Each row groups installations by their first observed UTC day. D1 is activity the next calendar day; D7 is activity exactly seven days later. — means that day is still in progress or has not arrived. Summary rates are weighted by eligible cohort size.

Install date is first seen in our event data, not an App Store installation timestamp. Highest reached uses level_start / level_end; highest completed uses level_end. Progress and session counts include all recorded activity through now, even after the selected install-date range. — means no qualifying level event.

Dates select first-seen cohorts. Only installations with 24 elapsed hours are eligible. Each step must occur at or after the previous step within those first 24 hours; equal timestamps are accepted for batched events. Missing or out-of-order events stop the funnel. Not reached does not prove the player quit. Timing uses event timestamps and first observed activity, not a verified installation time.

Dates select level starts. Outcomes are matched within the same installation and session, with matching level number/object and attempt number when supplied, up to the next start or 24 hours. Starts without an outcome remain unresolved; missing keys, simultaneous starts or conflicting outcomes are ambiguous. All rates use starts as the denominator, including unresolved and ambiguous starts. Stuck can overlap completion. Retry starts use reported attempt > 1. Duration and tool statistics use matched completions with valid properties. Object grouping combines that object across progression numbers.