Chuseok / Golden Week retention uplift vs paid spend efficiency
Illustrative lab separating festival organic retention uplift from paid efficiency reads for KR Chuseok and JP Golden Week windows.
Festival retention uplift confounds paid efficiency reads. Korea.net Holidays documents Chuseok as a major Korean holiday; JNTO’s holiday calendar documents Japan’s Golden Week cluster. AppsFlyer’s retention and cohort analysis help defines how retention cohorts are built in that MMP — use it as the metric-definition page for this lab. Exact holiday dates must be taken from the calendar year you analyse.
Define the calculation
Compare D7 retention for a geo during the festival window with the same geo’s pre-festival baseline. Split organic-heavy versus paid install cohorts. Paid efficiency is read on paid-only ROAS or CPI→payer conversion, not on blended retention that includes holiday organic uplift.
Synthetic example
Baseline D7 retention 28.0%. Festival-window blended D7 34.0% (+6.0 pp). Paid-only festival D7 29.0% (+1.0 pp). If media reports “+6 pp retention from UA”, the synthetic split shows most uplift sits outside paid.
| Output | Worked-example result |
|---|---|
| Baseline D7 % | 28.0000 |
| Festival blended D7 % | 34.0000 |
| Blended uplift pp | 6.0000 |
| Festival paid-only D7 % | 29.0000 |
| Paid uplift pp | 1.0000 |
| Uplift misattribution pp | 5.0000 |
Evidence and limits
Holiday definitions come from Korea.net / JNTO. Retention definition is AppsFlyer’s cohort article. All percentages above are synthetic. No claim that Chuseok or Golden Week always lifts retention by 6 pp.
