Skip to content
Primores AI wiki

Cohort Analysis — Reading the Shape, Not the Average

Cohort Analysis

TL;DR: Cohort analysis groups customers by acquisition date (or other shared attribute) and tracks them over time, instead of treating all customers as a single aggregate. The shape of the retention curve matters more than the Month-12 endpoint — specifically, the M0–M3 slope (the “onboarding cliff”) drives LTV more than any other input. Production-grade analysis uses surviving-cohort LTV (M3+) paired with NRR and CAC payback rather than a single aggregate LTV. 2026 benchmarks: B2B SaaS LTV:CAC median 3.2:1; DTC subscription 4.1:1 (replenishment categories reached SaaS parity in 2026 for the first time). Once a DTC customer buys twice, typical retention is 85–90% from that point forward. First-to-second purchase rate is the central operational lever.

Simple explanation

Suppose a company says “our customer LTV is $400 and CAC is $100, so LTV:CAC = 4:1. We’re healthy.”

That might be true. Or it might be hiding three different segments: enterprise customers with $2,000 LTV, SMB customers with $200 LTV, and consumers with $50 LTV — all averaged into “$400.” The aggregate looks healthy; two of the three segments are individually broken.

Cohort analysis fixes this by grouping customers and tracking each group over time. The basic unit: customers acquired in the same month (the “cohort”) tracked across subsequent months. You see retention, revenue, and contribution per cohort, separately.

Once you cohort the data, you see things that aggregate hides:

  • Which acquisition channels produce high-LTV cohorts vs. low-LTV ones
  • Whether retention is improving or declining over time
  • The shape of the retention curve — gentle slope vs. cliff
  • Which months / years acquired the “good” customers and which didn’t

This is the difference between “is the business making money?” (aggregate) and “is each cohort making money?” (cohort).

Why it matters for business

Three operational consequences:

  • Acquisition decisions get more honest. A channel showing “great CAC” in aggregate might be acquiring low-retention customers whose LTV doesn’t cover the CAC at the cohort level. Cohort analysis reveals this; aggregate hides it.
  • Retention investments get prioritized. If the M0–M3 onboarding cliff is what’s costing you LTV, fixing onboarding is higher-leverage than acquiring more customers. Cohort analysis shows you where the leak is.
  • Capital efficiency becomes legible. Investors and operators alike now expect cohort-level disclosure. Aggregate-LTV pitch decks read as evasive in 2026.

The business framing: cohort analysis is what tells you whether the business model is working. MMM and incrementality testing tell you whether the marketing engine is working — cohort/LTV tells you whether the engine is worth scaling at all.

The shape of the retention curve

The single most underappreciated finding in 2026 cohort literature:

The shape of the retention curve — not just the M12 endpoint — drives LTV more than any other input.

Two cohorts that hit the same Month-12 retention can have radically different LTVs. A cohort that retains 65% through a steep M0–M3 onboarding cliff is much less valuable than a cohort that gently slopes to the same 65% endpoint. The area under the curve is what generates LTV; the shape determines the area.

The M0–M3 “onboarding cliff” is the highest-leverage segment of the curve. Most customer churn happens in the first 90 days. If you can flatten the slope here, every later segment compounds.

SaaS contractual retention (2026 benchmarks):

  • Month 12: bottoms at 71%
  • Month 24: flattens to 64%
  • The flattening is the asset — once a cohort survives M12, the long-tail revenue is durable

DTC transactional retention (2026 benchmarks):

  • The shape is different — it’s about repeat purchase, not contractual renewal
  • Once a customer buys twice, retention is typically 85–90% from that point forward
  • First-to-second purchase rate is the central lever:
    • Supplements / consumables: 30–40%
    • Beauty: 25–35%
    • Fashion: 15–25%

The first-to-second conversion rate sets the ceiling on DTC LTV more than any post-purchase intervention. Improving onboarding to get the second purchase is structurally higher-leverage than improving retention among 3+-time buyers.

2026 LTV:CAC benchmarks

VerticalMedian LTV:CACTop quartileNotes
B2B SaaS3.2:14:1 – 6:1The 3:1 rule still holds as the baseline
DTC subscription4.1:1—Reached SaaS parity in 2026 for the first time (driven by 102% NRR + stable CAC)
DTC transactionalHighly variable—Depends on first-to-second purchase rate (15–40% by category)
Cross-industry median3.4:15.6:1The gap between median and top quartile has widened every year since 2023

The “3:1 rule” remains the rough benchmark — below 3:1 the business is paying too much to acquire customers; above 5:1 you may be underinvesting in growth. But the operational reality is more nuanced.

