Beta testing KESHO

I invited 92 people to play KESHO during the World Cup, and 29 made at least one prediction. I knew every one of them, so this was never going to prove anything at scale.

What it could do was point. After the beta, I rebuilt what happened from the production database and looked for the patterns worth designing around. Two stood out.

Groups went live on June 12. Players created ten of them, but only one really took off: ten members who made 486 predictions after joining, 397 of them by members other than its owner.

They chose to play together, so this shows where people stayed, not why.

When a player’s rank moved, up or down, about five in six came back the next day. When it held, closer to half did. Active players had more chances to move, so it’s a pattern, not proof.

Most of the activity came from one group of ten friends. I also hadn’t properly set up PostHog, so I had to piece together what happened using database records. I could see what people did, but not where they got stuck or dropped off.

Next time, I’ll make sure I can measure how people use the product before inviting anyone to test.