The leaking bucket
The metaphor is worn but exact. Acquisition pours water in; retention determines the size of the hole. Pour faster into a badly holed bucket and you get a wetter floor and the same water level — plus a larger bill.
The arithmetic is unforgiving. At five per cent monthly churn, the average customer stays about twenty months. At ten per cent, ten months — half the lifetime value from the same product, which means half the price you can afford to pay for a customer.
| 5% monthly churn | 10% monthly churn | |
|---|---|---|
| Average customer lifetime | ~20 months | ~10 months |
| LTV at ₹1,000/month | ₹20,000 | ₹10,000 |
| Affordable CAC at 3:1 | ₹6,600 | ₹3,300 |
| Customers to replace monthly at 500 users | 25 | 50 |
| Steady state at 30 new users/month | 600 users | 300 users |
Reading a cohort table
An overall churn number averages away the information. Cohorts — grouping users by when they joined — show whether the product is getting better, and where in the lifecycle people leave.
Joined Week 1 Week 2 Week 4 Week 8Mar 42% 28% 21% 20%Apr 45% 31% 24% 23%May 51% 39% 33% 32% <- onboarding change shippedJun 53% 41% 34% -- Reading down a column shows whether recent changes helped — May's improvement held into week eight, so it was real
- Reading across a row shows where in the lifecycle people leave — here, most of the loss is in week one
- A curve that flattens means you've found people for whom this is genuinely useful
- A curve still descending at week eight means nobody has settled, and acquisition is pouring into a hole
Most churn is decided in the first session
The largest drop is almost always between signup and first real use. Someone arrived, didn't reach the point where the product does its thing, and left. They aren't churned customers so much as people who never became customers.
- Define the moment the product becomes obviously usefulFor an itinerary tool, a completed itinerary they'd actually send. Name it precisely.
- Measure what share reach it in the first sessionThis single number is usually the highest-leverage metric in an early product.
- Remove every step before it that isn't requiredTeam invitations, profile setup, integrations, tours. All of it can come after.
- Solve the empty stateSeed a realistic example they can edit. Starting from a blank screen is where most first sessions end.
- Watch five people do it, in silenceYou'll find two structural problems in an hour that no funnel report would have named.
When acquisition is the right focus
This isn't an argument that retention always wins. There are real situations where acquisition is correct, and knowing them prevents the opposite error.
- Retention has flattened at a healthy level and the constraint is genuinely volume
- You have too few users to measure anything — ten users can't produce a cohort curve, so get to fifty
- The product is inherently low-frequency, where an annual usage pattern is normal and not a failure
- You're testing whether a new segment retains at all, which requires acquiring some of them first
Common mistakes
- Scaling acquisition while the cohort curve still descends to zero
- Measuring retention on logins rather than the valuable action
- Using one blended churn number instead of cohorts, which hides whether things are improving
- Treating first-session drop-off as marketing's problem rather than the product's
- Never calculating the steady-state ceiling implied by current numbers
Key takeaways
- Steady state is new users divided by churn — calculate it before planning growth
- Cohorts show both whether you're improving and where in the lifecycle people leave
- The biggest drop is usually signup to first real use
- Acquisition is right when retention has flattened, or when you have too few users to measure
Try it yourself
Calculate your steady state: monthly new users divided by monthly churn rate. That's the size your business settles at if nothing changes. For most early products the number is smaller than expected and immediately reorders priorities.