ExeechainExeechain

Detect

  • Churn prediction0-100 daily, with the drivers
  • Customer health scoreThe three reasons, quoted
  • Revenue forecastNext quarter, defensible

Act

  • Retention playbooksDrafted per customer
  • Workflow automationWorkflows you compose
  • CopilotAsk about any account

Prove

  • QBR automationEight hours to thirty seconds
  • Revenue leak benchmarkSize the leak from your MRR, free
  • Retention intelligenceThe whole loop

Alternatives

  • vs Gainsight
  • vs ChurnZero
  • vs Vitally
  • vs Planhat
  • vs Totango

What they cost

  • Gainsight pricingEvery reported figure, sourced
  • ChurnZero pricingListed vs actually paid
  • Buyer's guideHow to choose

Connect

  • Stripe
  • HubSpot
  • Intercom
  • Zendesk
  • Mixpanel
  • Slack

Bring your own

  • API & CSVREST, webhooks, import
Pricing

Read

  • Documentation
  • Blog
  • Getting started

Look around

  • Live demoA seeded workspace
  • Churn calculatorLogo and revenue churn, annualised
  • NRR calculatorNet and gross retention
  • SecurityWhat we have, and don't
Sign inStart free
  1. Home/
  2. Blog/
  3. SaaS retention curve
Blog/Retention metrics

SaaS retention curve: how to build the cohort table, the three shapes it takes, and what to do about each

How to build a SaaS retention curve from billing data (cohorts from first payment, one row per acquisition month), the three shapes (drop then flat, steady decline, cliff at renewal) and what each means, the customer curve against the revenue curve, and the four mistakes that produce a false curve.

Exeechain Research·September 20, 2026·8 min read

Short answer

What should a SaaS retention curve look like?

Drop, then flat. A healthy curve loses the customers who never reached value in months one to three, then levels into a plateau that barely moves. The plateau height is the number that matters: above 70% to 80% for B2B is strong. A curve that keeps sloping is a product being rented rather than adopted; a curve with a cliff at month twelve is a renewal problem, not an onboarding one.

A churn rate is one number about one month. A retention curve is the whole story of a cohort, and it answers the question the churn rate cannot: when do customers leave, and do the ones who stay ever stop leaving? This is how to build the curve from billing data, the three shapes it takes and what each one means, the four mistakes that produce a false curve, and what to do about each shape.

Building the cohort table

The retention curve is the average of a cohort table, so build the table first. Rows are acquisition months. Columns are months since acquisition. Each cell is the share of the row's cohort still active at that column.

Example cohort retention table, share of each acquisition cohort still paying by month
CohortSizeM1M2M3M6M12
Jan12091%85%82%80%78%
Feb13590%84%81%79%
Mar11088%81%78%77%
Apr14292%87%84%
Average90%84%81%79%78%

The illustrative table above is the healthy shape: a ten-point drop in month one, a further six by month two, three more by month three, and then almost nothing. The average row is the retention curve. Read the diagonal too: it is the most recent month for every cohort, and a diagonal that is worse than the column above it means something changed recently for everyone at once, which is usually a price change, an outage, or a competitor.

Three rules for the cells. Start the cohort at first payment, not signup. Define active as paying, from the billing system, unless you are deliberately measuring usage. And do not fill a cell whose month has not fully elapsed; the blanks in the table are honest, a partial month is not. The SQL for this table, with the CTEs, is in the cohort analysis guide.

The three shapes

Drop, then flat

The healthy curve. Early-life churn removes the customers who never activated, and the survivors stay. Two numbers describe it: the depth of the drop (how many never reached value) and the height of the plateau (how many the product holds). The plateau is worth more than the drop. A curve that falls to 60% and stays there for three years is a better business than one that falls to 80% and keeps sliding.

Steady decline

The same share of remaining customers leaves every month, at month three and at month thirty. Tenure is not building attachment. This is the shape of products used for one project, of customers who were the wrong fit from the start, and of pricing that charges before the habit forms. It is also the shape that makes lifetime value a fiction, because LTV projections assume a plateau exists.

Flat, then a cliff

Retention holds for eleven months and drops at twelve or thirteen. That is the annual renewal, and the drop is customers who were auto-renewed once and not twice. These customers stayed a full year; onboarding did its job. The fix is a renewal process that starts ninety days before the date, not more onboarding.

Customer curve and revenue curve

Build the same table with MRR instead of account counts and you get the revenue retention curve. On most SaaS books it sits above the customer curve, because the small accounts churn first and the survivors expand. A revenue curve that climbs above 100% is a cohort whose expansion outgrew its churn, which is net revenue retention over 100% measured the right way, on a cohort rather than a blended month. A revenue curve that sits below the customer curve means the large accounts are the ones leaving, and that is a different emergency from the one the customer curve shows.

Four ways to build a false curve

  1. Cohorts that start at signup. The month-one drop becomes the trial conversion rate wearing a retention costume. Start at first payment.
  2. Partial months filled in. A cohort from three weeks ago has no month-one figure yet. Filling it with the customers who have not left so far overstates the whole diagonal.
  3. Failed payments counted as churn. A card that expired is a customer who did not decide to leave. If a third of the drop is billing failures, a third of the curve is a payment recovery problem with a known fix. Split the involuntary share out or the curve will send the product team to fix a billing problem.
  4. Reactivations counted as retained. A customer who churned in month four and came back in month nine appears active at month twelve. That is a win-back, and it belongs on its own line, not in the retention curve.

