Appearance
Cohort Analysis & Retention
Averages lie. Cohorts tell the truth about whether your product is actually getting better at keeping customers.
Why This Matters
- 🏢 Owner: Cohort analysis reveals if your business is improving over time. Two companies with the same overall churn can have wildly different trajectories — one getting better, one getting worse — and only cohorts reveal which is which.
- 💻 Dev: Cohort data shows whether product changes actually impact retention. That big feature launch in March — did customers acquired after it stick around longer?
- 📋 PM: Retention curves show exactly when customers drop off, pointing you to the moments in the journey that need the most attention.
- 🎨 Designer: Cohort data connects design changes to behavioral outcomes. A redesigned onboarding in Q2 should show up as better retention for Q2+ cohorts.
The Concept (Simple)
Imagine you are a teacher tracking how many students stay enrolled in your class over time. Instead of looking at all students together, you group them by when they started:
- September class: 30 students enrolled, 25 still here in October, 22 in November...
- October class: 28 students enrolled, 26 still here in November, 24 in December...
If the October class retains better than the September class, something you changed (curriculum, teaching style, materials) is working. That is cohort analysis.
A "cohort" is simply a group of customers who share
a common starting point — usually the month they signed up.
┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐
│ Jan 2026 │ │ Feb 2026 │ │ Mar 2026 │ │ Apr 2026 │
│ Cohort │ │ Cohort │ │ Cohort │ │ Cohort │
│ │ │ │ │ │ │ │
│ 100 new │ │ 120 new │ │ 95 new │ │ 140 new │
│customers │ │customers │ │customers │ │customers │
└──────────┘ └──────────┘ └──────────┘ └──────────┘
│ │ │ │
▼ ▼ ▼ ▼
Track each group separately over time...How It Works (Detailed)
Building a Cohort Retention Table
A cohort table tracks what percentage of each cohort is still active (or still paying) at each subsequent time period.
Logo (Customer) Retention Cohort Table:
╔═══════════╦═══════╦═══════╦═══════╦═══════╦═══════╦═══════╦═══════╗
║ Cohort ║ Mo 0 ║ Mo 1 ║ Mo 2 ║ Mo 3 ║ Mo 4 ║ Mo 5 ║ Mo 6 ║
╠═══════════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╣
║ Jan 2026 ║ 100 ║ 82 ║ 71 ║ 65 ║ 61 ║ 58 ║ 56 ║
║ (100 cust)║ 100% ║ 82% ║ 71% ║ 65% ║ 61% ║ 58% ║ 56% ║
╠═══════════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╣
║ Feb 2026 ║ 120 ║ 101 ║ 89 ║ 82 ║ 78 ║ 75 ║ ║
║ (120 cust)║ 100% ║ 84% ║ 74% ║ 68% ║ 65% ║ 63% ║ ║
╠═══════════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╣
║ Mar 2026 ║ 95 ║ 83 ║ 74 ║ 69 ║ 66 ║ ║ ║
║ (95 cust) ║ 100% ║ 87% ║ 78% ║ 73% ║ 69% ║ ║ ║
╠═══════════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╬═══════╣
║ Apr 2026 ║ 140 ║ 126 ║ 115 ║ 108 ║ ║ ║ ║
║ (140 cust)║ 100% ║ 90% ║ 82% ║ 77% ║ ║ ║ ║
╚═══════════╩═══════╩═══════╩═══════╩═══════╩═══════╩═══════╩═══════╝How to read this table:
- Read across (→) to see how a single cohort ages. The Jan cohort lost 18% in month 1 but only lost 3% between months 5 and 6 — churn is front-loaded.
- Read down (↓) to see if retention is improving. At Month 1: Jan=82%, Feb=84%, Mar=87%, Apr=90%. Each cohort retains better, meaning the product is improving.
- The diagonal represents your current month — all cohorts at their most recent measurement.
Reading Retention Curves
The cohort table above translates into a retention curve — a visual representation of how each cohort decays over time.
