Cohort retention curves

The talk teaches how to use dot plots to see individual users’ product activity over time, complementing aggregate metrics such as DAUs and cohort retention curves. Each row represents a user, each column a time period, and symbols mark meaningful actions or user attributes. The resulting view can reveal usage patterns, retention problems, feature associations, and at-risk B2B accounts that aggregate charts might hide. Choose events that reflect real user value and use daily or finer time periods, especially when the user base is small.

David Lee explains how to measure cohort retention by grouping new users, choosing a meaningful active-user action, and matching the measurement period to how often the product should be used. He shows how to read retention tables and curves, arguing that whether a curve flattens matters more than its height because retained users can accumulate over time. He warns against misleading metrics, including overly broad time periods, shallow actions, isolated retention figures, and misunderstood analytics dashboards. He closes with ways to improve retention through product changes, better user targeting, onboarding, and network effects, and shows how retained cohorts build a growing user base.

David Lieb recounts how Bump grew through word of mouth, then explains why its low-frequency use made it difficult to build a venture-scale business. Talking to users led the team from Bump to Flock and then PhotoRoll, whose ideas and team were acquired by Google and helped shape Google Photos. The talk emphasizes finding durable user needs, measuring retention, and testing products with users rather than focusing too early on technical polish or growth. Lieb also discusses how to build support inside a large company and how AI can solve human tasks at scale.