Academy › Publishing › Analytics
How often repeat customers come back
You'll walk away with the average return interval for your customers, service by service, and with a rule for choosing the mean or the median depending on how much data you have.
By Yodai · Published September 6, 2026 · Updated October 12, 2026
Knowing what share of your customers come back isn't enough. You also need to know how often the ones who return actually return, and that changes everything you'll do with the data.
A customer who comes back every three weeks isn't the same as one who comes back every three months. Treat them the same, and you'll worry too late about the first one and bug the second one with a reminder before it's even due.
Think about a drop-in gym and a dental clinic. At the gym, a repeat customer might show up every week. At the clinic, a repeat patient comes back every six months. Using the same reminder threshold for both is an expensive mistake: you'll reach out too late to the gym member and too soon to the dental patient.
The number you need is the average time between one visit and the next, counting only customers who have already come back at least once. With it, in the next lesson, you'll be able to pin down the exact week when someone stops being a customer who's running late and becomes a customer you've lost.
One customer's visits, one by one
Sarah has come to your hair salon four times, and Jake, three. For each customer the math is the same: sort their visits by date and subtract each one from the one that comes after it.
Sarah gives you three intervals: 35, 35, and 38 days. Jake gives you two: 45 and 209. That second jump for Jake could be a slow season for him, a trip, or just someone who takes longer to come back. It's the kind of data point that, if you have few customers, can throw off the average for your whole business.
You don't need just one interval per customer. If someone has come in four times, that's three intervals, and all three count separately in the final tally. A customer with lots of visits carries more weight in the result than one with just two, and that's fine: it's real information about their rhythm, not an error to fix.
Do the same for every customer who has come back more than once. The more intervals you gather, from more customers, the more reliable your final number will be. With five customers and two intervals each, you have ten data points, and with ten data points you can still get a strange number, like Jake's jump. With fifty or a hundred intervals, you start to see the real rhythm of your business, not the quirk of two or three people.
- 1Filter out the ones who've come back
Leave out anyone who's only bought or booked once. They count toward figuring out what share of customers repeat, but they don't give you an interval here, since there's no second date to subtract from. - 2Sort visits by date
For each customer, line up their visits in order, oldest to most recent. If your calendar or booking software already stores the dates, export that list instead of rebuilding it by hand. - 3Subtract date from date
Calculate the days between each visit and the next one. A customer with four visits gives you three subtractions; one with two visits, just one. - 4Add it up and divide
Add up every interval from every customer into a single list. Divide that sum by the total number of intervals you added. If the result comes out with decimals, round to the nearest week: it's easier to work with later than a number like thirty-four point seven days.
Mean or median, depending on how much data you have
Once the ranges are grouped, you still have one decision left: average them using the mean or the median. Pick the wrong one and you'll get a warning line that doesn't match your business.
- Calculated by adding up all the intervals and dividing by how many there are
- One odd case can throw the whole thing off: Jake's 209-day gap can push the average for your entire business up by weeks if you have few customers. With Sarah's and Jake's five intervals combined, the average comes out to 72 days, and the median to 38.
- It's the value that falls right in the middle when you sort all the intervals from smallest to largest
- An extreme case doesn't affect it, so it holds up better when you have little data
Which one should I use? If you only have a few intervals, use the median: a couple of outliers, like Javier, shouldn't decide the warning line for your whole business. With a lot of data, calculate both: if they come out similar, either works; if the average is noticeably higher, the odd cases are pulling it up, and the median better describes the normal pace.
Don't mix services that aren't alike
If you average haircut visits and hair coloring visits together in the same calculation, the number you get is useless. One has a rhythm of a month and the other of three, and the average of the two doesn't truly describe anyone. Calculate the interval separately for each service that has its own rhythm, and save it for that service's own reminder threshold.
Don't mix a new customer, who hasn't settled into a rhythm yet, with one who's been coming for years, either. If you like, calculate the interval only for customers with three visits or more, so the number reflects a habit and not a coincidence. If your business offers a single service and all your customers are established, you can skip this step: you already have one number and that's enough.
With the average interval for each service, you now have the missing piece to decide, in the next lesson, the exact week you stop waiting for someone.
- You've calculated the average interval between visits for your repeat customers, service by service
- You know whether to use the mean or the median depending on how many intervals you're working with
- You've identified which services or customers you can't mix in the same calculation
- You know that intervals are pooled together before averaging, not averaged customer by customer
Test yourself
Answer before opening each one. If one gives you trouble, go back to the section that explains it.
1. Your dental clinic has 15 customers who have come back at least once, and one of them took 400 days for their last checkup because they moved to another city and came back the following year. Mean or median?
Median: with only 15 customers, that 400-day case could shift the average by a couple of weeks, and it doesn't reflect the normal pace of your checkups.
2. Your business offers haircuts and manicures, and you'd like to come up with a single interval number for both. Should you?
No. Each service has its own rhythm, and mixing them gives you an average that doesn't describe your haircut customers or your manicure customers. Calculate a separate interval for each service.
3. A customer has come to your business twice, the second time just ten days ago. Does that already count toward the interval calculation, or do you wait until they have more history?
Count it: two visits already give you one interval, even if it's just one, and it'll carry less weight than a customer with many visits. You'd only leave it out if you've decided to calculate only customers with three or more visits.
With that interval in hand, you can now set a deadline: the exact week after which someone who hasn't come back stops being late and becomes a lost customer. That line, different for each service, is what you'll set in the next lesson.
NextThe week you stop expecting them back