How repeat purchase rate is calculated
Take every customer who bought during a period, count how many of them placed more than one order in it, and divide the second number by the first. The arithmetic is the easy part. The two decisions wrapped around it are what make the figure trustworthy or worthless.
The first is the window. Measured over a short window, a shop selling something people buy once a year looks dreadful while a grocery shop looks brilliant, and neither reading means anything — the window has to be long enough for a normal repurchase to have had time to happen. The second is identity. If orders cannot be matched to a person, returning customers are counted as new ones and the rate reads far below the truth. Guest checkout, a changed email address, an order taken over the phone and a household sharing one number all break that match.
Why repeat purchase rate matters
Acquiring a customer costs money every single time. Selling again to someone who has already bought costs comparatively little, because you can reach them directly and they have already tested whether you deliver what you promised. A shop with a healthy repeat rate compounds. A shop without one has to buy every order it will ever receive, and growth stops the day the advertising does.
It is also the most honest signal you have about the product and the service around it. Advertising can make a first order happen. Only the product itself, the delivery and the way problems are handled can make a second one.
Common mistakes with repeat purchase rate
Counting orders rather than customers is the frequent one, and it flatters. A rising order count can sit on top of a shrinking base of buyers who each order more often, which is a very different business. Blending channels is the next: marketplace buyers usually cannot be matched to your own site’s customers at all, so a combined figure describes nobody.
The most expensive mistake is treating a low rate as an email problem. If the product disappoints, the delivery runs late or the parcel arrives damaged, no reminder sequence will rescue it — the sequence simply delivers the disappointment to more people, faster. In markets where much is paid on delivery there is a further distortion, because orders refused at the door were never really customers to begin with.
How to act on it
Fix identity before anything else. Capture a reliable email address or phone number at checkout, deduplicate customer records, and make sure repeat orders join to the same person in your reporting. Then read the number by cohort — customers who first bought in a given month, followed forward — rather than as one site-wide figure, because a cohort view shows whether the shop is getting better or worse while a single average hides it.
After that the levers are ordinary. Deliver when you said, tell people where the order is, ask for a review, remind buyers of consumables when they are due, and give a reason to return that is not always a discount. Most of that is a set of triggered messages, which is what ecommerce automation is for, and it should be read alongside customer lifetime value rather than on its own.
