# Cohort and Retention Analytics > Retention answers the single most important growth question: do users who join Parent: [Data Analysis](https://llms-explorer.com/tree/data-analysis/) · 22 facets · 132 facts · page: https://llms-explorer.com/tree/cohort-and-retention-analytics/ ## Overview - Retention answers the single most important growth question: do users who join keep coming back? Acquisition without retention is a leaky bucket - you pour users in the top and they fall out the bottom, so growth stalls no matter how much you spend. This skill covers the math and methods for measuring retention across cohorts (groups of users grouped by a shared start or behavior), reading retention curves, doing growth accounting (decomposing user and revenue change into its parts), and the SaaS revenue-retention metrics (NRR/GRR). It is the methods layer beneath product-led and SaaS growth. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#overview) - Scope boundary: this skill is the measurement of who stays and by how much. For funnels/activation/North-Star use da-21-product-analytics; for probabilistic CLV (BG/NBD, gamma-gamma) use da-23-customer-lifetime-value; for hazard-rate/survival modeling use da-24-survival-analysis; for experiment-driven lift use da-12-ab-testing-causal-inference. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#overview) ## 1. Acquisition vs behavioral cohorts - Acquisition (time) cohort - users grouped by when they first signed up or activated (same day/week/month). Answers when users churn and lets you compare cohort quality over time. It does not tell you why (Amplitude, Cohort Retention Analysis, 2024). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#1-acquisition-vs-behavioral-cohorts) - Behavioral cohort - users grouped by what they did (completed onboarding, used a key feature, invited a teammate), independent of join date. Answers which behaviors correlate with retention - the input to finding your activation/aha moment (Amplitude, Guide to Behavioral Cohorting, 2024; Amplitude Docs, Behavioral Cohorts). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#1-acquisition-vs-behavioral-cohorts) - Workflow: use acquisition cohorts to detect a retention problem, behavioral cohorts to diagnose and fix it (find the behavior that separates retained from churned users) (Chameleon, Cohort Analysis 101, 2024). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#1-acquisition-vs-behavioral-cohorts) ## 2. Retention-curve shapes - A retention curve plots % of a cohort still active against periods-since-start. Canonical shapes (Amplitude, Retention Curve; Churnkey, Retention Curves, 2024; Product Growth, Retention Curves Guide): — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#2-retention-curve-shapes) - Declining - slopes to zero; no group finds lasting value → no PMF. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#2-retention-curve-shapes) - Flattening - steep early drop, then levels off at a retention floor (the long-term stable %). A flattening curve is the classic signal of product/market fit: a stable set of users is hooked. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#2-retention-curve-shapes) - Smile - drops, flattens, then rises as churned users resurrect (often via network effects or re-engagement). The aspirational shape (Slack, Airbnb). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#2-retention-curve-shapes) - Dead-on-arrival - near-vertical drop to ~0 by period 2; users tried it once and never returned. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#2-retention-curve-shapes) - The retention floor (where the curve flattens) is your real long-term retention. Improving the floor (curve flattens higher) compounds far more than improving early-period retention that still decays to the same floor. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#2-retention-curve-shapes) ## 3. Retention definitions: N-day, unbounded, bracket, rolling - Choosing the definition changes the numbers dramatically - always state which you use (Amplitude, 3 Ways to Measure Retention, 2024; Mixpanel Docs, Retention; Amplitude, N-Day Retention for Mobile Games): — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#3-retention-definitions-n-day-unbounded-bracket-rolling) - N-day (classic / bounded) - % of cohort active on exactly day N. Strict; best for daily-use products. Day-2 retention = 50% means 50% came back specifically on day 2. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#3-retention-definitions-n-day-unbounded-bracket-rolling) - Unbounded ("rolling" in Mixpanel's loose sense) - % active on day N or any day after. Always ≥ N-day. Good for infrequent-use products. Note: this is not a true moving average despite the "rolling" label. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#3-retention-definitions-n-day-unbounded-bracket-rolling) - Bracket / range - % active within a custom window (Day 0; Days 1–7; Days 8–14). A flexible generalization of N-day; matches a product's natural cadence. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#3-retention-definitions-n-day-unbounded-bracket-rolling) - Rolling retention (classic survival sense) - % active on day N or later, used to estimate a survival/lifetime curve; conceptually same as unbounded. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#3-retention-definitions-n-day-unbounded-bracket-rolling) - Bounded vs unbounded asymmetry: bounded undercounts weekly/monthly-cadence products; unbounded inflates if you never re-baseline. Match the metric to the product's expected frequency. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#3-retention-definitions-n-day-unbounded-bracket-rolling) ## 4. Retention ↔ engagement - Retention is binary (active or not in a period) and is the output; engagement depth (frequency × breadth of actions) is the leading indicator. Deeper engagement → habit → retention → sustainable growth - Reforge frames retention/engagement as "the power plant of the growth model" (Reforge, Retention is the Silent Killer; Reforge, Growth Loops are the New Funnels; Conor Dewey, Reforge Recap: Engagement + Retention). Practical move: don't just track the retained/churned flag - track how deeply retained users engage, because depth predicts long-term value and is the lever you pull to raise the retention floor. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#4-retention-engagement) ## 5. Growth accounting (users) - Decompose period-over-period active users into additive components. The fundamental identity (Social Capital / Jonathan Hsu, Diligence Part 1, 2017; Amplitude, Growth Accounting, 2024): — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#5-growth-accounting-users) - new - first-ever active this period. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#5-growth-accounting-users) - retained - active last period AND this period (carried over). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#5-growth-accounting-users) - resurrected - active in some past period, inactive last period, active now. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#5-growth-accounting-users) - churned - active last period, inactive now (enters as a negative). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#5-growth-accounting-users) - User Quick Ratio (QR) = (new + resurrected) / churned. Users gained per user lost. QR > 1 means growing; rule of thumb QR ≥ ~1.5 is healthy (Hsu, Diligence Part 1, 2017; The SaaS CFO, SaaS Quick Ratio, 2024). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#5-growth-accounting-users) ## 6. Growth accounting (revenue / MRR) - Same identity applied to dollars (Social Capital / Hsu, Diligence Part 2, 2017; Lenny Rachitsky, Bottom-Up SaaS Metrics): — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#6-growth-accounting-revenue-mrr) - expansion - existing customers paying more (upsell/seats). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#6-growth-accounting-revenue-mrr) - contraction - existing customers paying less (downgrade) but not zero. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#6-growth-accounting-revenue-mrr) - churned - dropped to zero. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#6-growth-accounting-revenue-mrr) - SaaS Quick Ratio = (new MRR + expansion MRR) / (churned MRR + contraction MRR). Mamoon Hamid (Social Capital) popularized a target of QR ≥ 4 for early-stage SaaS - $4 of growth for every $1 lost (The SaaS CFO, 2024; Cobloom, SaaS Quick Ratio). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#6-growth-accounting-revenue-mrr) ## 7. DAU/WAU/MAU and stickiness - DAU / WAU / MAU - unique active users in a 1-day / 7-day / 30-day window (Mixpanel, MAU Benchmarks, 2026; Gainsight, DAU/MAU Guide). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#7-dauwaumau-and-stickiness) - Stickiness ratio = DAU/MAU (× 100) ≈ how many days/month an average monthly user shows up; DAU/MAU = 20% ≈ 6 days/month. Use WAU/MAU for products not meant for daily use (Statsig, Understanding DAU/MAU). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#7-dauwaumau-and-stickiness) - Benchmarks: daily-habit/social aim for DAU/MAU > 50%; B2B SaaS averages ~30%; <20% can be fine for infrequent products (CleverTap, DAU vs MAU, 2024). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#7-dauwaumau-and-stickiness) ## 8. Churn rate vs retention rate (and the asymmetry) - Single-period complements: Retention = 1 − Churn (Churnkey, Churn vs Retention; Orb, Churn vs Retention; Maxio, Retention vs Churn). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#8-churn-rate-vs-retention-rate-and-the-asymmetry) - The compounding asymmetry: churn compounds multiplicatively - 5%/mo churn ≈ 0.95¹² ≈ 54% retained after a year, NOT 1 − (5%×12). Always state the period and never linearly annualize. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#8-churn-rate-vs-retention-rate-and-the-asymmetry) - Customer (logo) churn ≠ revenue churn - a small customer and a whale count the same in logo churn but very differently in revenue churn. Track both. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#8-churn-rate-vs-retention-rate-and-the-asymmetry) ## 9. Revenue / dollar retention — NRR & GRR - Cohort the revenue of a customer group and measure it a year later (Drivetrain, GRR; Orb, NRR vs GRR; SaaS Capital, Good Retention Rate, 2025): — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#9-revenue-dollar-retention-nrr-grr) - GRR measures pure leak prevention (best-in-class 90–100%). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#9-revenue-dollar-retention-nrr-grr) - NRR measures retention plus expansion - NRR > 100% means a cohort grows in revenue with zero new logos (the "negative net churn" holy grail). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#9-revenue-dollar-retention-nrr-grr) - Benchmarks (private B2B SaaS, ~2024): median NRR ~100–106%; enterprise (>$100K ACV) ~118%; SMB (<$25K ACV) ~97%; best-in-class NRR > 120–130%; median GRR ~88–90% (Optifai, B2B NRR Benchmarks; Ordway, NRR Guide). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#9-revenue-dollar-retention-nrr-grr) ## 10. Sean Ellis test & power-user curve - Sean Ellis (40%) test - survey: "How would you feel if you could no longer use [product]?" If ≥ 40% say "very disappointed," you likely have product/market fit (FitSignal, Sean Ellis 40% Test; LearningLoop, Sean Ellis Score; StartupArchive, Sean Ellis on PMF). