How do you track new versus returning customer revenue?

In a web analytics tool, a returning visitor is a browser that has been seen before, within whatever period identifiers survive. Somebody who has bought from you four times, then clears their cookies or switches to a phone, is a new visitor. Somebody who browsed twice without buying is a returning one.

That measure has a legitimate use in understanding site behaviour, and it is the wrong input for a question about customers. Using it to calculate acquisition cost or repeat revenue produces numbers that will not survive contact with your order data.

Your ecommerce platform holds the right information. Every order can be checked against previous orders for the same customer, and most platforms expose exactly this split in their own reporting.

So the rule is simple. Behaviour questions go to analytics; customer questions go to the order records. Mixing them is the source of most disagreements between a marketing dashboard and a finance report.

The obvious definition is a customer whose order is their first at your store. It is the right starting point and it needs decisions attached before it can be applied consistently.

Decide whether orders from other channels count. Retail point of sale, marketplace orders imported into the platform, wholesale and business orders each change the picture, and treating them differently in different reports is how two people end up with two answers.

Decide how long a lapsed customer stays a customer. Someone who bought once four years ago and returns is technically returning, and treating them as new for acquisition purposes may better describe what actually happened. Either choice is defensible; leaving it undecided is not.

Decide how subscriptions are counted. A recurring charge is an order in the platform, and counting every renewal as a returning purchase will make your repeat metrics look excellent while telling you nothing about whether new people are joining.

Write the definition into the report itself so nobody has to reconstruct it later.

Guest checkout with a different email address is the largest source of error. The same person orders once with a personal address and once with a work address, and the platform records two customers. Matching on address or phone as well as email recovers some of it.

Multiple stores or regions are the second. A store running separate storefronts for different countries will count a customer twice if they buy from both, unless the reporting joins them.

Point of sale and marketplaces are the third. Orders arriving through those channels may or may not be attached to the customer record, and the choice changes both the count and the revenue split.

Refunds and cancellations are the fourth. A first order that was refunded in full arguably did not acquire a customer, and leaving refunded orders in the new customer count inflates acquisition and understates cost.

None of these needs to be perfect. They need to be handled the same way every month, so the trend is real even if the absolute number carries some error.

New customer count is the first, because it is the growth number. A store can hold revenue steady for a long time on repeat purchases while acquiring nobody, and this is the figure that shows it.

Blended acquisition cost is the second: total marketing spend divided by new customers. It ignores attribution entirely, which is exactly what makes it reliable, and it is the number to compare against your first-order contribution and your lifetime value.

Share of revenue from new customers is the third. A rising share means growth is coming from acquisition; a falling share means the base is carrying the business. Neither is wrong, and knowing which one you are looking at changes where the next budget should go.

Underneath those, keep a cohort view. Customers grouped by the month of their first order, tracked for repeat revenue over time, is the report that tells you whether the customers you are buying are getting better or worse. That is a quarterly review rather than a weekly one.