Retention and cohorts: why keeping customers beats winning them
How a customer base settles at new customers divided by churn, how to read cohort curves, and why retention usually beats acquisition.
Key takeaways
- A business that wins the same number of customers each month and loses a steady share settles at a size of new customers divided by the monthly churn rate.
- Common mistakes: Reading one company-wide churn figure: it mixes loyal old customers with new ones who leave fast, and hides whether things are getting better or worse.
- You met churn (the share of customers who leave each period) and customer lifetime value in Business basics.
Key idea
A business that wins the same number of customers each month and loses a steady share settles at a size of new customers divided by the monthly churn rate. Halving churn doubles that size; doubling acquisition only does the same at twice the marketing cost. That is why retention usually beats acquisition.
You met churn (the share of customers who leave each period) and customer lifetime value in Business basics. Here is the bigger picture. Think of a customer base as a bucket with a hole in it. Marketing pours water in; churn lets it out. When the water flowing in equals the water flowing out, the level stops rising. That level is new customers per month divided by monthly churn.
Harvard Business Review (2014) reports that winning a new customer costs 5 to 25 times as much as keeping one, and cites Bain & Company research that raising retention by 5 percent raises profits by 25 to 95 percent. These are broad findings, not measurements of your client. Use them as a reason to ask about retention, never as a number to put in your answer.
Cohorts: following one group over time
A cohort is a group of customers who started in the same month. A cohort table shows what share of each group is still buying after 1, 3, 6 and 12 months. Reading across a row shows how one group fades. Reading down a column compares groups at the same age: if newer cohorts keep fewer customers, growth in sign-ups may be hiding a business that is getting worse.
| Cohort (month joined) | After 1 month | After 3 months | After 6 months | After 12 months |
|---|---|---|---|---|
| January 2025 | 70 | 50 | 40 | 35 |
| April 2025 | 68 | 47 | 36 | 31 |
| July 2025 | 62 | 41 | 30 | Not yet |
| October 2025 | 58 | 37 | Not yet | Not yet |
So-what
Each newer cohort keeps fewer customers at the same age: 50 percent after 3 months for January but 37 percent for October. The company is buying worse customers, perhaps with bigger first-order discounts.
Line chart: The January 2025 cohort over its first year: the curve flattens. Values in percent still ordering. Month 0: 100; Month 1: 70; Month 2: 58; Month 3: 50; Month 6: 40; Month 9: 37; Month 12: 35.
So-what
Most losses happen in the first three months; after that the curve flattens near 35 percent. A curve that flattens means a loyal core exists. A curve that keeps falling towards zero means the product has not found one.
Worked case
Cut churn or buy more customers?
The prompt
The fictional meal kit company wins 10,000 new customers a month and loses 8 percent of its customers each month. Each customer brings SGD 36 of contribution a month. Option A: a better first month (free menu changes, a call from a chef) cuts churn to 6 percent and costs SGD 200,000 a month. Option B: more advertising raises new customers by 25 percent at a cost of SGD 150 for each extra customer. Which option earns more once the customer base settles? (Assume churn stays steady.)
The structure
- Settled customer base = new customers per month / monthly churn
- Base today and under each option
- Monthly contribution = base x SGD 36
- Key: Minus the cost of each option
Working it through
1. Base today
10,000 new a month divided by 8 percent churn.
Settled base today (customers):10,000 ÷ 0.08 = 125,0002. Contribution today
125,000 customers at SGD 36.
Monthly contribution today (SGD):125,000 × 36 = 4,500,0003. Option A base
10,000 new a month at 6 percent churn.
Settled base with option A (customers):10,000 ÷ 0.06 = 166,6674. Option A result
Contribution minus SGD 200,000 a month.
Option A monthly result (SGD):10,000 ÷ 0.06 × 36 - 200,000 = 5,800,0005. Option B base
12,500 new a month at 8 percent churn.
Settled base with option B (customers):12,500 ÷ 0.08 = 156,2506. Option B cost
2,500 extra customers a month at SGD 150 each.
Option B monthly cost (SGD):2,500 × 150 = 375,0007. Option B result
Contribution minus the extra advertising.
Option B monthly result (SGD):156,250 × 36 - 375,000 = 5,250,000
The recommendation
The company should cut churn first: option A settles at about SGD 5.8 million a month after its cost, against SGD 5.25 million for option B and SGD 4.5 million today. First, cutting churn from 8 to 6 percent grows the settled base by a third, to about 166,700 customers, with no extra advertising. Second, the extra customers in option B cost SGD 375,000 a month, and the cohort table suggests newer customers are already leaving faster, so the real churn on them may be worse than 8 percent. The risk is that the first-month fixes do not cut churn as much as hoped. As a next step, test them on one month's cohort and compare its 3-month retention with the cohorts before.
Risks: The settled base takes many months to reach; the gap opens slowly; If extra advertising brings lower quality customers, option B is worse than shown.
A mobile operator in Malaysia wins 50,000 new prepaid customers a month and loses 2.5 percent of its customers each month. Around what number of customers does its base settle?
Reading one company-wide churn figure: it mixes loyal old customers with new ones who leave fast, and hides whether things are getting better or worse. Celebrating sign-up growth while each new cohort keeps fewer customers. Assuming retention is free: it has costs too, so compare the cost of each option with the customers it keeps. Treating 1 divided by churn as exact: real curves fall fast then flatten.
A company wins 1,000 customers a month and loses 5 percent a month. Where does its base settle?
Reading down a cohort table, each newer cohort keeps fewer customers after 3 months. What does that suggest?
Sources for this lesson (2)
- Recognized public explanations of case-interview concepts and terms
- Harvard Business Review, "The Value of Keeping the Right Customers", October 2014
My notes on this lesson
0 of 5,000 characters. Saves automatically.
Try the 3 remaining checks and drills above to complete this lesson (0 of 3 done).
Spotted something wrong or out of date? Report a mistake. We check every report and correct the page.