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#10-sean-ellis-test-power-user-curve) - Power-user curve (L28 / L30) - histogram of users by active days in the month (1 of 30 … 30 of 30), coined by the Facebook growth team ("Ln" = active n of last 30). A right-skewed "smile" with a heavy right tail signals strong engagement a single DAU/MAU average hides (a16z / Andrew Chen, Power User Curve, 2018; andrewchen.com). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#10-sean-ellis-test-power-user-curve) - Use together: Sean Ellis = attitudinal PMF; flattening curve + heavy-tailed power-user curve + healthy stickiness = behavioral PMF. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#10-sean-ellis-test-power-user-curve) ## 11. Building cohort tables in SQL - Canonical three-step pattern (Holistics, Cohort Retention with SQL; Cube, Cohort Retention Recipe; O'Reilly, SQL for Data Analysis ch.4): — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#11-building-cohort-tables-in-sql) - Assign each user a cohort (their first-activity period). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#11-building-cohort-tables-in-sql) - Compute period offset for every activity (period − cohort_period). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#11-building-cohort-tables-in-sql) - Pivot/aggregate counts per (cohort, offset) and divide by cohort size. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#11-building-cohort-tables-in-sql) - For zero-activity periods (gaps), build a date spine with generate_series/recursive CTE, LEFT JOIN activity, and COALESCE(...,0) so missing months render as 0 rather than vanishing (Holistics, 2024). — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#11-building-cohort-tables-in-sql) - Self-joins read more clearly for N-day retention; window functions are terser but harder to review. Pick legibility for shared analytics code. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#11-building-cohort-tables-in-sql) ## Tools / Frameworks - Amplitude / Mixpanel - built-in N-day/unbounded/bracket retention, behavioral cohorts, stickiness, power-user curves. Read the docs for the exact retention definition each uses before comparing dashboards. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#tools-frameworks) - SQL warehouse (BigQuery / Snowflake / Postgres) - DATE_TRUNC, DATE_DIFF/DATE_PART, generate_series, window functions; the portable ground truth behind any BI tool. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#tools-frameworks) - Reforge / Lenny's Newsletter / a16z (Andrew Chen) - growth loops, retention/engagement engine, power-user curve. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#tools-frameworks) - Social Capital "8-ball" growth accounting - the canonical new/resurrected/churned decomposition and quick ratio. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#tools-frameworks) - BI layer (Looker/Cube/Metabase) - cohort retention as a reusable model; Cube ships a retention recipe. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#tools-frameworks) ## Methodology - Define the active event explicitly (login? key action? value moment?). Everything downstream depends on this. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) - Pick the retention definition (N-day vs unbounded vs bracket) to match product usage frequency. State it on every chart. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) - Build acquisition cohorts, plot curves, find the retention floor. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) - Segment by behavioral cohort to find the activation behavior that lifts the floor. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) - Run growth accounting (users and MRR) to see whether growth is new-driven or retention-driven; compute the quick ratio. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) - Layer revenue retention (NRR/GRR) for monetized products. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) - Validate PMF with flattening curve + power-user tail + Sean Ellis test. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#methodology) ## Practical Patterns - Lead every retention chart with the definition + active-event + cohort granularity; otherwise numbers are uncomparable. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#practical-patterns) - Optimize the retention floor (curve shape), not just Day-1 - a higher asymptote compounds. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#practical-patterns) - Report NRR and GRR together: NRR can mask churn that expansion papers over; GRR exposes the underlying leak. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#practical-patterns) - Use WAU/MAU (not DAU/MAU) for weekly-cadence products so stickiness isn't artificially low. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#practical-patterns) - Decompose growth with the 8-ball / growth-accounting view in every business review so "we grew 10%" reveals new vs resurrected vs reduced churn. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#practical-patterns) ## Anti-Patterns - Linearly annualizing churn (5%/mo ≠ 60%/yr). Compound it. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#anti-patterns) - Comparing N-day to unbounded numbers as if equivalent - unbounded is always higher. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#anti-patterns) - Reporting only NRR and hiding gross churn behind expansion. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#anti-patterns) - A single DAU/MAU average masking a bimodal power-user split - show the histogram. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#anti-patterns) - Day-1-retention obsession while the curve still decays to the same floor. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#anti-patterns) - Ignoring the survivorship of recent cohorts - the newest cohort has no long-tail data yet; don't compare its Day-30 to an old cohort's before 30 days have elapsed. