A daily rewards wheel, measured with a causal study

The problem

H2Bet is a licensed betting brand in Brazil. Acquiring players was getting more expensive, and the game catalog was much the same as every competitor’s. Growth had to come from players coming back more often, not only from buying new ones.

Bonuses were the obvious lever, and the obvious risk. Bonuses cost money, attract people who only want to extract them, and in a regulated market every incentive has to respect responsible gambling rules. The goal I set was a daily habit with the cost under control, measured against a control group from the start.

What I decided

Start with one mechanic, not a program. Levels and missions were on the table. I chose a daily wheel first because it could launch fast on the rewards platform we already had, and we could learn from it quickly.

One spin a day, for everyone, expiring daily. Every logged-in player got one spin a day, with free spins as prizes. Spins expired every day, so the incentive was to come back, not to hoard. Daily players were included too: leaving out the most loyal players would feel unfair to them.

Tie the budget to revenue. The cost of each tier was set as a share of what that tier deposited, and the total budget as a share of revenue, so the wheel could never cost more than the business allowed.

Segment with data and protect against abuse from day one. Eight tiers, built on the data team’s segmentation, decided the prizes. People who deposit only to extract bonuses were identified and kept out of the paying wheels, and every tier had a weekly cap.

Write the compliance rules into the design. Rewards only for players who had already bet real money, never conditioned on a deposit, no messages that create urgency or fear of missing out, and the legal warnings in place. My own line was simple: no tactics that push people to deposit.

Roll out gradually and measure against a fair control group. The wheel went from 5% to 10% to 30% of players between July and September 2025, then to everyone. Early numbers looked great, but the players who spun were already more engaged, so I compared them only with players from the same tier who had no wheel.

Listen to players, then change the wheel. A survey with 2,541 answers showed that players preferred more spins of lower value over fewer, bigger ones. The next version raised the minimum prize from 1 to 10 free spins and moved the tiers from deposits to amount wagered, because a deposit-play-withdraw cycle was inflating the old tiers.

I owned the product; many teams made it work. I proposed the wheel, designed the cost model, tiers and prizes, configured the rewards platform, ran the survey and the experiments, and wrote the weekly reports. The data team built the segmentation, the list of bonus abusers, the dashboards and the causal study. Operations created and renewed the bonuses, engineering built the in-app entry point, and CRM, creative, compliance and support each had a part.

What went wrong, and what changed

The launch to everyone overloaded the rewards platform. We started sending the wheels in batches a few minutes apart and warning the platform ahead of every big launch.

A bonus ran out because two prize slices shared it. The fix was one bonus per slice, sized for 30 days. Later incidents added a rule: a second person reviews every configuration before it goes live.

Support tickets spiked when segments were misconfigured. Five of seventeen segments had errors. After fixing them and renewing the bonuses, tickets dropped sharply, and engineering cut calls to the rewards platform by 95%.

The result

The data team ran a causal study comparing 14 weeks before and after the full rollout, with 116,000+ players matched to similar players who did not use the wheel. It was validated with a placebo test and a check that both groups followed the same trend before the wheel.

• Deposits: +39%.
• Return on the wheel’s cost: up to 5.6x.
• Active days per week among engaged players: +55%.
• Players who used the wheel were more likely to stay in the highest-value segments.

In the survey, 57% of players were satisfied or very satisfied, and 54% said they would be very disappointed if the wheel went away.

What I took from it

Measure against a fair comparison, or you measure your own optimism. The first numbers looked better than the real effect, because engaged players spin more. The control group by tier, and later the causal study, are what made the result defensible.

A cheap mechanic can carry a strategy if the economics are designed in. The wheel worked because cost, abuse and compliance were part of the design, not fixes added later.