What Is a Good Churn Rate for Ecommerce? Wrong First Question. Your Churn Number Is a Definition, Not a Fact.

I sat in a meeting once where two teams argued about churn for the better part of an hour.

The email team said lapsed meant no purchase in 90 days. The paid media team said 30 days. Both had dashboards. Both dashboards were built correctly. They were just counting different people and calling them by the same name.

By the end of the meeting we had not decided anything, because we were not really disagreeing about customers. We were disagreeing about a definition nobody had written down.

I have seen some version of that meeting more than once. It is why I now get suspicious when someone asks me what a good churn rate is. For most B2C brands, there is no number to compare against until you know how the number was made.

A subscription cancels. A shopper just stops.

Churn means the share of customers you lose in a period. If you run a subscription, churn is easy to point at. A person cancels, or a payment fails and never gets fixed. There is a moment when they leave.

Most ecommerce and retail brands do not get that moment. Nobody cancels a shop. A customer buys in March, does not buy in April, and by June you are guessing. Are they gone, or are they just between purchases?

So every non-subscription brand invents an answer. Usually it is a fixed window. No purchase in 90 days means churned. Sometimes 60 days. Sometimes a year. That window is a choice, not a discovery. Change the window and your churn rate changes, even though not one customer behaved differently.

That is the whole lesson. Churn, for these brands, is a definition you set. Then the number follows.

How teams usually set it, and where it goes wrong

The common way is one window for everybody. Ninety days is popular because it is round and it fits a quarter. It is also wrong in both directions.

Think about two customers. One buys a supplement every month. If she has not bought in 90 days, she has missed two or three orders. She is probably gone. The other buys a good winter jacket. He may not buy anything again for a year, and he is still a perfectly happy customer. At 90 days, your report calls both of them churned. One call is probably right. The other is nonsense.

The better-run teams I have worked with do it differently, in three ways.

First, they set the window from the customer's normal gap between purchases, not from the calendar. If people usually buy every 30 days, being gone for 90 means something. If they usually buy every six months, it means nothing yet.

Second, they set different windows for different buying patterns. Monthly buyers get a short window. Seasonal or big-ticket buyers get a long one. It is more work. It is also the only way the number means anything.

Third, they check the window against what actually happens. A useful rule I have seen practitioners use: look at the point after which only about 3 to 5 out of every 100 lapsed customers ever come back and buy again. Past that point, calling someone gone is fair. Before it, you are writing people off too early.

One small example

Say you sell a product people normally use up in about a month. One pack lasts roughly 30 days.

A customer buys four packs in one order in January, because there was an offer. She then buys nothing in February, March, or April.

Under a 90-day rule, she churns in April. She goes on the win-back list. She gets the "we miss you" email and maybe a discount code.

But she has not gone anywhere. She bought four months of product in January. She was still using what she bought. She was never lost, so there was nothing to win back. If she buys again in May at full price, your report may even count her as a successful win-back, and the discount you sent her was money given to someone who only needed time, not an offer.

Now multiply her by a few thousand customers after a big sale month. Your churn looks worse than it is. Your win-back looks better than it is. Both mistakes come from the same place: a window that ignores how much product the customer took home.

What changed

A few years ago, most of this was calculated by hand, in a spreadsheet, by someone who knew the business well enough to argue about the window.

Now your tools do it for you. Your email platform, your analytics tool, and your dashboard each come with their own built-in idea of who counts as lapsed or at risk. Most teams never look at those settings. So the old meeting problem has got quieter and worse. You no longer have two teams using two definitions. You have three tools using three definitions, and nobody in the room knows any of them.

AI reporting has sped this up. Asking a tool "what is our churn?" gets you a clean number in seconds. It does not get you the argument about the window, and the argument is where the truth is.

Nothing about the customer's behaviour changed. What changed is that the definition is now hidden inside a default.

What I would do on Monday

Write your definition down, in one sentence, where everyone can see it. Ours is something like: a customer counts as lapsed when they have gone twice as long as their usual gap between purchases, and gone for good when they pass the point where almost nobody comes back.

Then rebuild the number once, using that definition, and compare it with the one you have been reporting. If the two numbers are far apart, you have not found a churn problem. You have found a counting problem, and that one is much cheaper to fix.

Do this before you spend anything on win-backs. There is no point trying to win back people who never left.

This is the kind of work I now do with B2C teams: fixing the definitions underneath the numbers, so the decisions made on top of them are sound.

Sources: Klaviyo Community discussions on funnel-aware segmentation and win-back flow timing; Shopify, Churn rate in ecommerce; practitioner guidance on defining the lapse period by return probability (Experimental).