Write up a retention analysis with cohort-level insight
advancedClaude OpusData & AnalyticsData Analystretentioncohort-analysisproduct-analyticsltvdata-analyst
Use case
Use this prompt when you have a retention pull and need to deliver a memo to product or growth leadership. Pushes past 'retention curve flattens at week 6' into the segmentation and product hypotheses that drive retention work.
The prompt
You are a senior product analyst writing a retention analysis memo for a product or growth leadership audience. Inputs: - Retention metric definition:{{retention_definition}}(what counts as "retained" — e.g., "logged in week N", "took core action X in week N", "still subscribed") - Cohort grid:{{cohort_grid}}(cohorts × periods) - Behavioral segmentation if available:{{behavioral_segments}}(e.g., users who activated feature X in week 1; users who invited a teammate; users on plan tier Y) - Persona / acquisition cuts available:{{persona_cuts}}- Time range:{{time_range}}- Business context:{{business_context}}(stage, what retention drives — LTV, NRR, virality) - Audience:{{audience}}Write the memo with these sections: ## Headline (3–4 sentences) Where the retention curve is now, where it has moved, and the single most important question this analysis answers — or fails to answer. ## The Retention Curve Describe the shape: early dropoff (week 0–2), mid (week 2–8), and long-tail floor (week 8+). For each, the rate and what it tells you about the product. Note where the curve flattens — the long-tail floor is the load-bearing number for LTV. ## The Behavioral Splits That Matter Compare retention curves for users who do vs. do not take a candidate "magic moment" action in week 1. Use the behavioral segments in{{behavioral_segments}}. For each split: - Lift in week-4 and week-12 retention from doing the action - Population size of the segment that does it - Whether the lift survives controlling for plan or persona - Whether the action is plausibly causal or just a correlate of intent Be honest about correlation vs. causation. The user who invites a teammate is also a more committed user; the invite may not be the lever. ## Persona / Acquisition Differences Retention by persona, plan, or acquisition channel. Identify the segment that retains best (and at what scale) and the segment that retains worst. Quantify the gap. ## Long-Tail Floor and LTV Implications Estimate where the long-tail retention floor sits. Translate to a rough LTV implication: if floor is X% and ARPU is Y, lifetime value lands roughly in Z range. Avoid false precision — give a band. ## Where to Bet Three retention bets, ranked. For each: - The lever (e.g., make X happen for more new users in week 1) - The expected size of the prize (move floor from X% to Y%) - The risk it's a correlation, not a cause, and how to test before committing Tone: rigorous, plainspoken, willing to say "we don't know yet." The memo's job is to focus the next quarter of retention work, not to declare victory.
Variables
{{{{retention_definition}}}}Replace with your {{retention definition}}{{{{cohort_grid}}}}Replace with your {{cohort grid}}{{{{behavioral_segments}}}}Replace with your {{behavioral segments}}{{{{persona_cuts}}}}Replace with your {{persona cuts}}{{{{time_range}}}}Replace with your {{time range}}{{{{business_context}}}}Replace with your {{business context}}{{{{audience}}}}Replace with your {{audience}}
Example
Input
retention_definition: User logged in and performed at least one "core action" (created or edited a doc) in the calendar week cohort_grid: | Cohort | W1 | W2 | W4 | W8 | W12 | W24 2023-Q3 | 100% | 56% | 38% | 28% | 24% | 20% 2023-Q4 | 100% | 58% | 40% | 30% | 26% | 22% 2024-Q1 | 100% | 61% | 43% | 32% | 28% | - behavioral_segments: | Invited at least 1 teammate in week 1: 22% of users Created 3+ docs in week 1: 31% of users Used the AI assist feature at least once: 18% of users Connected an integration (Slack/Drive/etc.) in week 1: 14% of users persona_cuts: | Self-serve solo: 41% of signups Self-serve team (2+ seats): 36% Sales-assisted team (paid): 23% By acquisition: organic 38%, paid 27%, referral 22%, partner 13% time_range: Q3 2023 – Q1 2024 cohorts; observed through May 2024 business_context: B2B productivity SaaS, freemium with paid team plan, $20/user/month, current ARR $18M, retention drives both expansion and virality audience: VP of Product and Growth PM
Output
## Headline Retention has improved meaningfully across cohorts: week-4 retention rose from 38% (Q3'23) to 43% (Q1'24), and the week-24 floor moved from 20% to a projected 23–24%. The analysis points to three behavioral correlates — teammate invite, integration connection, and AI assist usage — but only the first two appear plausibly causal after controlling for plan. The unanswered question: is the long-tail floor lift driven by product changes or by an acquisition mix shift toward team plans? ## The Retention Curve - Week 0→2: drops to 56–61%. This is the activation cliff. The 5-point improvement across cohorts is the largest gain anywhere in the curve and likely reflects 2024 onboarding work. - Week 2→8: glides from ~58% to ~30%. Standard productivity-SaaS shape; the slope hasn't changed materially. - Week 8→24: flattens from ~30% to ~22%. The long-tail floor is the load-bearing LTV number, and the 2-point lift here is more economically meaningful than it looks. Curve flattens around week 8–10. Users still active at week 8 are largely long-term users. ## The Behavioral Splits That Matter - **Invited a teammate in week 1 (22% of users):** week-4 retention 71% (vs. 36% for non-inviters); week-12 retention 52% (vs. 22%). 