We shipped a seven-day sign-in streak. Retention didn't move. What moved was the number of people opening the app at 11:58pm, tapping a single button, and leaving — a little herd of users diligently maintaining a streak for a product they weren't using. We'd built a second product on top of the first one, and it was busier than the real thing.
Here's the thing: the industry line is that gamification rescues retention. It mostly doesn't. Points, badges and streaks are cheap to add and easy to put on a slide, which is exactly why they show up in so many roadmaps and fix so few products. If the core action isn't worth doing on Tuesday, a coin animation won't make it worth doing on Wednesday.
The streak that did nothing
The problem wasn't that we built it badly. It's that the reward was detached from the core action. A sign-in rewards opening, not finishing the thing you came to finish. So what you end up optimising is the gesture of opening, not the value. Users learn fast to collect the reward at minimum cost: one tap at midnight. At that point the incentive has stopped being a means and become the goal, and the product is being held hostage by its own game.
Be suspicious of the phrase gamification lifts engagement. Engagement with what? If all you count is opens, then yes, a streak lifts it — but that number is manufactured by the mechanic itself. We've watched teams present the daily actives a streak brought in, which in plain terms is waking the same people up one extra time a day without them doing a single extra thing. The number isn't lying. It's the wrong question.
Rewards that sit on top vs. loops that live inside
The thing that actually works isn't a game, it's a loop, and it has to live inside the core action: the person finishes what they came for, the system gives back a signal with meaning (not coins — the thing you care about changed, someone replied, the progress moved), and that signal is itself the reason to come back. In a food-ordering mini-program, the place you always use has a new dish beats five points for showing up. In a group buy, one person short is a loop, not a badge.
There's a brutal but useful test: strip the whole rewards layer away and ask whether the core action is still worth doing. If it is, the loop already came with the product and the points are dead weight. If it isn't, anything you add is rented noise, and people leave the moment the lease runs out.
| Mechanic | What it looks like it does | What behaviour it actually changes | When to delete it |
|---|---|---|---|
| Daily sign-in streak | Brings people back every day | Trains a midnight tap-and-leave habit | When streaks rise and completed tasks don't |
| Points and a redeem shelf | Rewards loyalty | Pulls in bargain hunters who leave at payout | When redemptions cluster on one cheap item |
| Badges and levels | Signals progress | Collecting badges instead of using the product | When badges outlive any real milestone |
| A progress bar tied to the real task | Shows how far along you are | Gets the core action finished, nothing extra | Keep it; it isn't a game, it's feedback |
| A reason that comes from the content | Makes the next open obvious | Returns because something they care about changed | Never; this is the loop itself |
What it costs to keep the game running
Now the costs. Sometimes the right move is to delete a step, not add one. Users aren't complaining that things aren't fun enough, they're complaining that there are too many steps — and a sign-in is one more step. The most effective retention work we ever did was removing two screens and adding no game at all. That's an awkward thing to say in a planning meeting, because a game is easy to approve and easy to present, while deleting screens doesn't look like anything.
It can be farmed, first of all. As soon as points turn into something convertible, scripts and throwaway accounts show up. Defending against that costs money, and when you tighten the rules the first people caught are usually real ones. Points are also a liability: once issued, they have to be honoured, and they sit on the books. Change one rule and users feel owed, which is why changing the game is far harder than changing a feature — they read it as a promise, not a setting.
Then there's diminishing returns for old users. Day seven's coin isn't as interesting as day one's, so you keep raising the stake, it starts to look like discounting, and eventually the brand is just the one that hands out coupons. Not every incentive is bad: during cold start and first-run guidance, a little nudge genuinely helps. But that's scaffolding with an expiry date, not a permanent load-bearing wall. Take it down when it's due.
Try one honest test before the next brainstorm: quietly turn the points layer off for a week and watch whether anyone stops doing the core action. If they don't, you've been paying for a second product nobody needed. If they do, the problem isn't a missing badge — it's that the core action wasn't worth coming back for, and that's the part to fix.
Sources
- Nielsen Norman Group — research — why engagement metrics can end up measuring the wrong behaviour.
- Think with Google — consumer insights — what actually brings people back to a service.
- Baymard Institute — checkout research — the cost of extra steps between intent and completion.
- Apple — Human Interface Guidelines — rewards that belong to the content rather than bolted on top.