Retention
%
100% ─┬─────────────────────────────────────────────────
│╲╲
│ ╲╲╲
90% ─┤ ╲╲╲──── Apr cohort (best)
│ ╲╲╲╲
│ ╲╲ ╲╲──── Mar cohort
80% ─┤ ╲╲ ╲╲
│ ╲╲ ╲╲──── Feb cohort
│ ╲╲ ╲╲
70% ─┤ ╲╲ ╲╲──── Jan cohort (worst)
│ ╲╲ ╲╲
│ ╲╲ ╲───────────────── Flattening
60% ─┤ ╲────────────────────── = good sign
│ ╲──────────────────────
│
50% ─┤
│
└─────┬─────┬─────┬─────┬─────┬─────┬──────▶ Month
1 2 3 4 5 6
GOOD SIGNS: BAD SIGNS:
✅ Curves flatten over time ❌ Curves keep declining steadily
✅ Newer cohorts sit higher ❌ Newer cohorts sit lower
✅ Steep only in month 1-2 ❌ Steep drops at months 3-6+The Three Shapes of Retention Curves
1. FLATTENING (Healthy) 2. DECLINING (Dangerous) 3. SMILING (Exceptional)
100%│╲ 100%│╲ 100%│╲
│ ╲ │ ╲ │ ╲
│ ╲ │ ╲ │ ╲
│ ╲ │ ╲ │ ╲
│ ╲─────────── │ ╲ │ ╲
│ (plateau) │ ╲ │ ╲
│ │ ╲ │ ╲ ╱── (expansion
│ │ ╲ │ ╲╱ drives NRR
└──────────────▶ │ ╲ │ ╱ above 100%)
└─────────╲──▶ └──────────▶
You found your core Never finds a floor. Revenue grows even as
users. They stick. Product-market fit issue. some customers leave.Dollar Retention vs. Logo Retention
These two views tell fundamentally different stories.
Logo retention = What percentage of customers are still active? Dollar retention = What percentage of revenue is retained (including expansion)?
╔══════════════════════════════════════════════════════════════════════╗
║ SAME COHORT, TWO STORIES ║
║ ║
║ Jan 2026 Cohort: 100 customers, $15,000 starting MRR ║
║ ║
║ After 6 months: ║
║ ───────────────────────────────────────────────── ║
║ Customers remaining: 56 out of 100 ║
║ Logo retention: 56% ← Looks bad ║
║ ║
║ BUT: ║
║ - 10 of the 56 upgraded: +$3,000/mo ║
║ - 5 added seats/modules: +$1,200/mo ║
║ - Revenue from remaining 56: $13,500/mo ║
║ - Including expansions: $17,700/mo ║
║ ║
║ Dollar retention: 118% ← Looks great ║
║ ║
║ Interpretation: You're losing many small customers but your ║
║ retained customers are becoming MORE valuable. This may be fine ║
║ (or it may hide a problem with your entry-level tier). ║
╚══════════════════════════════════════════════════════════════════════╝Dollar Retention Cohort Table:
╔═══════════╦════════╦════════╦════════╦════════╦════════╦════════╗
║ Cohort ║ Mo 0 ║ Mo 1 ║ Mo 2 ║ Mo 3 ║ Mo 4 ║ Mo 5 ║
╠═══════════╬════════╬════════╬════════╬════════╬════════╬════════╣
║ Jan 2026 ║ 100% ║ 92% ║ 89% ║ 91% ║ 96% ║ 103% ║
║ ($15,000) ║ ║ ║ ║ ↑ ║ ↑ ║ ↑ ║
║ ║ ║ ║ ║ expansions kicking in ║
╠═══════════╬════════╬════════╬════════╬════════╬════════╬════════╣
║ Feb 2026 ║ 100% ║ 94% ║ 92% ║ 95% ║ 101% ║ ║
║ ($18,000) ║ ║ ║ ║ ║ ║ ║
╠═══════════╬════════╬════════╬════════╬════════╬════════╬════════╣
║ Mar 2026 ║ 100% ║ 95% ║ 94% ║ 98% ║ ║ ║
║ ($14,250) ║ ║ ║ ║ ║ ║ ║
╚═══════════╩════════╩════════╩════════╩════════╩════════╩════════╝Dollar retention rising above 100% in later months is the hallmark of excellent SaaS businesses. It means expansion revenue from surviving customers exceeds the revenue lost to churn.
Identifying Retention Inflection Points
An inflection point is a moment in the customer lifecycle where churn behavior changes significantly. Finding and fixing these is one of the highest-leverage activities in SaaS.
