XP drip feeds at level 3—competence compounds at 30

Picture a Sydney café owner who has just paid a developer to bolt a loyalty widget onto her site. Points per coffee, a progress bar, a little confetti at checkout. Six weeks later, engagement is flat. The question worth asking isn't whether the widget was built well. It's whether the reward structure was built for the wrong stage of the relationship. That gap between "give people a reason to act" and "give people a reason to get good" is where behavioural psychology and website development start overlapping in ways most business owners never plan for.

The psychology of thin rewards

B.F. Skinner's work on variable-ratio reinforcement is the quiet engine behind most engagement mechanics on the web. Rewards delivered on an unpredictable schedule produce persistent behaviour — that's the finding, and it's robust. But there's a catch the pop-science versions tend to skip: variable reinforcement works hardest on behaviours that are already established and low-effort. It sustains tapping. It doesn't build skill.

Daniel Kahneman and Amos Tversky's loss aversion research adds a second layer. People weigh a potential loss roughly twice as heavily as an equivalent gain, which is why streaks, expiring points and "you'll lose your progress" nudges feel so potent. Again, though, the effect is strongest when the thing at risk is cheap to keep. A daily login streak costs nothing to maintain. It also teaches nothing.

So you get what I'd call XP drip feeds: small, frequent, predictable-ish rewards that keep a user returning without ever moving them up a competence curve. For a café loyalty scheme that might be fine — the goal is frequency, not expertise. For a business website where the user needs to learn something (configure a product, complete a multi-step application, build confidence in a professional service), drip-fed XP is a trap. You're paying attention rent on a tenant who never upgrades.

Where competence actually compounds

The shift happens somewhere around what we might call level 30 — not a literal number, but the point where a user has enough accumulated skill that the product becomes self-rewarding. They're no longer coming back for the confetti. They're coming back because they're good at it, and being good at it feels better than the confetti ever did.

This is close to what Deci and Ryan's self-determination theory describes: intrinsic motivation rests on competence, autonomy and relatedness. Extrinsic rewards can crowd out intrinsic ones if applied too early or too thickly — the classic overjustification effect. A user who starts doing something because it's interesting and then gets paid in points for it may quietly reclassify the activity as "work I do for points" and disengage when the points stop.

The practical implication for website development is uncomfortable. The mechanics that maximise early engagement are often the same ones that suppress later mastery. If your onboarding is a slot-machine of micro-rewards, you may be training users to expect a payout for every action — and the moment the payout schedule thins, so does the behaviour.

Designing for the long curve

What does a competence-first build look like? A few principles worth holding onto.

Scaffold, don't sprinkle

Instead of rewarding every micro-action, structure the experience so each step visibly increases the user's capability. A well-built quoting tool, for instance, might show a business owner their own inputs translating into a clearer picture of their costs — the reward is the insight, not a badge. Progressive disclosure (revealing complexity as the user demonstrates readiness) does more for retention than a hundred notification pings, because it respects the competence curve rather than short-circuiting it.

Make progress legible

Users need to see that they're getting better, and that's a design problem, not a psychology problem. Dashboards that show accumulated work, saved configurations, comparison history — anything that reflects investment back to the user — create what researchers call the sunk-cost-as-competence effect: the more skill you've built, the more the tool feels like yours. That's a healthier lock-in than a streak counter, because it survives a lapse in visits.

Reserve variable rewards for exploration

Variable reinforcement isn't evil; it's just misapplied. Use it to encourage discovery — surfacing a relevant case study, a template the user hasn't tried, a feature that fits their pattern of use — rather than to farm daily logins. The reward becomes "there's more here than I knew," which feeds competence instead of substituting for it.

A concrete example worth knowing

In 2010, a team of researchers led by Dan Ariely and colleagues published a study on what they called "the IKEA effect": people who assembled their own products valued them significantly more than identical pre-assembled ones. The effect held even when the assembly was clumsy and the result objectively worse. Effort, in other words, breeds attachment — but only when the effort produces something the person recognises as their own competent work. When the assembly failed entirely, the effect vanished.

That's the whole argument in miniature. Give users a real thing to build, and let them feel the build. The website's job isn't to hand out XP for showing up. It's to make the user's growing competence visible, usable and worth returning to. A café loyalty widget can't do that. A tool that helps someone run their business better can.

What to build next

If you're commissioning or rebuilding a business website this year, the forward-looking question isn't "how do we gamify this?" It's "what does our user get better at by using us, and how do we make that improvement obvious?" Start by mapping the competence curve for your main user type — what they can do on day one, day thirty, day three hundred. Then design the interface to reward the climb rather than the arrival. Audit your existing reward mechanics for overjustification risk: anything that pays out for behaviour the user would do anyway is probably training the wrong reflex. And build at least one feature that only becomes valuable after repeated use — a saved history, a comparison engine, a personalised recommendation that improves with data. That's the compounding asset. The confetti is optional.