Audiences AIRepeat Rate Is a Container
A 60% repeat rate. Profit down 38%. Same brand, same twelve months, and both numbers checked twice because the founder assumed one of them had to be wrong. Neither was. He runs a D2C fashion label, and what he had was a measurement problem, one of the most common ones I get called about. The figures throughout are a composite of client engagements, details altered.
Start on the profit side, because that half is arithmetic. Revenue had grown 37% over the year. Procurement costs had grown 40.8%. When input costs outrun the top line by almost four points, the margin a loyal base is supposed to throw off gets absorbed before it reaches the bottom of the statement. Deeper inventory buys ahead of sale events, richer codes to keep the repeat number moving, a heavier cost mix across the assortment: each looked defensible on its own. Together they meant the most celebrated metric in the business was also the most expensive thing in it.
Now the 60% itself. Repeat rate is not loyalty. It is a container, and tipping it out spills at least six different customers onto the table: full-price loyalists, sale-driven buyers, discount-trained buyers who have learned to wait for a code, habit buyers, replenishment buyers, and people who were never at risk of leaving at all. The dashboard counts every one of them as the same event. The margin line disagrees on every single order. A customer returning inside a sale and a customer returning at full price look identical in the retention report and behave nothing alike on contribution, and that gap is exactly where the decision problem lives.
Make it concrete with a split I will label illustrative, because every brand's mix is different. Suppose those 60 points break into 15 of full-price loyalists, 20 of sale-driven and discount-trained buyers, 15 of habit and replenishment, and 10 who would have repurchased with no prompting whatsoever. On that mix, a third of the celebrated retention is being bought at a discount, and a sixth of it was free. The blended rate can hold perfectly steady while the composition underneath swings by ten points in a quarter, and composition is what decides whether the next retention dollar earns anything.
Worse, the bought portion compounds. Every sale cycle that rescues the repeat number teaches another slice of the base to wait, so holding 60% next quarter takes a slightly deeper code, a slightly earlier event, a slightly bigger inventory buy behind it. That is how revenue up 37% and procurement up 40.8% happen to the same company at the same time. The brand was paying a rising price for a flat number and reading the flat number as stability.
There is a second way the number flatters you, and it involves no customer behavior at all. Repeat rate is a ratio. Slow down new-customer acquisition and the base tilts toward people who already know the brand, so the percentage climbs while nothing underneath improves. I have sat in a review where a rising retention line drew a round of nods, and the line was describing an acquisition slowdown. The same chart was telling the wrong story.
The way out begins with one separation: organic repeat versus offer-assisted repeat. Tag each returning order by whether an offer touched that customer in the run-up to the purchase. Read contribution for the two groups separately. Then check how much of your repeat revenue lands inside sale windows. For most brands this is an afternoon of analyst work, and it changes the retention conversation entirely, because it shows which customers the discounts converted and which ones they merely accompanied. The standard failure is over-discounting people who were coming anyway while under-nudging the ones who needed the push, and a blended repeat rate will hide that for a decade of reporting.
The fashion brand rebuilt its view this way and found just under half of its repeat revenue had an offer attached. The retention budget did not shrink; it got redirected toward the customers whose behavior it could actually change. Which is the whole point of measuring anything. So before the next retention push goes out, make the number answer for itself: are you increasing repeat, or buying it?
If a number in your business needs taking apart like this, that is the work of Interpret.