30+ point lift sustained through week 24. Population is large enough to matter. The lift partially survives controlling for plan: even within the self-serve solo segment, inviters retain at ~58% week-4 vs. 33% non-inviters. Plausibly causal — the act of inviting creates social commitment and content others depend on. Strong candidate. - **Created 3+ docs in week 1 (31%):** week-4 retention 64% vs. 30%. Almost certainly a correlate of intent rather than a lever — users who were going to retain anyway create more docs. Suggestive but should not be the bet. - **Connected an integration (14%):** week-4 retention 68% vs. 39%. Population is small but lift is large and survives plan controls. Plausibly causal — integrations create daily-driver behavior. Worth a serious shipping bet. - **Used AI assist at least once (18%):** week-4 retention 51% vs. 41%. 10-point lift; smaller than the others and may largely be intent. Not the highest-leverage bet from this cut. ## Persona / Acquisition Differences - **Plan:** sales-assisted team plans retain at 78% W4 / 58% W12. Self-serve team at 52% / 31%. Self-serve solo at 33% / 18%. Plan tier is the largest single retention discriminator in the data. - **Channel:** referral cohort retains best at W12 (39%); paid worst (21%). Organic and partner sit between (28% and 32%). Channel-level retention gaps are stable across the three cohorts. - The Q1 cohort has a 4-point higher share of self-serve team plans than Q3 — which mechanically lifts blended retention. This is part of the apparent improvement. ## Long-Tail Floor and LTV Implications Long-tail floor lands in the 22–24% range based on the W24 cell and the slope of W12→W24 in earlier cohorts. At $20/user/month and the observed seat-expansion rate, lifetime gross revenue per signup is roughly in the $80–$130 range — a band, not a point. The Q1 lift, if durable, moves the band ~$10 higher. Worth noting that LTV is dominated by team plans; solo retention lifts have small LTV impact. ## Where to Bet 1. **Drive teammate invite in week 1.** The lift is large, the population is meaningful, and the correlation is more plausibly causal than the others. Test: a randomized prompt that boosts invite rate among solo users; measure W4 retention lift in the prompted group. Risk: invitees may be lower-quality and dilute the effect; measure both halves. 2. **Drive integration connection in week 1.** Smaller population but enormous per-user lift. Test: a streamlined integration setup in onboarding for the most-used tools (Slack, Drive). Risk: harder to instrument as a clean experiment because of integration availability differences. 3. **Move solo signups toward team plans through targeted prompts at the moment of collaboration intent.** Plan tier is the biggest retention discriminator in the data; team-plan conversion is more leverage than within-plan retention work for the solo cohort. Open question to settle before committing: how much of the Q1 retention lift is mix shift toward team plans vs. real product improvement on the underlying curve. A like-for-like cut by plan tier across cohorts would resolve this in a day.
Tips for best results
- 1Always cut the curve into early / mid / long-tail. The early cliff is an activation problem; the long-tail floor is an LTV story. Different teams own them.
- 2Behavioral splits are useful only after controlling for at least one lurking variable (usually plan or intent proxy). Otherwise you'll mistake correlation for a lever.
- 3When the company changes acquisition mix between cohorts, blended retention movement can be entirely a mix story. Always do one like-for-like cut to rule this out.
- 4Translate floor changes into LTV bands, not point estimates. A 2-point floor lift sounds small until you do the math; a point estimate sounds more confident than the data supports.
- 5Be honest about which behavioral correlates are plausibly causal. The cleanest way to mislead a product team is to label intent proxies as levers.
Related prompts
Write a narrative summary of a cohort analysis
intermediateTurn a raw cohort retention table into a clear narrative that explains what's happening across cohorts and why it matters.
Data & Analyticscohort-analysisretentionproduct-analytics
Frame a churn prediction problem
advancedFrame a churn prediction problem clearly: target definition, prediction horizon, label edge cases, and the tradeoffs between competing formulations.
Data & Analyticschurnproblem-framingmachine-learning
Generate exec narrative on top of a dashboard
advancedProduce a one-page exec narrative that sits on top of a dashboard data dump — the story leadership reads instead of squinting at the dashboard.
Data & Analyticsexecutive-reportingkpinarrative
Need help implementing this prompt in your workflow?
Book a call