Note · Incentives
What a rewards system actually pays for
The short answer
A rewards system pays for whatever it counts, never for what you meant. If the counter is "deposits" you are buying deposits, including the ones that were going to happen anyway and the ones that reverse next week. Before launching one, write down the behaviour you want, then write down what the system will literally increment, and see whether those are the same sentence. They usually are not.
The counter is the product
Everything else in a rewards programme, the tiers, the copy, the badge art, is decoration on top of one decision: what increments. People find the counter faster than you think, and they optimise against it honestly, without any intent to cheat. A programme that counts logins gets logins. A programme that counts referrals gets referrals from people who would have come anyway. Neither outcome is a failure of the users.
Subtract what was going to happen anyway
The single most common error is crediting the reward with the whole number instead of the difference. If a cohort was already converting at some rate, the programme is only worth what it added on top of that rate. Without a holdout, a group that gets no reward, you cannot compute the difference, and every report you write afterwards will be a description of your own traffic rather than of your programme.
A snapshot audience cannot honour a later decision
If eligibility is frozen into a list at the start of a campaign, that list will eventually include someone who has since opted out, unsubscribed, closed their account or asked you to stop. The fix is not a better list. It is re-reading membership at the moment you act, every time, because membership is a fact about right now and a file is a fact about a Tuesday.
Sample size before you crown a winner
A lane with eight participants and a lane with four hundred will both produce a percentage, and the small one will produce a more exciting percentage more often. If your ranking logic does not require a minimum sample before it awards credit, you will systematically fund noise and defund the things that worked slowly.
Design the abuse case on day one
Not because most people cheat, they do not, but because the handful who do will find the cheapest path in hours, and the cost lands on your budget, not theirs. The useful exercise is to ask what you would do to maximise the counter if you were paid to and had no scruples, then decide whether you are willing to pay for that.
Keep reading