What to do with each shape

  • Deep drop, healthy plateau: the product holds the right customers and loses the wrong ones early. Work on time to first value and on qualifying signups, and measure the month-one figure by acquisition channel, because one channel is usually producing most of the drop.
  • Steady decline: find the customers still there at month eighteen and learn what they use that the leavers did not. That feature or workflow is the product; the rest is what people rent. Then make it the onboarding.
  • Cliff at renewal: build the renewal pipeline, and read the health of every annual account at ninety days out. The customers who leave at the anniversary usually looked fine on logins and terrible on breadth of use.

Frequently asked questions

What is a SaaS retention curve?

A SaaS retention curve plots the share of a signup cohort still paying at month 1, 2, 3 and onward. Each acquisition month is its own line, or the lines are averaged into one. The height of the curve at any month is cohort retention at that month; the shape of the curve tells you when customers leave and whether the ones who stay, stay for good.

What does a good SaaS retention curve look like?

A good SaaS retention curve drops in the first one to three months, then flattens into a near-horizontal line. The early drop is customers who never reached value; the flat part is the customers the product actually holds. For B2B SaaS, a curve that flattens above 70% to 80% is strong; for self-serve and SMB products the plateau is lower and the first-month drop steeper. The plateau matters more than the starting slope.

How do you build a cohort retention table?

One row per acquisition month, one column per month since acquisition, each cell holding the share of that cohort still active. Start the cohort at first payment, not at signup, or the trial conversion rate blends into the curve. Define active as paying (billing) rather than logged in (product), unless you are deliberately measuring usage retention. The retention curve is the average down each column, and the diagonal of the table is the most recent month for every cohort.

Why does my retention curve keep falling instead of flattening?

A curve that declines at a steady rate every month means the same share of remaining customers leaves each period regardless of tenure, which means tenure is not building any attachment. Common causes are a product used for a single project rather than an ongoing process, a customer base that was never the right fit, or a pricing model that charges before the habit forms. It is the shape that makes lifetime-value projections fiction, because there is no plateau to project from.

What causes a cliff in the retention curve at month 12 or 13?

The annual renewal. A curve that holds and then drops at a fixed month is customers who were auto-renewed once and chose not to renew twice, or who reviewed the contract at its first anniversary and left. The fix is a renewal process that starts ninety days before the date, not onboarding, because those customers stayed for a year.

Should the retention curve use customers or revenue?

Build both. The customer curve tells you when accounts leave; the revenue curve tells you whether the accounts that leave are the ones that matter. In most SaaS books the revenue curve sits above the customer curve, because small accounts churn first and expansion lifts the survivors. A revenue curve that ends above 100% is net revenue retention over 100% on that cohort, which is what a board wants to see.

Where this comes from

The cohort table is illustrative, not a customer's data. The benchmark plateau (70% to 80% for B2B) is the widely quoted industry range, not a measurement Exeechain took. The four false-curve mistakes are the ones Exeechain's own churn accounting is built to avoid: involuntary churn is recorded in its own lane, and a reactivation is recorded as a win-back rather than as retention. The formula side of retention is in customer retention rate.

Evaluating cohort retention against other platforms? See how Exeechain compares head-to-head with Gainsight, ChurnZero, Vitally, and Planhat.

Keep reading

More from the blog.

Customer success

Renewal management: the 90-day process that stops silent churn, with stages, exit criteria and four KPIs

9 min read · Sep 20, 2026

Customer success

SaaS onboarding metrics: the seven that predict retention, defined, and the day-14 rule

8 min read · Sep 20, 2026

Churn prediction

Churn management: the definition, the four types of churn, the loop, and the strategies in order of return

9 min read · Sep 20, 2026

The curve, from your billing data,
with the failed payments split out.

First scores in 15 minutes. Full accuracy in 24 hours. From $299/mo. Cohorts from first payment, involuntary churn in its own lane, the save drafted for approval.

Or look at the live demo

Not ready to switch? Size your leak from your MRR and churn rate, with no billing access at all.

ExeechainExeechain

Find the revenue you never decided to lose. Failed payments, cards expiring before renewal, and renewals slipping, by customer, with the dollars attached.

Product

  • Free revenue leak scan
  • Churn prediction
  • Health scores
  • Copilot
  • Playbooks
  • QBR automation
  • Forecast
  • Retention intelligence
  • Pricing

Integrations

  • Stripe
  • HubSpot
  • Intercom
  • Zendesk
  • Mixpanel
  • Slack

Free tools

  • Churn rate calculator
  • NRR calculator

Compare

  • vs Gainsight
  • vs ChurnZero
  • vs Vitally
  • vs Planhat
  • vs Totango
  • Gainsight pricing
  • ChurnZero pricing
  • Planhat pricing
  • Vitally pricing
  • Totango pricing
  • CS software pricing
  • Buyer's guide

Resources

  • Live demo
  • Blog
  • Failed payment recovery
  • Involuntary churn
  • API reference

Company

  • About
  • Contact
  • Status

Legal

  • Privacy
  • Terms
  • Security
AES-256 at restTLS 1.2+DPA available
© 2026 Exeechain·Built for SaaS teams who take retention seriously.
Sign in