Reward campaign users already spend 326% more than standard UA users.
Only if that was the number you were chasing.
Most teams launching a CPE (reward) campaign in Japan can’t say what they were actually chasing β not until three weeks in, when someone asks “did it work?” and nobody agreed on what “worked” meant before the budget went out.
The KPI Nobody Can Name
For CPI, everyone answers instantly. Target CPI. Done.
Ask the same about a CPE campaign, and the answer gets vague fast β “engagement,” “quality users,” “better than CPI.” None of these are a KPI. They’re directions without a destination.
Here’s where real numbers should set the finish line, not vague words.
What the Benchmarks Actually Say
Averages can mislead more than they help. Adjust’s 2026 Gaming App Insights Report puts average D1 retention across genres at 27% β but the same report warns that top-quartile games regularly exceed 40β50%, and that benchmarking against the average is “somewhat misleading.”
| Segment | D1 Retention | Note |
|---|---|---|
| Industry Average | 27% | All genres, 2025 data |
| Top-Quartile Games | 40β50%+ | The average was never the target |
| Japan (Country Average) | ~25% | World’s highest-retention market |
Source: Adjust, 2026 Gaming App Insights Report
A generic industry average was never the target. Your own game’s number, defined before launch, is.
When Can You Actually Judge a Campaign?
Judging a campaign takes longer than most teams give it. AppsFlyer’s own dashboards, updated this January, now treat D7 revenue β blending its data with SKAN β as the earliest checkpoint worth acting on. Adjust’s cohort-analysis guidance goes further: a cohort that’s too small can mislead no matter how many days have passed, which is why a full D0βD30 cohort view matters more than any single early snapshot.
| Checkpoint | Timing | Why It Matters |
|---|---|---|
| Day 0β2 | Launch window | Cohort too small β signal is noise, not a verdict |
| Day 7 | Earliest real checkpoint | AppsFlyer’s earliest actionable signal (Jan 2026) |
| Day 28β30 | Reliable judgment window | Full D0βD30 cohort needed to judge with confidence (Adjust) |
Sources: AppsFlyer product update (Jan 2026); Adjust cohort-analysis guide
Two days of data isn’t a verdict. It’s noise pretending to be a signal β and knowing when your data becomes trustworthy enough to judge should be decided before day one, not after the dashboard starts looking disappointing.
326% Doesn’t Happen by Chance
Reward campaign users already spend 326% more than standard UA users. But that number only means something if it was the number you were solving for β a target set before launch, judged at the right checkpoint, against your own game’s benchmark instead of an industry average.
Planning a CPE (Reward) campaign in Japan? Let’s define the finish line before you start running.
For more data and insights, visit aixpost or our LinkedIn page.
If you’re interested in running campaigns across Japan’s local media, visit the Rocket A website. You’ll find introductory materials, campaign references, and estimated simulation data tailored to your app.
π rocket-a.com