How to spot inflection points in the data:
Monthly churn rate BY lifecycle month (Jan 2026 cohort):
Month Retained Lost Period Churn
────── ──────── ──── ───────────
0 → 1 82 18 18.0% ◄── INFLECTION: Activation problem
1 → 2 71 11 13.4%
2 → 3 65 6 8.5% ◄── Still elevated
3 → 4 61 4 6.2%
4 → 5 58 3 4.9% ◄── Stabilizing
5 → 6 56 2 3.4% ◄── Approaching steady-state
┌─────────────────────────────────────────────────────┐
│ Period Churn by Month │
│ │
│ 18% │ ██ │
│ │ ██ │
│ 15% │ ██ │
│ │ ██ ██ │
│ 12% │ ██ ██ │
│ │ ██ ██ │
│ 9% │ ██ ██ ██ │
│ │ ██ ██ ██ │
│ 6% │ ██ ██ ██ ██ │
│ │ ██ ██ ██ ██ ██ │
│ 3% │ ██ ██ ██ ██ ██ ██ │
│ │ ██ ██ ██ ██ ██ ██ │
│ 0% └──┴──┴──┴──┴──┴──┴── │
│ M1 M2 M3 M4 M5 M6 │
└─────────────────────────────────────────────────────┘Common inflection points and their causes:
| When (Month) | Likely Cause | Fix |
|---|---|---|
| Month 0–1 | Failed activation / poor onboarding | Redesign first-run experience, add guided setup |
| Month 1–3 | Did not reach "aha moment" | Identify and shorten time-to-value |
| Month 3–6 | Initial use case solved, no deeper value | Introduce advanced features, integrations |
| Month 6–12 | Annual planning cycle — budget review | Demonstrate ROI before renewal conversations |
| Month 12 | Annual contract renewal decision | Proactive success outreach at month 10 |
In Practice
Step-by-Step: Building Your First Cohort Analysis
Step 1: Define your cohort grouping
Most common: Month of first subscription payment
Alternatives:
- Week of signup (for high-volume consumer SaaS)
- Quarter of signup (for enterprise with fewer customers)
- Channel of acquisition (organic vs. paid vs. referral)
- Plan tier at signup (free vs. starter vs. pro)Step 2: Define your retention metric
┌────────────────────────────────────────────────────────────────┐
│ Choose ONE primary metric for your cohort table: │
│ │
│ [ ] Logo retention — Is the customer still active? │
│ Best for: Understanding activation and engagement │
│ │
│ [ ] Revenue retention — What % of MRR is retained? │
│ Best for: Understanding business health and NRR │
│ │
│ [ ] Usage retention — Is the customer still using the product?│
│ Best for: Leading indicator of future churn │
│ │
│ (Track all three, but pick one as your North Star) │
└────────────────────────────────────────────────────────────────┘Step 3: Pull the data
You need two data points per customer:
- When they started (cohort assignment)
- Whether they were active in each subsequent period
sql
-- Simplified SQL for building a cohort table
SELECT
DATE_TRUNC('month', first_payment_date) AS cohort_month,
DATEDIFF('month', first_payment_date, activity_date) AS months_since_start,
COUNT(DISTINCT customer_id) AS active_customers
FROM subscriptions
WHERE status = 'active'
GROUP BY 1, 2
ORDER BY 1, 2;Step 4: Calculate percentages
Divide each cell by the Month 0 count for that cohort to get retention percentages.
Step 5: Visualize and compare
Plot the retention curves for each cohort on the same chart. Look for the patterns described above.
How to Act on Cohort Data
The whole point of cohort analysis is to drive action. Here is a decision framework:
┌───────────────────────────────────────────────────────────────────┐
│ COHORT DATA → ACTION FRAMEWORK │
│ │
│ OBSERVATION ACTION │
│ ──────────────────────────────── ─────────────────────────── │
│ │
│ Month 1 drop > 15% → Fix onboarding │
│ (Activation failure) See ../03-product/*.md │
│ │
│ Newer cohorts worse than older → Investigate acquisition │
│ (Quality declining) channel quality changes │
│ │
│ Newer cohorts better than older → Recent product changes │
│ (Improvement signal) are working — double down │
│ │
│ All cohorts drop at same month → Lifecycle-stage problem │
│ (Structural churn point) Add intervention at that │
│ month (email, call, etc.) │
│ │
│ Revenue retention > logo retention → Expansion is healthy, but │
│ by wide margin investigate SMB/entry churn │
│ │
│ Revenue retention < 100% at 12 mo → Need expansion motion │
│ (Net contraction) See ../06-growth/*.md │
│ │
│ One channel's cohorts retain → Shift spend toward that │
│ significantly better channel │
└───────────────────────────────────────────────────────────────────┘Example: Using Cohorts to Evaluate a Product Change
Context: In March 2026, you redesigned the onboarding flow.
