Boosting active-user engagement across markets

01The Challenge

The team was running campaigns that hit install targets every month — but the post-install numbers told a different story. Retention was flat, ROAS was inconsistent across markets, and the agency before us was reporting on metrics that didn't change a single decision.

The goal was simple to state and harder to operate: stop chasing installs, start chasing behaviour. Make every dollar reachable from the LTV it produces.

02Uptrix's Approach

We rebuilt the optimisation goal from the floor up. Installs were demoted to a leading indicator. The new optimisation event: confirmed booking, with a 7-day attribution window. Every model — bidding, creative scoring, audience expansion — was retrained against it.

From there, the engine handled the rest: predictive LTV scoring, real-time budget reallocation, fraud screens before the auction. The team made one strategic decision per week; the model made the other 9,999.

03What We Did

• Migrated the event taxonomy to booking-anchored measurement across MMP and SKAN postbacks.
• Built lookalike audiences from the top-decile LTV cohort, refreshed weekly.
• Replaced static creatives with a 12-variant rotation tuned on post-install behaviour.
• Enabled the Uptrix fraud shield — publisher, IP, device, and CTIT signals filtered pre-auction.
• Reallocated budget on a 5-minute clock based on incremental booking lift.

04The Results

CPI dropped 34% by month two. ROAS on day-7 hit 3.8× by month three, above the original benchmark we set. D7 retention came up 62% on the cohorts the model was targeting — the acquisition was finally bringing in the right users, and the funnel started behaving like the organic one.

The team scaled into three new markets on the same engine in the following quarter. Same model, localised creative, predictable ramp.