Mobile Gaming
Fewer events, better retention.
How a top-100 casual mobile game lifted D30 retention by eight points in 18 weeks by shrinking its LiveOps calendar and targeting the real churn cliff.
CASE / 07
Top-100 casual mobile game
The game was three years old, profitable, and bleeding D30 retention at about two points a month. Nothing dramatic was broken. That was the problem. The compound effect across twelve months of a quiet leak is worse than a noisy crisis, and nobody internally had time to stop and look at the curve.
GEO
Global
Setup
LiveOps + economy team
Duration
18 weeks
Shipped
Q3 2023
Where the leak actually was
Where the leak actually was
We pulled cohort data back 14 months and ran session-count retention, not day-count. D7 was fine. D14 was soft. D30 was the cliff. Day-count hid the real story, because session cadence varied wildly by cohort. On a session basis, the drop-off was between session eight and session fourteen, which mapped directly onto the first meaningful grind wall in the progression curve.
This was not a content problem. The game had plenty of content past the wall. It was a pacing problem. Players who hit the wall without the right resource state quit inside two sessions.
The economy was the content calendar
The economy was the content calendar
LiveOps was dropping events every two weeks. None of them were tied to progression state. Events were rewarding players who already had the resource base to participate, which was, by definition, not the players at the grind wall. Whales got more whale content. Stuck players got a push notification they ignored.
Three event types, not twelve
Three event types, not twelve
We killed nine of twelve event templates. The three that stayed were rebuilt around cohort state: a catch-up loop for players approaching the grind wall, a progression-accelerator for players stuck at it, and a returner loop for players who had lapsed between D14 and D30. Each had a written trigger condition and a rollback rule. The publisher hated this plan for the first six weeks. The first catch-up event's data shipped in week seven and settled the argument.
Segmented offers, capped at six variants
Segmented offers, capped at six variants
The publisher's ask was a 40-variant IAP matrix. The team of four could not instrument that, let alone read the results. We pushed back and settled on six variants, segmented on session count and spend state. Six was what the analysts could read cleanly inside a two-week review. Forty would have shipped, produced unreadable results, and been abandoned by month three. This was the most uncomfortable conversation of the engagement and the single most load-bearing decision.
Launch With Us