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#anti-patterns) ## Troubleshooting - Retention "improved" suspiciously → check whether the active-event definition or the retention type (bounded↔unbounded) changed. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#troubleshooting) - Recent cohorts look worse → likely right-censoring, not real decline; only compare offsets that have fully matured for all cohorts shown. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#troubleshooting) - NRR > 100% but business feels shaky → inspect GRR and logo churn; expansion from a few whales can hide broad SMB churn. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#troubleshooting) - Curve never flattens → no PMF in that segment; re-segment by behavioral cohort to find a sub-population that does flatten. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#troubleshooting) - SQL retention has gaps/jumps → you're missing zero-activity periods; add a date spine + COALESCE. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#troubleshooting) - Stickiness looks terrible → wrong window; switch DAU/MAU → WAU/MAU for infrequent products. — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#troubleshooting) ## References - Amplitude - Cohort Retention Analysis (2024): https://amplitude.com/blog/cohorts-to-improve-your-retention — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Amplitude - Guide to Behavioral Cohorting (2024): https://amplitude.com/blog/guide-to-behavioral-cohorting — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Amplitude Docs - Behavioral Cohorts: https://amplitude.com/docs/analytics/behavioral-cohorts — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Amplitude - Retention Curve: https://amplitude.com/explore/analytics/retention-curve — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Amplitude - 3 Ways to Measure Retention (2024): https://medium.com/@amplitudeHQ/3-ways-to-measure-user-retention-2af5e4e82a45 — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Amplitude - N-Day Retention for Mobile Games: https://amplitude.com/blog/n-day-retention-for-mobile-games — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Amplitude - Growth Accounting (2024): https://amplitude.com/blog/growth-accounting — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Mixpanel Docs - Retention: https://docs.mixpanel.com/docs/analysis/reports/retention — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Mixpanel - MAU Definition & 2026 Benchmarks: https://mixpanel.com/blog/mau/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Churnkey - Retention Curves (2024): https://churnkey.co/blog/retention-curves/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Churnkey - Churn Rate vs Retention Rate: https://churnkey.co/blog/churn-rate-vs-retention-rate/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Product Growth - Retention Curves Guide: https://productgrowth.in/resources/guides/retention-curves-guide/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Chameleon - Cohort Analysis 101 (2024): https://www.chameleon.io/blog/cohort-analysis — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Reforge - Retention is the Silent Killer: https://www.reforge.com/blog/retention-engagement-growth-silent-killer — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Reforge - Growth Loops are the New Funnels: https://www.reforge.com/blog/growth-loops — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Conor Dewey - Reforge Recap: Engagement + Retention: https://www.conordewey.com/blog/reforge-engagement-retention — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Social Capital / Jonathan Hsu - Diligence Part 1: Accounting for User Growth (2017): https://medium.com/swlh/diligence-at-social-capital-part-1-accounting-for-user-growth-4a8a449fddfc — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Social Capital / Jonathan Hsu - Diligence Part 2: Accounting for Revenue Growth (2017): https://medium.com/swlh/diligence-at-social-capital-part-2-accounting-for-revenue-growth-551fa07dd972 — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Lenny Rachitsky - Most Important Bottom-Up SaaS Metrics: https://www.lennysnewsletter.com/p/the-most-important-bottom-up-saas-69d — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - The SaaS CFO - SaaS Quick Ratio (2024): https://www.thesaascfo.com/saas-quick-ratio/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Cobloom - SaaS Quick Ratio: https://www.cobloom.com/blog/saas-quick-ratio-how-to-measure-your-startups-revenue-health — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Gainsight - DAU/MAU Guide: https://www.gainsight.com/essential-guide/product-management-metrics/dau-mau/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Statsig - Understanding DAU/MAU: https://www.statsig.com/perspectives/understanding-daumau-key-metrics-for-product-success — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - CleverTap - DAU vs MAU (2024): https://clevertap.com/blog/dau-vs-mau-app-stickiness-metrics/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Orb - Churn vs Retention Rate: https://www.withorb.com/blog/churn-rate-vs-retention-rate — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Orb - NRR vs GRR: https://www.withorb.com/blog/nrr-vs-grr — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Maxio - Retention vs Churn: https://www.maxio.com/saaspedia/retention-rate-vs-churn-rate — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Drivetrain - Gross Revenue Retention: https://www.drivetrain.ai/strategic-finance-glossary/what-is-gross-revenue-retention-formula-benchmarks — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - SaaS Capital - Good Retention Rate (2025): https://www.saas-capital.com/blog-posts/what-is-a-good-retention-rate-for-a-private-saas-company/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Optifai - B2B SaaS NRR Benchmarks: https://optif.ai/learn/questions/b2b-saas-net-revenue-retention-benchmark/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Ordway - NRR Guide: https://ordwaylabs.com/resources/guides/net-revenue-retention-guide/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - FitSignal - Sean Ellis 40% Test: https://www.fitsignal.com/blog/sean-ellis-40-percent-test — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - LearningLoop - Sean Ellis Score: https://learningloop.io/glossary/sean-ellis-score — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - StartupArchive - Sean Ellis on PMF: https://www.startuparchive.org/p/sean-ellis-on-how-to-tell-if-you-have-product-market-fit — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - a16z / Andrew Chen - The Power User Curve (2018): https://a16z.com/the-power-user-curve-the-best-way-to-understand-your-most-engaged-users/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - andrewchen.com - The Power User Curve: https://andrewchen.com/power-user-curve/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Holistics - Calculate Cohort Retention with SQL (2024): https://www.holistics.io/blog/calculate-cohort-retention-analysis-with-sql/ — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - Cube - Cohort Retention Recipe: https://cube.dev/docs/product/data-modeling/recipes/cohort-retention — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) - O'Reilly - SQL for Data Analysis, ch.4 Cohort Analysis: https://www.oreilly.com/library/view/sql-for-data/9781492088776/ch04.html — [source](https://llms-explorer.com/sources/mdb-context-hub/da-34-cohort-retention-analytics/#references) ## Where this helps - Answering whether users who join a product keep coming back, the core growth question retention analytics is built to measure, distinct from acquisition or activation. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Deciding which retention definition to report — N-day, unbounded, bracket, or rolling — and needing to know they produce materially different numbers, so the choice has to be stated explicitly on every chart. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Revenue is growing but churn might be masked underneath it; comparing NRR, which includes expansion, against GRR, which measures pure leak prevention, surfaces that gap. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Sanity-checking a claimed product/market-fit signal using the Sean Ellis 40% test alongside a flattening retention curve and a heavy-tailed power-user curve, rather than trusting any one signal alone. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Project ideas - Build a cohort table in SQL using the canonical three-step pattern: assign each user a cohort by first-activity period, compute period offsets, and aggregate into a retention curve. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Implement both N-day and unbounded retention definitions on the same dataset and quantify how much the two numbers diverge for a given usage pattern. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Compute growth accounting on both users and revenue, new, retained, resurrected, churned, expansion, contraction, to decompose a period-over-period change into its component parts. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Build a DAU/WAU/MAU stickiness dashboard and benchmark the resulting ratio against category norms, such as roughly 30% for B2B SaaS, to judge whether it's healthy for the product type. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Common mistakes - Linearly annualizing a monthly churn rate, treating 5% per month as 60% per year, instead of compounding it correctly, which understates the true annual churn. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Comparing an N-day retention number directly to an unbounded retention number as if they were the same metric; unbounded is always higher by construction. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Reporting only NRR and omitting gross churn, which lets expansion revenue from a few large accounts mask broad churn among smaller ones. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Treating a recent cohort's worse-looking numbers as a real decline without checking for right-censoring; recent cohorts haven't had time to mature, so only fully-matured offsets should be compared across cohorts. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Known issues - Retention numbers are only comparable when the active-event definition and the retention type, bounded versus unbounded, are held constant; a metric that improved can just reflect a silent definition change, not real behavior change. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - NRR above 100% can still coexist with a shaky business if a few whale accounts' expansion is papering over broad SMB churn; the aggregate metric alone won't reveal that. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - The Sean Ellis 40% test measures attitudinal product/market fit, what users say, which can diverge from behavioral PMF, what the retention and power-user curves actually show, so the two should be read together. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* - Growth accounting and cohort tables require a precisely defined active event; everything downstream, curves, stickiness ratios, churn, inherits any ambiguity in that definition. — [source](https://llms-explorer.com/tree/cohort-and-retention-analytics/) *(AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.)* ## Context files - [Cohort and Retention Analytics](https://llms-explorer.com/downloads/sources/mdb-context-hub/da-34-cohort-retention-analytics.md)