Google's On-Device Measurement: what we actually found
Google's On-Device Measurement changed how we see iOS installs — and had a real impact on our Google App Install Campaign performance. Better post-install visibility, smarter bidding, fewer invisible conversions.
Originally published on Medium.
The headline numbers
Post-install growth tracked at nearly 3× the rate of spend, a clear signal that ODM surfaced a large cohort of high-intent users who were completely invisible under SKAN.
Finding 1 — ODM didn’t create installs, it reclassified them
Before ODM, 96% of our iOS installs were attributed through SKAN. ODM didn’t add new volume, but it changed how installs are seen: direct attribution jumped from 4% to 38%.
Same install volume. Better visibility.
Finding 2 — Post-install events nearly doubled
Key post-install events grew 92.7% while spend grew only 33.8% — a 3× efficiency ratio. Install growth tracked spend closely, ruling out a quality or market effect. The post-install uplift came from ODM surfacing previously invisible conversions.
The most plausible explanation: ODM gave Smart Bidding a much richer signal. Users whose in-app behaviour had been completely invisible under SKAN were now attributable, improving both measurement and algorithmic targeting.
Finding 3 — The Google vs MMP gap? Still there
ODM did not close the discrepancy between Google’s reported installs and our MMP. If anything, it widened slightly — consistent with more direct attribution flowing through Google’s ecosystem rather than the MMP.
This isn’t a red flag — it’s a known measurement artefact, consistent with our incrementality tests. Anyone reading CPI figures should treat them as a floor, not a ceiling.
Geographic constraint — EEA remains largely unchanged
ODM’s impact was concentrated in Non-EEA markets. EEA installs still fall predominantly under SKAN so the familiar measurement gaps remain in place there. Switch between tabs in the chart above to compare.
Bottom line — Worth implementing, with realistic expectations
ODM is a meaningful step forward for iOS app measurement. Better install visibility, a real uplift in post-install signals, and richer bidding data, especially outside the EEA. It won’t fix everything, and Apple’s privacy framework isn’t going anywhere. But if you’re running iOS app campaigns at scale, this is worth paying attention to.
If you’re working through your own measurement setup and want to compare notes, find me on LinkedIn.