'Use every bit of the conversion value, more granularity wins.' Overfitting the schema actively hurts you under the privacy threshold.
SKAN's 6-bit value gives 64 fine values, but each finer bucket splinters your installs into smaller groups — and any group under Apple's crowd-anonymity threshold reports null. Map all 64 and you scatter installs so thin that most postbacks zero out. You traded granularity for blanks.
✓ Coarse-grained mapping (revenue tiers, key funnel steps) clears the threshold
✓ Fewer, meaningful buckets give reliable signal
✗ 64 micro-buckets fall below privacy threshold and return null
✗ Granularity you can't read is worse than coarse data you can
Model your value schema on what actually predicts LTV — usually 6-15 buckets — and leave headroom above the privacy floor.
Verdict: use a coarse, LTV-predictive schema, skip the full 64-value map — over-granularity nulls your data.
Best for: analytics teams maximizing conversion-value resolution by default.
In-App Bench
@InAppBench
Myth: map all 64 SKAdNetwork conversion values for maximum granularity
This post was published in the Telegram channel In-App Bench (@InAppBench). All its posts are on the channel page.
Open on Telegram