Country mix shift: standardising F2P ROAS before declaring creative win
Creative wins are often geo mix. Standardise ROAS with labelled country weights before celebrating a variant.
A creative “win” that is mostly a shift toward higher-ARPU countries is a mix artefact. Standardise before you scale.
Method
Use country_code on attributed installs/revenue (raw data field dictionary) and compare variants on a fixed reference mix. Attribution still assigns media-source credit under the attribution model; standardisation is a BI overlay, not a new MMP mode.
Synthetic example
Variant A looks stronger on blended ROAS (1.19 vs 0.95) because it is 70% US. On a 55% US reference mix, A is 1.085 vs B 1.049 — still ahead, but the gap shrinks. If standardised ROAS reverses, do not ship the creative as a universal winner.
Facts vs limits
| Fact | Limit |
|---|---|
| Country is available on raw attributed rows | Two-country toy mix is illustrative |
| Standardisation is operator math | Not causal; does not replace geo holdouts |
CSV.
