SKAN null-rate charts need population and definition before a “healthy” target
AppsFlyer’s ~11% average null CV rate is a vendor-sample observation with a dated window—not a universal health threshold.
AppsFlyer has publicly described an average SKAN null conversion-value rate around 11% across networks in its privacy-threshold analysis, and its dashboard docs define null rate as null-CV postbacks divided by installs. That is a vendor definition and a dated sample observation—not a “healthy” target every game should hit. Modeling post; SKAN dashboard.
Definition before target
Null is not zero. AppsFlyer’s materials distinguish null (CV withheld) from zero (install with no measurable action in the window). Null rates also vary with install volume against Apple’s undisclosed privacy threshold, campaign structure and SKAN version. A chart that omits population, period, OS version and whether modeled metrics are included cannot set a studio SLA.
The ~11% figure comes from an AppsFlyer analysis published in 2022 describing conditions after 2021 spikes. It is not automatically current for every network in 2026.
Refuse unmatched health scores
Our recommendation is to track your own null rate with the same formula and filters over time. Do not import a vendor average as a pass/fail threshold without matching population and definition.
Inspect the evidence
Download the source-population and comparability card. The extraction separates documented facts from our analysis and unknowns.
| Field | Evidence or limit |
|---|---|
| Report / publisher | AppsFlyer SKAN null-CV analysis and dashboard docs / AppsFlyer |
| Period | Vendor analysis describing post-2021 stabilisation around ~11% (article dated 2022-11-07); not a 2026 census |
| Population | AppsFlyer-observed SKAN postbacks across networks in the described analysis |
| Metric under review | Null conversion value rate ≈ postbacks with null CV / installs × 100; ~11% average cited in analysis |
| Aggregation | Cross-network average in vendor analysis; account dashboards are filterable |
| Critical limit | The ~11% figure is not a universal healthy target; period, network mix and modeled vs unmodeled reporting must match before comparison. |
| Comparable with | The same null-rate formula on a matched iOS/SKAN population and period |
| Not comparable with | A studio SLA copied from an unmatched vendor chart or a zero-CV rate |
Further reading in the existing archive: Blended ROAS Calculation: Where It Quietly Goes Wrong; The benchmark report is selling you something. These links provide background; this check does not independently verify their full contents.
