# The plateau: retention and product-market fit

> What a flattening retention curve says about the product, and what it doesn't.

- Course: [Product Analytics: From Data to Decisions](https://productovi.com/courses/product-analytics.md)
- Module: Do people come back?
- Lesson 15/32, ~28 min
- Source: https://productovi.com/courses/product-analytics/retention-plateau

## Does the curve stop falling?

**Alex Tutor:** Maya is back from an investor call with a question only Halves data can answer, so it comes to you.

**Maya:** An investor asked me straight out: “Do you have product-market fit?” I said yes, obviously. What in our data would actually back that up?

**Alex Tutor:** **Product-market fit** means a real market needs the product enough to keep using it without being pushed. One of its clearest signals is a **retention plateau**: the level where a cohort's curve stops falling.

**Alex Tutor:** Every curve drops at first: curious people leave. If it keeps sliding toward zero, nobody found lasting value. If it flattens, you've found people who need the product.

![Balls roll down a slope; many fall off the edge, but some come to rest on a flat ledge partway down.](https://productovi.com/courses/product-analytics/inline/retention-plateau-illustration-2.webp)

*If the curve flattens, some users have found lasting value.*

**Alex Tutor:** Before our data, a quick check that the shape is clear: the level a curve starts at matters less than where it ends up heading.

> **Exercise:** Two cohorts, followed for a year. Which curve is the **stronger** product-market fit signal?

**Alex Tutor:** Here are last summer's sign-ups, July to September 2025, by the type of their first group, followed for a year. Monthly bracket retention: opened the app at least once in month k.

**Summer 2025 sign-ups: share active by month since sign-up**

| Months since sign-up | Couple, opened the app | Flat, opened the app | Flat, any event | Trip, opened the app |
| --- | --- | --- | --- | --- |
| 0 | 100 | 100 | 100 | 100 |
| 1 | 72.2 | 65.9 | 68.9 | 33.5 |
| 2 | 64.5 | 54.9 | 59.9 | 12.4 |
| 3 | 60.3 | 48.9 | 55.9 | 8.4 |
| 4 | 57.3 | 45 | 53 | 7.4 |
| 5 | 55.8 | 43 | 52 | 5.6 |
| 6 | 55 | 41.6 | 51.6 | 6.1 |
| 7 | 55.3 | 40.6 | 50.6 | 7.2 |
| 8 | 53.8 | 40.9 | 50.9 | 7.6 |
| 9 | 53.8 | 40.4 | 51.4 | 10.3 |
| 10 | 53.7 | 40 | 51 | 13.9 |
| 11 | 52.5 | 39.6 | 50.6 | 19.4 |
| 12 | 52.7 | 39.4 | 50.4 | 21.8 |

*600 couples, 800 flats, 1,600 trips. “Flat, any event” also counts users whose only activity was a recurring bill created by the server.*

**Sofia:** Couples drop from 72.2% in month 1 to about 55% by month 6, and then barely move.

**Alex Tutor:** That's a plateau. Measure its level as the average of months 6 to 12, where the curve is flat.

> **Exercise:** From the chart, compute the **couples' plateau level**: the average of months 6–12 (opened the app). In %, one decimal.

## What a plateau doesn't prove

**Maya:** Great, 54% of couples stay for good. So that's product-market fit, done?

**Alex Tutor:** It's strong evidence that couples who try Halves find lasting value. It doesn't say how many couples there are to win, whether they'll pay for Plus, or whether newer cohorts will flatten too. One summer's cohort is one data point.

**Alex Tutor:** And size matters. A plateau at 50% of 20 hand-picked beta users says much less than one at 30% of thousands who came from the store.

> **Exercise:** The retention curve of Halves users who signed up in summer 2025 with a couple's group flattens at about 54% (months 6–12, app opens). Which of these does that plateau **not** show on its own? Select all that apply.

## A plateau built by the server

**Sofia:** Flats have two lines. By any event they level off around 51%, by app opens around 40%.

**Alex Tutor:** The gap is people who set up the rent as a recurring bill and never open the app again. The server keeps logging expenses for them, so an “any event” curve looks flatter and higher than real use.

