Why Users Do Not Come Back: Early Retention When You Have Twenty of Them
At twenty users churn is not a metric, it is twenty stories you can read individually. The most common cause is not a missing feature, it is that the problem does not happen often enough to build a habit around.
Posted by
RouzbehRelated reading
The First Five Minutes: Onboarding When You Have No Onboarding Team
Most onboarding advice assumes a product team and a tooling budget. At ten users the fix is not a guided tour, it is the empty state, one obvious first action, and doing it by hand while you still can.
Using Directories for SEO: Getting Listed, and Building Your Own
Two different tactics get called the same thing. Submitting to directories is a small one-off. Running one is a content asset that compounds, and it only works if the pages say something nobody else has.
MCP or RAG: Which One Does Your AI Feature Actually Need?
MCP connects a model to live data. RAG compresses a corpus that will not fit in a prompt. They solve different problems, and picking the wrong one produces confident wrong answers.
Someone signed up, used the product properly, and never opened it again. That is the most demoralising result in software, and it is also the most informative, because it means everything before the product worked. The instinct is to add features. The more common cause is that the problem you solve does not happen often enough for anyone to build a habit around it, and no feature fixes that.
Table of contents
Retention is a property of the problem
Before diagnosing anything, answer one question honestly: how often does the problem your product solves actually occur for one person?
If the answer is daily, daily usage is a reasonable expectation and weekly usage is a warning. If the answer is once a quarter, then somebody using it once a quarter is not churn, it is the product working exactly as it should. Founders lose months to this, watching a weekly active number for a product that solves an annual problem and concluding they have a retention crisis.
This has a hard consequence worth facing early. If the problem genuinely occurs rarely, subscription pricing is fighting the shape of your own product, because you are asking for monthly payment for something used quarterly. Either find the adjacent problem that happens weekly, or price it as a one-off or an annual thing. Retention tactics will not out-argue frequency. There is more on matching the model to the product in how to price your first indie SaaS.
The three shapes of not coming back
They look identical in a chart and they need completely different responses.
| What happened | What it means | What to do |
|---|---|---|
| Signed up, never really used it | The first five minutes failed. This is not retention, it is activation wearing a disguise. | Fix the empty state and the first action, per the first five minutes. |
| Used it properly once, job done | The product worked. There was simply no second occasion, or not yet. | Check the frequency question above before changing anything. |
| Used it, came back, then stopped | The real thing. The value did not justify the effort on the second or third go. | Ask them. This is the group worth emailing. |
The third row is where the useful information lives, and it is usually the smallest group. Most founders average all three together and get a number that describes nobody.
At this size, read the rows
Cohort tables are for products with enough users that individuals stop mattering. You do not have that problem. With twenty or fifty accounts, open the list and go through it person by person: when they signed up, what they did, when they stopped, and what they told you at signup about why they were there.
A percentage of twenty is not a rate, it is a rounding error with a decimal point. Four people leaving looks like twenty percent churn and is actually four specific stories, at least two of which you can probably reconstruct from what they did.
What you are looking for is a pattern that is not visible in aggregate: everyone who stopped arrived from the same channel, everyone who stayed answered the signup question the same way, everyone who left never got past one particular screen. Those patterns are legible at twenty users and invisible at twenty thousand, which makes this a genuine advantage of being small.
If you asked a question at signup, this is where it pays for itself, because you can split the list by intent rather than by date. That is the same mechanic as segmenting by survey answers, applied to people who left.
The email that actually gets answered
Most founders send a churn survey with five multiple-choice options and get nothing back. A short, plainly human email to one person gets replies at a rate that surprises people the first time.
Subject: Did Lighthouse not fit? Hi Sam, You set up a waitlist in March and I noticed you have not been back since. No pitch here, and I am not trying to win you back. I am just trying to work out whether it did not do what you needed, or whether it did the job and there was nothing more to do. Either answer is genuinely useful. One line is plenty. Rouzbeh
Why this works: it names the specific thing they did, it removes the fear of a sales conversation, and it offers "it worked and I was done" as an acceptable answer. That last part matters because it is frequently the true answer, and a survey full of complaint options makes it impossible to give.
What not to do: do not ask what feature would bring them back. People invent an answer to be helpful, you build it, and they still do not return. Ask what happened, not what would fix it. The same rule as any other user conversation.
What actually brings people back
In descending order of how much they matter, which is roughly the reverse of the order founders try them.
- The value recurring. Something changed since they last looked, and it is theirs. New signups on their waitlist, new feedback, a number that moved. This is the only durable one, and it is a product question rather than a messaging one.
- Their data accumulating. A product holding six months of someone's work is harder to abandon than one holding nothing. This is why an empty account churns and a full one rarely does.
- A notification that carries the value itself. Not "you have unread items", but the thing: the three new responses, in the email. A message that is useful without clicking earns the click.
- A reason tied to their calendar. If the problem is monthly, arriving the day before it becomes urgent is worth more than any weekly digest.
- Re-engagement campaigns. Last, and honestly close to worthless without the four above. A reminder to use something that gave them nothing the first time is just a reminder of that.
The pattern is that retention is mostly built into what the product does between visits, not into what you send. If nothing happens while they are away, there is nothing to come back to.
When it means wrong customer, not bad product
Sometimes retention is fine for one kind of user and terrible for another, and the average hides both. A product might work beautifully for people running something continuously and not at all for people trying it once out of curiosity.
If that split exists, you have a positioning problem rather than a product problem, and the fix is upstream: change who you attract rather than what you build. Which channel is sending the users who stay is answerable with the same setup as which channel brought your signups, just measured further down.
This is a genuinely good outcome when you find it. It means the product works and you have been pointing it at the wrong room, which is a much cheaper problem than the alternative.
Frequently asked questions
What churn rate should I be worried about?
At twenty users, none of them, because the number is not measuring anything stable. Worry about the individual stories instead. Rates start meaning something around the point where you can no longer remember each customer's name.
Should I offer a discount to people cancelling?
Rarely, and never as the first move. It buys a month and destroys the information, since somebody who stays for half price has not told you the product was worth the full price. Ask why first.
Is it worth building notifications to pull people back?
Only once something happens while they are away that is worth telling them about. Building the notification system before the recurring value exists is building the delivery mechanism for an empty envelope.
They said they would come back later. Do I believe them?
Treat it as politeness rather than a plan, the same way you treat "I would definitely use that" before launch. What people did is evidence; what they intend is decoration.
My product genuinely is used once. Is that fatal?
No, but it changes the business rather than the tactics. One-off usage means one-off pricing, and growth that comes from new customers instead of retained ones, which needs a repeatable acquisition channel more than it needs a retention plan.
The reason this is worth doing carefully is that the three shapes of leaving are indistinguishable in a chart and demand opposite responses. Work out how often the problem happens, separate the people who never started from the people who stopped, then email a handful of them like a person rather than a dashboard. Twenty users is few enough to know the real answer, which is a position you only get to occupy once.
Lighthouse keeps the answers people gave you at signup next to their account, so when someone stops showing up you can see why they came in the first place. From an indie dev, for indie devs and makers.