Usage Metric Definitions

For a cohort defined by the interval of their start event (often first-ever occurrence), the share who performed the return event in each subsequent interval.

• Output: a matrix — rows = entry interval, columns = “N intervals later”, cells = retained count / %.
• Recurring retention: active in interval N (independent per interval). The common default.
• Cumulative / “rolling” variant: active in N and every interval up to N.
• Start event and return event can differ (e.g. signed_up → ran_query).

Stickiness

For the chosen event over a fixed span (e.g. last 7 days / 4 weeks), bucket each unit by how many distinct intervals they were active, then count units per bucket.

• Output: distribution — x = number of active intervals, y = number of units.
• Surfaces power users (high-interval buckets) and feature stickiness.
• The DAU/WAU or DAU/MAU ratio is a scalar summary of the same idea.

Lifecycle

Classify each unit’s activity in each interval relative to the previous one and its first-ever activity:

• Output: stacked bars per interval (dormant negative). New + returning + resurrecting = active users; dormant shows churn out.
• Reading: dormant outpacing returning + resurrecting = leaky bucket (growth masks churn); resurrection spikes = win-back working; returning plateau = engagement ceiling.