App / Mobile Gaming · Performance Marketing
PocketForge Labs dropped install costs 34% by paying for players, not installs
34% lower cost per paying user
A mobile puzzle studio was scaling installs that never monetised. We rebuilt UA around day-7 predicted LTV.
34%
Lower cost per paying user
118%
Day-30 ROAS
40
Creative variants per month
11
Focused markets
The challenge
PocketForge Labs celebrated a ₹93 CPI while day-30 ROAS languished at 41%. Post-ATT, their measurement was thin and their creative testing was manual, so they had no idea which of their 11 campaigns were buying players versus tourists.
Strategy
Shift the optimisation event from install to a predicted-value proxy, then use SKAN cohorts and a probabilistic model to grade creatives on day-7 revenue rather than install volume.
Execution
We instrumented a day-1 engagement signal that predicted 78% of day-30 revenue variance, moved all campaigns to value-based bidding on that event, and rebuilt the creative pipeline to ship 40 playable and video variants a month. Underperforming geos were cut from 34 markets to 11.
Results
CPI rose to ₹153 — and cost per paying user fell 34%. Day-30 ROAS reached 118%, taking the studio contribution-positive on paid UA for the first time in its history.
“Cheap installs were the most expensive thing we ever bought.”
Rune Halvorsen — CEO, PocketForge
