Cohort analysis (events)

Flat attendance can conceal two opposite realities: a loyal audience quietly renewing, or a leaky bucket where heavy marketing replaces everyone who churned. Cohort analysis is the only way to tell them apart. Group each edition's first-time attendees, then measure what share returned one, two, and three editions later; do the same for exhibitors. The resulting curves answer the questions that matter commercially — does our event get more valuable to people over time, or do most first-timers sample it once and leave? Which acquisition channels produce attendees who stick, and which produce one-off badge counts? That last cut changes marketing budgets: a channel with cheap registrations and terrible second-year return is expensive once you price in replacement. The practical requirement is unglamorous — stable person-level identifiers across editions, which means deduplicating registration records across name changes, job moves, and email switches before any analysis starts. The common mistake is running cohorts on email address alone and mistaking job-changers for churn. One honest nuance: annual events give you one data point per cohort per year, so curves take three or four editions to become trustworthy. Start collecting clean data now precisely because the payoff is slow.