I built a hidden-numbers system to stop grinder behavior in my referral program

In an effort to combat grinder behavior within his referral program, a developer implemented a hidden-numbers system. Initially, the program provided visible numbers such as total invites sent, successful invites, days earned, rank, and multiplier. However, it was soon discovered that these visible numbers were enticing users to grind the system. The developer decided to hide these numbers and replace them with qualitative messages and utility-based information.

Within 30 days of implementing the hidden-numbers system, there was a significant reduction in grinder activity. The 12% grinder rate dropped to 4%, the 8% rapid-fire rate decreased to 2%, the 4% disposable email rate diminished to 1.5%, and the 2% cluster rate fell to 0.8%. User satisfaction, measured through an NPS-style survey, also improved, with visible numbers users scoring an NPS of 31, compared to 58 for hidden numbers users.

The developer found that visible numbers created an incentive for users to optimize their stats, leading to grinding behavior. By hiding these numbers, the program removed this optimization target and reduced the incentive for grinders. Users were satisfied with qualitative feedback, such as messages like "Your invites are working" or "Your invites are starting to come together."

The developer also maintained a few visible elements, such as the last invite timestamp, cooldown reset time, and trial end date, which had utility for the users.

The hidden-numbers system proved successful over an 8-month period, with minimal adjustments needed. The final implementation of the system includes only three visible elements: the last invite timestamp, cooldown reset time, and trial end date. All other metrics, including counts, ranks, leaderboards, multipliers, and public recognition, have been removed. This approach fosters honest engagement and discourages grinding by removing the visible targets that previously attracted the worst actors.