Analyze user retention by cohort
cohort-analysisskillsetup L2★11,239
phuryn/pm-skills ↗What it does
Analyze user engagement retention patterns by cohort
Best for
Understanding long-term user engagement and identifying which cohorts churn early vs. stay engaged, informing retention experiments and feature adoption strategies.
Inputs
- · CSV/Excel/JSON cohort data (user id, cohort, time periods, engagement metrics)
- · feature adoption rates
- · retention rates over time
- · missing values check
Outputs
- · data validation summary (structure, quality, date ranges, cohort sizes)
- · quantitative analysis: retention rates, engagement trends, feature adoption curves
- · retention heatmaps (cohorts vs. time periods)
- · line charts showing cohort progression
- · feature adoption comparison charts
- · anomaly highlights and drop-off points
- · cohort performance baselines
- · Python pandas/numpy scripts (if raw data provided)
- · follow-up research recommendations: user interviews, surveys, session replays, win/loss analysis
Requires
- · pandas, numpy (for cohort calculations)
- · matplotlib, plotly (for visualization)
Preconditions
- · CSV/Excel/JSON data with cohort identifier, time periods, engagement metrics
- · At least 3–4 cohorts for meaningful pattern identification
- · Multiple time periods (for retention curves)
Failure modes
- · Insufficient data (fewer than 3 cohorts, can't identify patterns)
- · Retention curves only one period (no trajectory, no drop-off analysis)
- · Missing values not checked or dropped
- · Anomaly detection omitted (unusual patterns not flagged)
- · No follow-up research recommendations (only quantitative, not qualitative)
- · Cohort grouping unclear (which column defines cohort?)
- · Visualization colors clash or are hard to distinguish
Trust signals
- · Five-step structured process (validate → quantify → visualize → identify → recommend)
- · Heatmap + line chart + comparison chart (three visualization types)
- · Anomaly detection named explicitly
- · Cohort baseline comparison metric
- · Four research methods suggested: interviews, surveys, session replays, win/loss
- · Python scripts offered for reusability
- · Retention curve drop-off point emphasis