Sign-up testing bursts (several requests in the first days) are excluded; only gaps over 180 days count. "Estimated" because only Apple can confirm an actual renewal, so true counts may differ.
Each prior-year request is assumed to produce a certificate coming due 12 months later (an upper bound on what's due), so the implied rate is a lower bound. Full years only.
The <2d spike is testing noise; 180–400d matches the expected annual-renewal pattern.
Blank cells lack full follow-up.
Anonymized identifier.
Country is self-reported at signup.
YOY shape shows within-year seasonality; compare levels across years. 2016 starts mid-year; 2026 ends Sep 15.