nothing

Picture a skill score that jumps 12% with every win and barely budges after a loss. The number feels alive, it rewards you for showing up, and it quietly teaches you nothing about what went wrong. If you build products for Australian businesses, that pattern should feel uncomfortably familiar — because plenty of onboarding flows are wired exactly the same way.

The asymmetry hiding in your first-run experience

Behavioural economists have spent decades documenting a simple truth: humans are not symmetrical learners. Daniel Kahneman and Amos Tversky's work on prospect theory showed that losses loom larger than equivalent gains — a $100 loss stings roughly twice as much as a $100 win pleases. You'd think that asymmetry would make us obsessive students of failure. In some contexts it does. In software onboarding, it usually doesn't, because we've built the feedback loop to only fire on success.

Consider the typical activation sequence: a new user signs up, completes a setup step, sees a progress bar tick forward, gets a confetti animation. Every completed action produces a reward. Every abandoned action produces… silence. The user who quits at step three of five gets no signal at all — not from the product, not from the team, not from the analytics dashboard unless someone has deliberately instrumented the drop-off.

That's the skill-score problem in miniature. The system is calibrated to reward wins and ignore losses, so the only learning that accumulates belongs to the person having the wins. Your onboarding, meanwhile, stays exactly as dumb as it was on day one.

Variable-ratio reinforcement, or why your progress bar is doing more than you think

B.F. Skinner's work on operant conditioning gave us the term variable-ratio reinforcement: behaviour rewarded on an unpredictable schedule is the most persistent behaviour of all. It's the mechanism behind pull-to-refresh, notification badges, and the mild dopamine hum of a progress bar that fills at uneven intervals.

Here's the uncomfortable part. Variable-ratio schedules are excellent at producing repetition and terrible at producing understanding. A user will keep tapping through onboarding steps because the rewards keep arriving at unpredictable moments — but they won't learn anything about the product's actual value, and neither will you. You've optimised for completion, not comprehension.

Australian SaaS teams are particularly prone to this because our market is small enough that a few hundred activations feels like real signal. A 68% completion rate on a five-step onboarding flow looks respectable in a board deck. It tells you almost nothing about whether the 68% who finished can actually do the thing they signed up for.

What a loss actually looks like in a funnel

Losses in onboarding are rarely dramatic. They're the user who opened the pricing page three times and never returned. The one who connected their accounting software and then went quiet for eleven days. The one who filled in every field except the one that mattered and closed the tab. None of these produce a notification. All of them are teaching moments you're currently throwing away.

The 12% number and what it's really measuring

The "12% per win" framing comes from gamified skill-rating systems, where a rating climbs steadily with each success and adjusts only slightly, if at all, after a defeat. It's a design choice, and a revealing one: it makes players feel competent, which keeps them playing, which is good for engagement metrics. But it produces a rating that measures persistence more than skill.

Map that onto your product. If your onboarding only rewards completion, what you're measuring is persistence through your funnel — not whether the user has learned to use your product well. Those are different things, and the gap between them is where churn lives.

There's a well-known study in this territory worth knowing about. Research on "false feedback" in learning environments has repeatedly found that learners given inflated positive feedback perform worse on later transfer tasks than learners given accurate feedback, even when the inflated group reports higher confidence. Confidence and competence diverge precisely when the feedback loop is rigged toward encouragement. Your onboarding can be warm, welcoming, and quietly counterproductive all at once.

Designing for the loss signal

The fix isn't to make onboarding punishing. It's to make failure legible — to both the user and your team.

Instrument abandonment as a first-class event. Not just "step 3 drop-off" but why. Session replay, exit surveys triggered at the moment of hesitation, and a short "what stopped you?" prompt on the abandonment screen will tell you more in a fortnight than six months of completion-rate dashboards.

Give users a recovery path, not a restart. A user who fails a setup step and is returned to the beginning learns that failure means starting over. A user who is shown the specific field that caused the problem, with a plain-English explanation, learns something. The second experience is rarer than it should be, and it's the one that produces competent users rather than persistent ones.

Reward the recovery, not just the completion. If your progress bar only advances on success, add a visible acknowledgement when someone fixes a problem. It sounds small. It changes what the system teaches.

Run a loss audit quarterly. Pull every session where a user abandoned a critical step and read twenty of them end to end. Not the aggregate funnel — the individual sessions. You'll find patterns your dashboards have been smoothing away for months.

The competitive angle

There's a reason competitive games with honest rating systems feel stressful but fair: the rating moves down when you lose, so it means something when it moves up. Your onboarding can borrow that honesty without borrowing the stress. The goal is a system where finishing means something because the path wasn't rigged — where a completed setup actually predicts a user who can do the job.

Australian businesses competing against larger international players have a real advantage here. You can afford to talk to your users individually, read their sessions, and rebuild the loop with actual learning baked in. The big players optimise for volume and can't afford that granularity. You can.

What to change this quarter

Pick one critical onboarding step and rebuild its feedback loop so that failure produces a signal — to the user and to you. Add an abandonment prompt. Read twenty failed sessions. Change one thing based on what you find. Then measure whether the quality of completions improved, not just the rate.

The skill score that rises 12% per win is a mirror. It shows you a system that has confused encouragement with teaching. Your onboarding doesn't have to make the same mistake — but it will, quietly, for as long as losses stay invisible.