Surviving-cohort LTV > aggregate LTV

The 2026 production pattern: track LTV across multiple slices, not as a single number.

Surviving-cohort LTV (M3+) is the LTV calculated only for customers who survived past Month 3. This separates the customers who churned in the onboarding cliff (whose LTV is essentially their first purchase) from the customers who are actually engaging with the product (whose LTV is the realized long-tail value).

The argument for the M3+ cut: customers who haven’t survived the onboarding cliff are different as customers from those who have. Mixing them produces a misleading average. Reporting both — aggregate and M3+ — gives an honest picture.

Paired with surviving-cohort LTV:

  • NRR (Net Revenue Retention) — how much revenue from a cohort is left after a year, including expansion (upsells, cross-sells) and contractions (downgrades, partial churn). NRR >100% means cohorts grow over time.
  • CAC payback period — how many months until a cohort’s contribution profit recovers the CAC. Short payback (<12 months) means the business is capital-efficient; long payback (>24 months) means it’s slow.

Together, surviving-cohort LTV + NRR + CAC payback = a more honest business-health signal than any single LTV figure.

LTV:CAC failure modes the dashboard hides

The aggregate dashboard often hides these:

  • Mixed segments averaging into “healthy.” Enterprise + SMB cohorts averaged together can produce a 3:1 LTV:CAC where enterprise is 6:1 and SMB is 1.5:1 (and SMB is dragging the business). Segment-level analysis reveals this.
  • Survivorship bias in improving LTV. “Our LTV is going up” might be a real improvement, or might just be older cohorts (which have had more time to generate revenue) dominating the data. Vintage-level analysis controls for this.
  • CAC that excludes failed-channel spend. If you report CAC as “successful-campaign-spend ÷ customers acquired,” you’re hiding the cost of the failures. Total marketing spend ÷ customers is the honest denominator.
  • Payback period exceeding runway. A business with 18 months of cash and a 24-month CAC payback runs out of money before the unit economics work. Payback period vs. runway is the brutal check.
  • Acquisition-mix shifts that mask cohort-quality changes. Adding paid channels can inflate aggregate CAC even if cohort-level retention is improving (or vice versa). Channel-attributed cohort analysis catches this.

Connection to wiki frameworks

Honest limits

  • Cohort analysis requires cohort-level data infrastructure. Aggregate LTV is easy; cohort-level LTV requires data warehouse work most marketing teams haven’t done.
  • Small cohorts produce noisy curves. Cohort analysis with <500 customers per cohort is statistically weak. Aggregation by quarter or year may be necessary.
  • Cohort definitions are choices. Acquisition date is the most common, but cohort by channel, by geography, by product, by campaign can each surface different patterns. No single cohort cut is “right.”
  • Retention curves get smoother (look better) over time because failures drop out. This isn’t always “things are improving”; it’s sometimes just survivorship bias. Always check the cohort count alongside the retention rate.
  • The 85–90% post-second-purchase retention number is a rough rule-of-thumb, not a precise benchmark. Specific verticals (luxury, low-frequency replenishment, B2B with long buy cycles) deviate substantially.

Key Takeaways

  • Cohort analysis groups customers and tracks them over time instead of treating all customers as a single aggregate. The aggregate hides what cohorts reveal.
  • The shape of the retention curve > the M12 endpoint. Two cohorts with the same Month-12 retention can have radically different LTVs depending on the M0–M3 onboarding cliff.
  • Once a DTC customer buys twice, retention is typically 85–90% from that point forward. First-to-second purchase rate is the central lever.
  • First-to-second purchase rates by vertical: supplements/consumables 30–40%, beauty 25–35%, fashion 15–25%.
  • SaaS contractual retention bottoms at 71% Month 12, flattens to 64% Month 24. The flattening is the asset.
  • 2026 LTV:CAC benchmarks: B2B SaaS 3.2:1 median (top quartile 4:1–6:1); DTC subscription 4.1:1 (parity with SaaS reached in 2026); cross-industry median 3.4 / top quartile 5.6 (gap widening since 2023).
  • Surviving-cohort LTV (M3+) + NRR + CAC payback > single aggregate LTV. The trio gives an honest business-health signal.
  • The dashboard hides mixed-segment averaging, survivorship bias, failed-channel CAC, payback period exceeding runway, acquisition-mix shifts masking cohort-quality changes.

Sources