![An empty flat with nobody home, where a small machine on the table keeps printing rent receipts on its own.](https://productovi.com/courses/product-analytics/inline/retention-plateau-illustration-3.webp)

*Rent added by the server can look like people who stayed.*

**Maya:** Automation is value too, though. They don't need to open the app.

**Alex Tutor:** Maybe, and that's worth checking: do they still settle up, or will they cancel when they move out? Until you know, report the curve by app opens and name the other one.

> **Exercise:** In **month 12**, how many percentage points higher is the flats' curve by **any event** than by **app opens**? One decimal.

**Alex Tutor:** So which plateau goes into the answer for flats? Compute it the same way as for couples, by app opens.

> **Exercise:** Compute the **flats' plateau level** by **app opens**: the average of months 6–12 on the chart. In %, one decimal.

## Trips end by design

**Maya:** And trips? 5.6% by month 5. That's no plateau at all. Should we stop marketing to travelers?

**Alex Tutor:** Look at the end of the line: from month 9 it climbs back to 21.8% by month 12, and most of those users are in a new group. A trip group ends when the trip ends. The group's lifetime isn't **user churn**; these people came back for next summer's trip.

**Sofia:** So for trips the question isn't “are they active every month” but “do they come back for the next trip”.

![A traveler pulls the same dusty suitcase out of a cupboard months later, with friends waiting at the door for a new trip.](https://productovi.com/courses/product-analytics/inline/retention-plateau-illustration-4.webp)

*Trip users don't vanish; many come back for the next trip.*

> **Exercise:** Halves users who signed up in summer 2025 with a trip group fall to 5.6% monthly activity by month 5, then rise to 21.8% by month 12, mostly in new groups. What's the best reading?

**Theo:** Can't we just show the investor one curve for everyone? Simpler.

**Alex Tutor:** Blended, the three types go from 29.6% in month 3 to 32.7% in month 12. It looks like retention improving, when it's trips returning for summer on top of two flat curves.

**Sofia:** So one curve would tell the investor retention grows, while couples and flats are slowly slipping.

> **Exercise:** **True or false:** the blended curve for all three types rising from 29.6% in month 3 to 32.7% in month 12 shows that Halves retention improved over the year.

**Theo:** Fine, no blend. But the investor still wants one picture. What do we put on the slide?

> **Exercise:** An investor asked Maya, the Halves product lead, whether Halves has product-market fit. Summer 2025 sign-ups split into couples, flats and trip groups, which use the app at different rhythms. Theo, the growth marketer, wants one simple picture for the investor. What should Maya show instead?

## The answer to the investor

**Maya:** Then help me answer properly: do we have product-market fit, and for whom?

**Alex Tutor:** Answer by segment, with the plateau levels, the measure you used, and what you still don't know.

**Sofia:** Couples and flats by app opens, trips by return for the next trip, and one line on the limits.

> **Exercise:** Write Maya's 3–5 sentence answer to the investor: does Halves have product-market fit, and for whom? Use the plateau levels, name the measure, and state one limit.

> **Key ideas**
>
> - A **retention plateau** is where a cohort's curve stops falling: the users who stay. A plateau at a meaningful level is a strong **product-market fit** signal.
> - Summer 2025 plateaus (months 6–12, app opens): couples 53.8%, flats 40.4%.
> - A plateau doesn't prove market size, willingness to pay, or that newer cohorts will behave the same. Small or hand-picked cohorts prove little.
> - Check the qualifying action: counting server-created bills lifts flats to 50.9%. Report real use and name the other measure.
> - Products that end by design (trips) fall and then come back: group lifetime isn't **user churn**. Judge them by the return for the next trip (21.8% active in month 12, mostly in new groups).
> - Don't blend segments with different rhythms into one curve.

**Alex Tutor:** “Obviously yes” wasn't wrong for couples. It just needed a curve, a measure and a “for whom”.

- Previous lesson: [Reading a cohort table](https://productovi.com/courses/product-analytics/cohort-table.md)
- Next lesson: [Case: retention analysis by segment](https://productovi.com/courses/product-analytics/retention-by-segment.md)

---
Productovi Learning. Free, no sign-up needed to read.