BEFORE (Jan & Feb cohorts):
┌──────────┬───────┬───────┬───────┐
│ Cohort │ Mo 0 │ Mo 1 │ Mo 2 │
├──────────┼───────┼───────┼───────┤
│ Jan 2026 │ 100% │ 82% │ 71% │ ← 18% first-month drop
│ Feb 2026 │ 100% │ 84% │ 74% │ ← 16% first-month drop
└──────────┴───────┴───────┴───────┘
AFTER (Mar & Apr cohorts):
┌──────────┬───────┬───────┬───────┐
│ Cohort │ Mo 0 │ Mo 1 │ Mo 2 │
├──────────┼───────┼───────┼───────┤
│ Mar 2026 │ 100% │ 87% │ 78% │ ← 13% first-month drop ✅
│ Apr 2026 │ 100% │ 90% │ 82% │ ← 10% first-month drop ✅
└──────────┴───────┴───────┴───────┘
The onboarding redesign improved Month 1 retention by 5-8
percentage points. At 100+ customers per cohort, this
translates to 5-8 additional retained customers per month.
Annual impact estimate:
~6.5 extra retained customers/month × 12 months × $150 ARPA
= $11,700 additional ARR from reduced churn alone.Common Mistakes in Cohort Analysis
| Mistake | Problem | Solution |
|---|---|---|
| Cohorts too large (quarterly) | Hides month-to-month variation | Use monthly cohorts unless volume is very low |
| Cohorts too small (weekly) | Noisy data, hard to draw conclusions | Weekly only if 100+ customers per week |
| Only tracking logo retention | Misses expansion/contraction dynamics | Always pair with dollar retention |
| Not segmenting beyond time | All customers treated equally | Slice by plan, channel, company size |
| Waiting too long to analyze | Decisions made on gut feel instead of data | Set up automated monthly cohort reports |
| Confusing correlation with causation | "We launched X in March, March cohort is better" | Control for other variables (seasonality, etc.) |
Key Takeaways
- Cohort analysis groups customers by their start date and tracks their behavior over time. It reveals trends that aggregate metrics completely hide.
- Read the table in two directions: across (how a cohort ages) and down (whether newer cohorts are better or worse).
- Retention curves should flatten — if they keep declining linearly, you have not yet found product-market fit for a core user segment.
- Dollar retention and logo retention tell different stories — you need both. High dollar retention with low logo retention means your product works for power users but fails for casual ones.
- Inflection points in the retention curve point to specific lifecycle moments where intervention is most valuable.
- The purpose of cohort analysis is action — every insight should lead to a product, marketing, or operational change.
Action Items
- [ ] 🏢 Owner: Request a monthly cohort retention report from your data or product team. Review it in every monthly business review alongside the metrics from SaaS Metrics That Matter.
- [ ] 🏢 Owner: Set a company target for Month 1 and Month 12 retention rates. Track them as OKRs alongside MRR and NRR.
- [ ] 💻 Dev: Build the data pipeline to support cohort analysis — ensure every customer has a reliable
first_payment_dateorsignup_dateand that subscription status is tracked historically, not just as current state. - [ ] 💻 Dev: Create an internal dashboard or reporting query that auto-generates the cohort table monthly. Manual spreadsheets do not scale.
- [ ] 📋 PM: Identify the top inflection point in your retention curve. Build a cross-functional initiative to address it — this is likely your highest-leverage retention project.
- [ ] 📋 PM: After every major feature launch, compare pre-launch and post-launch cohorts at the same lifecycle stage. Use cohort data as the definitive measure of feature success.
- [ ] 🎨 Designer: If Month 0-1 is the biggest drop, audit the first-time user experience. Map every screen a new user sees in their first session and identify where confusion or friction occurs.
- [ ] 🎨 Designer: Design lifecycle touchpoints (in-app messages, emails, milestone celebrations) that correspond to the retention inflection points identified by the PM.
Next: See Product Strategy for how to translate retention insights into product roadmap priorities.Previous: Unit Economics — ensuring each customer relationship is profitable before optimizing retention.