Why users stop coming back
Signups were growing. Retention after the first purchase was not. The brief called it a retention problem. The first thing I did was argue with the brief.
Written for Daleel Store at the opening of the engagement, before any access to their internal data, so every claim below is framed as a hypothesis with a way to test it, not a finding. Originally delivered in Arabic; this is my own translation, tightened. Names and commercial details of the people involved are omitted.
01The reframe: the Finished/Churned Split
Churn is a user who had a reason to come back and didn't. "Finished" is a user whose job is done. They look identical in a retention chart. They are opposite problems.
This is the whole argument. If you treat a finished user as churned, you build re-engagement machinery for people who were never coming back, and you make the product heavier for everyone who was. If you treat a churned user as finished, you accept a leak as a law of nature.
So before any initiative: split the base behaviourally into single-purpose usage and plausible repeat usage, and only optimise against the second.
02Three diagnoses that all look like "retention is down"
| Diagnosis | What happened | What it demands |
|---|---|---|
| Churn | The user reached core value and had a clear reason to return, and didn't. | Fix the return path. This is the only one re-engagement work belongs to. |
| Low engagement | The user never reached core value in the first place, or not enough of it to judge. | Fix activation and time-to-first-value. Reminders here just annoy someone who never got the point. |
| Weak PMF | The product doesn't solve a recurring or high-priority problem at its current price. | Fix positioning, pricing, or segment. No amount of UX work moves this. |
Naming which of the three you're in is the highest-leverage half-hour in the whole quarter. Most teams skip it and spend the quarter anyway.
03Hypotheses, written before looking at data
Writing them first is a discipline, not a formality: it stops the data from being read as confirmation of whatever you already wanted to build.
- Post-sale support and customer service are weak enough to kill the second purchase.
- The perceived value is available cheaper or better from a competitor.
- Core value doesn't surface clearly, or doesn't surface at the right moment after first use.
- The first session closes the user's goal but opens no next scenario.
- The user doesn't fail. They finish. From where they stand, the story ended well.
That last one is the uncomfortable hypothesis, which is exactly why it goes on the list.
04Three metrics, and no more
- Activation rate. Did the user reach core value?
- Time to First Value (TTFV). How much time and effort before the product feels useful?
- Repeat usage / re-engagement. Does a natural return pattern exist at all?
Alongside those, three things to trace: the last step completed before the drop, the gap between first use and any return attempt, and what returners did differently in their first session.
The goal isn't to collect more data. It's to locate the moment of separation.
05The journey, and where it breaks
Signup → browse → purchase → first real use → first value → the post-value moment → second and third return attempts, if any.
Every break I'd expect sits in that sixth step, the one most journey maps don't draw:
- Immediately after the first goal is met.
- When there is no clear next step: no behaviour-triggered nudge, no timed prompt, nothing.
- When the experience feels like it ended successfully. Success with no sequel is still an exit.
Why doesn't the user come back, given the product worked? Because they see no logical repeat use, or they don't believe a return adds new value, or the product never tied this use to a future context.
06Three initiatives
| Initiative | Problem | The move | Measured by |
|---|---|---|---|
| 01 · First-Value redesign | Value arrives late, once, then disappears. | Surface core value earlier and tie it explicitly to a later scenario. Quick-buy in the Daleel Store app is the obvious first candidate. | TTFV falls |
| 02 · Post-purchase simplification | Unnecessary friction between paying and using. | Connect purchase directly to use with no intermediate steps. In a category like game top-ups this is expensive to build and to market. It also buys a customer who trusts the store with their account, which in that market is the whole purchase decision. | Share completing first use; repeat purchase of the same service |
| 03 · A real reason to return | No reminder, and no natural occasion. | Make "when and why you should come back" explicit. Behaviour-triggered prompts, or timing tied to the user's capacity to buy. For salaried users that means a monthly prompt landing a day or two before payday, not on a marketing calendar. | Repeat usage; re-engagement within 60 days |
Note what these have in common: light interventions tied to a specific value, not a generic "we miss you." A nudge that names nothing is a nudge that teaches users to ignore you.
07What I'd defer, and why
- Loyalty programmes and subscription changes. Not before core value is stable. Layering a rewards scheme on an experience people don't return to just makes churn more expensive.
- Secondary polish that doesn't touch the main drop point. It always feels productive. It moves nothing.
The sequencing rule I'd apply: prioritise anything that touches the first experience, needs no deep technical change, and has a legible commercial line to retention. Cheap tests, reversible decisions, scale only after impact is proven.
0890 days
| Window | Work |
|---|---|
| Days 0–30 | Diagnose. Hypotheses, journey analysis, and the Finished/Churned split made real in the data. |
| Days 30–60 | Test. First-use and post-purchase experiments, small and reversible. |
| Days 60–90 | Scale what moved, and measure the commercial impact rather than the engagement metric. |
And in week one, before any of it: share the problem with Engineering, Data, Marketing and Customer Experience, say plainly what we're chasing, which is the return, and listen. Data defines the metrics and instruments the behaviour. Marketing stops promising what the product doesn't deliver and starts targeting lookalikes of people who did come back. CX brings the qualitative half: why people say they didn't return.
09What I left out on purpose
A plan is also a list of things you decided not to do. These were cut for priority or for lack of information, not because they're bad:
- Customer service and support channels. Real, but a different project.
- Points and loyalty systems. The common failure is a currency whose value nobody can compute. "You have 340 points" means nothing. "This covers your next delivery" means something.
- Need-based bundles. For Daleel Store, something like a home-and-school essentials bundle with a games voucher included as the gift. Parents carry real guilt about discretionary spend, and a bundle absorbs it.
- User involvement through surveys or user-generated content incentivised with free coupons, as long as you listen for the need behind the answer. Users are reliable about their problems and unreliable about their solutions.
- More payment options, including instalments, for segments the current set excludes.
The plan was built inside a general hypothetical frame. Change the assumptions and the diagnosis changes with them. That's not a hedge. It's the reason the first 30 days go on proving which of the three diagnoses is true before anyone writes a ticket.
A retention chart can't tell you the difference between a user you lost and a user you finished serving. That distinction is the entire job.