Churn is a symptom
I have watched a lot of people attack churn directly, with discounts, pauses, win back offers and loyalty schemes, and get nowhere. The reason is simple. Churn is not a thing you can push on. It is the visible end of four or five separate problems, each of which has its own fix.
So the work is diagnosis first. It is slower and much less satisfying than launching a save offer, and it is the only thing I have seen produce a durable change.
The five reasons people actually leave
Written in the order they usually appear in a member's life.
They never arrived
They joined, looked once, and never came back. The membership never really started. This is the single biggest bucket on most sites and it is an onboarding problem, not a content problem. It is the whole subject of onboarding a new member well.
They got what they came for
They joined to solve one thing, solved it, and left satisfied. This is not failure. It is a reason to think about what a second year is for, and whether your site has one.
The habit broke
Something happened in their life, they missed a month, and the site quietly became a thing they used to do. Rejoining feels like admitting the lapse. Most of these people never say goodbye.
The value drifted
You moved, or they did. Your subject has narrowed or widened, their work has changed, and the match has come apart. This churn is healthy and you should not fight it.
Something went wrong
A charge they did not expect, a question that never got answered, an argument in the community. Small in number, loud in effect, and the only bucket where an apology is the fix.
From practice
The first year I actually asked leaving members why, in one open question with no options to pick from, forty per cent of the answers were some version of I never got started. I had spent the previous eighteen months adding content for people who had not read the content I already had.
Getting the evidence
Three sources, and you need all three because each one lies in its own direction.
- The cancellation question. One open box, on the way out, not required. Never a list of tick boxes, because a tick box tells you which of your guesses was least wrong.
- Last activity before cancelling. If most cancellations come from people who last logged in three months earlier, you do not have a cancellation problem, you have a silence problem that ended in a cancellation.
- Conversations. Ten notes to members who left, asking one question, offering nothing. A few will reply and their replies will be worth more than any dashboard.
Do this for a full quarter before you change anything. Churn is noisy month to month and you will chase ghosts if you react to a single bad four weeks.
If you cannot say which of the five reasons your churn is made of, any fix you apply is a guess wearing a plan.
The fixes that hold
For members who never arrived
A first fortnight with one specific thing to finish, and a real person noticing whether they finished it. Nothing else comes close.
For the broken habit
A predictable rhythm. Same thing, same day, so a member can build you into a week. Sites that publish whenever inspiration lands have worse retention than sites that publish something smaller on a fixed day, and it is not close.
For getting what they came for
Give the second year a shape. A deeper track, a different room, a role in the community for people who have been around. If you cannot describe what year two is, expect people to leave at the end of year one.
For something going wrong
Answer faster and apologise properly. Most of this bucket is recoverable by a reply within a day that does not read like a policy.
The tactics that quietly make it worse
- Discount to stay. It teaches members that leaving is how you get a better price, and it converts a clean exit into a cheaper version of the same exit three months later.
- Cancellation flows with obstacles. Every extra screen buys you a few days and costs you the recommendation. People remember how they were allowed to leave.
- Win back campaigns aimed at everyone. Members who left because the value drifted do not want to be chased. Aim only at the never arrived group, and only once.
- Adding more content. The most common response and the least effective. If people are not finishing what exists, more of it is not a gift.
A pause option is the one exception I recommend without reservation. Two or three months off, taken freely, saves a meaningful number of the broken habit group and costs you nothing but patience.
Measuring it without fooling yourself
Monthly churn as a single percentage is close to useless on a small site, because one member leaving moves it. Two better views:
- Retention by joining month. Take everyone who joined in March, and see what share are still here at three, six and twelve months. Compare cohorts rather than months. Changes to onboarding show up here first.
- Voluntary churn only. Strip out the failed payments, or you will read a billing problem as a content problem. failed payments and the churn nobody chose covers that split.
What I would do first
- Add one open question to the cancellation screen.
- Record last login date at the moment of cancellation.
- Collect a full quarter before changing anything.
- Sort the answers into the five reasons and see which bucket is largest.
- Fix that bucket only, and leave the other four alone until it is done.
