Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Growth Marketing › Intermediate

Reading Retention Cohorts for a Project Management Tool

An exercise in reading a cohort table the way a growth lead would before anyone blames the product or the price. You read rows, columns and a diagonal, separate a promotion's cliff from a real change in onboarding, check the acquisition mix before crediting the product, and turn what you find into retention experiments with holdouts and fixed cohort ages.
ExerciseIntermediate levelModule 2: Cohorts and retention5 minUpdated

Why this matters

A blended retention rate mixes customers of every age, so it can hold steady while newer customers leave faster, or dip for a reason that has nothing to do with the product. The cohort analysis tutorial builds the table that separates customers by age and shows the three readings it supports. The guide to churn and retention metrics sets the definitions that make one cohort comparable with another. A misread table caps LTV at the wrong level and sends the budget the wrong way.

The tool and its cohorts

Taskloom is an invented project management tool for small agencies and studios, with a free plan and paid plans priced per seat. Sign-ups have risen all year, yet monthly recurring revenue has barely moved. In March a webinar partnership offered new paying teams their first three months at half price. In April Taskloom shipped a new onboarding flow: a setup checklist and template projects that prompt the owner to invite teammates. A team is "retained" while it is still paying; each cohort is the month of a team's first payment.

Illustrative example — not a real result

Taskloom's paying-team cohorts, data to the end of June

CohortTeamsM0M1M2M3M4M5One-seat teams at first payment
January220100%78%66%60%57%56%45%
February240100%76%65%59%57%—47%
March410100%80%72%41%——70%
April260100%84%75%———44%
May270100%85%————46%
June280100%—————45%
MonthJanFebMarAprMayJun
Free sign-ups1,8001,9502,9002,3002,5002,700
Monthly recurring revenue£38,200£39,100£40,300£41,800£42,400£40,900
Blended monthly retention on the dashboard—92%93%92%93%86%

Retention at M2 for the January and February cohorts combined, by seats at first payment: one seat 52%, two or more seats 79%.

Reading the curve

Read the triangle in this order, annotating before you interpret.

  • Along a row. Where does each cohort's curve drop, and does it flatten?
  • Down a column. At the same age, are newer cohorts doing better or worse, and has the mix changed?
  • Across a diagonal. Did one calendar month move every cohort at once, or only one?
  • By segment. Is the curve one population or an average of two?

Task

  1. Describe the shape of the January and February curves: where they drop, where they flatten and what that implies for projecting lifetime value.
  2. Explain the March row and the June dashboard figure, using the diagonal to decide whether something happened to every customer in June.
  3. Judge whether the April onboarding change improved retention, checking the mix before crediting the product.
  4. Explain why sign-ups rose while revenue stayed flat.
  5. Define a retention KPI that a promotion cannot move without changing behaviour.
  6. Propose two retention experiments, each with a primary metric, a holdout and the cohort ages at which it is read.

Worked answer

The early curves flatten. January and February lose most teams in the first month and less each month after, levelling off in the high fifties from age three onwards. That is a flattening curve with a retained core, so lifetime value can be projected from the plateau for these cohorts, labelled as a projection.

March is a cliff, not a collapse. The March row looks strong at ages one and two because those teams paid half price, then falls hard at age three, the month the price doubled. The June diagonal confirms it: the January and February cohorts hold their plateau, so nothing hit every customer in June. The dashboard's fall is March's cliff averaged into a blend.

Illustrative example — not a real result

The onboarding comparison and the mix check

AgeJanuary and FebruaryApril and MayDifference
M178% and 76%84% and 85%About 7 points higher
M266% and 65%75% (April only)About 9 points higher
One-seat share45% and 47%44% and 46%Unchanged

March is excluded from the comparison: its discount and mix make it a different population.

The onboarding change probably worked. Newer cohorts retain better at the same ages, and the one-seat share is unchanged, so a shift in mix does not explain it. The evidence is still young: two cohorts, read no later than age two. Treat it as provisional until April reaches age three.

Why revenue is flat. The extra sign-ups came largely from the March promotion, whose teams paid half price for three months, were mostly one seat, and then left in large numbers at full price. Revenue churned out as fast as promotion revenue came in.

The KPI. Age-three retention for the latest complete cohort, with promotional cohorts annotated and reported separately. A blended rate rises with acquisition and falls with a cliff; a fixed cell moves only when behaviour does.

Experiment one: the price step. For future promotional cohorts, test a transition programme: a reminder before the step, an offer to move to annual billing and a smaller plan offered in place of cancelling. Hold out a share of the cohort on the current experience. Primary metric: still paying in the month after the step, with a later age read for durability and revenue per team as a guardrail.

Experiment two: the second seat. One-seat teams retain far worse. Test an in-product prompt and invitation template in the first week for one-seat teams, with a persistent holdout. First validate the leading metric on past cohorts: did teams that added a seat after paying, not only teams that started with several, retain better? Read at ages two and three, since a gap that shrinks between the two is novelty, as the experimentation loop for retention explains.

Common mistakes

  • Trusting the blended rate. It hid the March cliff until it arrived, then made it look like a company-wide fall.
  • Crediting onboarding without checking the mix. A falling one-seat share would have produced the same gain.
  • Running a short split for a retention change. Call the long comparison a holdout; reserve A/B Test for short tests inside the programme.

Next, take the cohorts and retention checkpoint or return to the intermediate growth level.

Pages this lesson applies

Terms used on this page

Also related

Related