What a cohort analysis is and why it decides your LTV
A cohort is the group of customers that started paying in the same month. Cohort analysis measures what percentage of that cohort is still active each month after subscribing. The resulting table — one cohort per row, months of life per column — is the product's X-ray. Two SaaS with identical MRR can have opposite balance sheets: one retains 85% at month 12, the other 42%. The second one doesn't have a recurring business — it has a very expensive acquisition funnel dressed up as a subscription.
The aggregate metric ("4% monthly churn") lies. It hides that the January cohort retained 78% at month 6 and the July cohort barely 54% because they came through a different channel or a different pricing. Cohort analysis is the only honest way to know whether what you changed (product, price, channel, onboarding) moved real retention, or if you only grew the numerator while the denominator was collapsing.
Formula and numeric example
Retention in month N = Active users of the cohort in month N ÷ Original cohort size NRR = (MRR at start of period + expansion − contraction − churn) ÷ MRR at start of period LTV proxy (bottom-up per cohort) = Σ (ARPU × retention_month_N × gross margin) across the modeled useful life
Example. January cohort: 500 new customers at 49 USD/month. Retention curve: month 1 = 92% (460 active), month 3 = 78% (390), month 6 = 66% (330), month 12 = 58% (290). ARPU stays flat at 49 USD, gross margin 80%. Cumulative revenue at 12 months = monthly sum of (users × 49 × 0.80). The quick calculation on the interpolated curve yields ~170,000 USD of recurring gross margin from that cohort in its first year, with an LTV proxy at 24 months close to 240,000 USD if the curve flattens at ~50% after month 12.
If that same cohort pays a blended CAC of 160 USD per customer (80,000 USD total), the payback period is month 5.8 and the 24-month LTV:CAC proxy ratio comes out at 3.0x. Three simulations with the same marketing spend, changing only the month-6 curve (from 66% to 72%), move LTV:CAC to 3.6x — a 20% difference in financial outcome for six points of retention that most dashboards don't look at.
Benchmarks by stage and segment (2025)
| Segment | 12m logo retention | 2025 median NRR | Typical flattening |
|---|---|---|---|
| B2B SMB (<50 emp.) | 72-82% | 95-105% | Month 4-6 |
| B2B mid-market | 85-92% | 108-115% | Month 8-10 |
| B2B enterprise | 92-97% | 115-130% | Month 12+ |
| B2C monthly | 40-55% | 85-95% | Month 2-3 |
| B2C annual | 65-78% | 90-100% | Month 6-9 |
| Vertical PLG | 78-88% | 110-125% | Month 6-8 |
Sources: ChartMogul SaaS Retention Report 2025, OpenView 2024 SaaS Benchmarks, Bessemer Cloud Index Q4 2025. Public-SaaS top-quartile median NRR still sits at 115-120% after the 2024-2025 macro adjustment, below the 130%+ of 2021 but stable.
Logo retention vs gross dollar retention vs NRR
- Logo retention. Accounts still active divided by accounts at start of period. Ignores how much each was paying.
- Gross dollar retention (GDR). MRR remaining from a cohort without counting expansion. Capped at 100%.
- Net revenue retention (NRR). GDR plus expansion (upsell, cross-sell, seat growth). Can exceed 100% and is the metric Wall Street weighs most for public SaaS.
A SaaS with 78% logo retention and 125% NRR has the best possible engine: it loses small accounts and the ones that stay grow fast. One with 95% logo retention and 96% NRR retains accounts but doesn't expand — the ARR-per-account ceiling is the initial price. Cohort analysis is the only way to see both signals together.
The month that matters is not month 1
Key counterintuition, documented by ChartMogul and Profitwell since 2023: the drop between month 1 and month 3 is onboarding noise and buyer's remorse — a signal of product-offer fit, not product-market fit. The drop between month 3 and month 6 is the one that predicts long-term LTV, because by then the customer has lived a full usage cycle and decides whether the product is worth renewing. Over 60% of the variance in 24-month LTV is explained by what happens in the month-3 to month-6 window, not by month 1.
Operational implication: investing in aggressive onboarding to 'save' month 1 gives a lower return than investing in activation milestones reached between weeks 6 and 18 of the contract. Most teams do the opposite because month 1 is what gets reported most.
Mistakes that destroy the business
- Looking only at average churn. A 4% monthly churn can be 2% in enterprise and 9% in SMB. The average hides where the curve is really breaking.
- Confusing expansion with retention. An NRR of 115% with 72% logo retention is an excellent enterprise land-and-expand business. The same NRR with 92% logo retention is a mediocre business that barely expands. Without cohorts, both look identical on the board slide.
- Refreshing the cohort with reactivations. If a customer churns in month 4 and comes back in month 9, they don't count as part of the original cohort: they are a new cohort. Mixing them inflates retention 8-12 points and breaks the forecast.
- Not segmenting by channel. Google Ads, Product Hunt, organic referrals, and B2B outbound produce cohorts with radically different curves. A single aggregate table convinces the CFO and confuses the growth lead.
- Stopping analysis at 12 months. True LTV reveals itself between month 18 and month 36. Cutting at 12 months underestimates enterprise cohorts and overestimates SMB cohorts.
- Ignoring net new ARR within the cohort. Seat growth inside an existing account is the cheapest growth lever. If the cohort calculator doesn't separate intra-account expansion from logo churn, the sales team has no clear targets.
When to use the simulator and when not
Use it when: you're modeling the impact of a pricing, onboarding, or channel change on LTV; you need to justify CAC to your board; you're comparing cohorts by segment, plan, or channel; you want to project MRR at 24 months bottom-up by active and projected cohorts.
It's not the fit when: you operate annual prepaid with sub-3% churn and LTV dominated by expansion (use it anyway, but the differential comes from NRR per account, not the retention curve); you run a marketplace or transactional business with highly variable month-to-month ARPU (you need frequency + basket modeling, not just a retention curve); you're pre-revenue and don't yet have 3-4 cohorts with 6+ months of history (first accumulate data, then model).
Related niches
If you operate recurring SaaS, cohort analysis pairs with monthly churn rate, the 18-24 month MRR projection, the CAC vs LTV by channel relationship, and your subscription plan architecture. The four metrics feed each other: changing pricing moves cohorts, moving cohorts changes LTV, and the new LTV rewrites the CAC you can afford to pay.
North-Star metric per business model
Not all retention cohorts should track the same engagement signal. The North-Star metric that best predicts long-run cohort LTV varies by model:
- B2C mobile apps: D1 (Day 1), D7, and D30 retention are the canonical signals. World-class D30 for consumer apps runs 20-40% (Adjust Mobile Benchmarks 2025); below 10% at D30 signals a monetization ceiling that cohort expansion cannot overcome.
- Productivity and collaboration SaaS: Weekly active users (WAU) within the first 30 days. Tools like Notion, Figma, and Linear use 'days active in first 30' as the strongest single predictor of 12-month retention. Benchmark: 60%+ of seats active in 3 of first 4 weeks.
- B2B SaaS (multi-seat, monthly active): Monthly active accounts (MAA) normalized by seat count. An account with 12 seats and 10 monthly-active users (83% seat utilization) has fundamentally different churn risk than one with 12 seats and 3 monthly active (25%). 70%+ seat utilization in month 3 predicts annual logo retention above 90% in mid-market B2B (OpenView 2024).