For a brand that earns its money through people coming back, rather than through a constant supply of new customers, the leading number is the repeat purchase rate. Acquisition cost is real, and it is secondary. A brand with a low acquisition cost and a poor repeat rate is buying customers it cannot keep, which is not a problem a marketing budget can solve. A brand with a high acquisition cost and a strong repeat rate has a channel problem instead of a product problem. The two look similar on a profit and loss account for several months, and then they stop looking similar at all.
The useful diagnosis is to read the repeat rate by the channel that brought each group of customers in. If the customers who arrived through one channel come back at twice the rate of those who arrived through another, the two channels are not interchangeable, even when their acquisition costs look the same. The expensive channel that brings loyal customers is cheaper than the cheap channel that brings people who never return, and the budget should move accordingly rather than towards whichever number reports fastest.
The second diagnosis is by product. Where a small number of lines produce most of the repeat purchases, the range is probably too wide. Either the rest of the range earns its place by bringing in customers who go on to buy those lines, or it should be retired. Most brands carry more complexity than they need, and they discover it only when they discontinue something and nothing happens.
The difficulty with leading on repeat purchase is that it takes a year or more to read properly. Acquisition cost can be measured in days. Anyone under pressure drifts towards the number that answers quickly, even when the slower number is the one that decides whether the business works. Building the patience to wait for the repeat data, and to make range and channel decisions on it, is most of the discipline.
In this region there is a further complication. Where orders are paid on delivery, a refused parcel looks like a sale and then unwinds, so the first version of every number is optimistic. Repeat rates have to be read after returns and refusals have settled, which adds weeks to the wait and changes which channels look good.
We run our own brand, wearon.co, this way. Repeat purchase is the number the team watches, the range is kept deliberately narrow, and the channel mix changes only when the repeat data supports it. It is also why we sometimes advise against a campaign that would lift visits, because a customer who buys once and never returns costs more than the first invoice suggests.
One practical note. Measuring any of this properly needs the order history joined to the source of the first order, which many businesses have never set up. That work is unglamorous, it usually takes a few weeks, and without it every conversation about channels is a matter of opinion. It is also the first thing we set up on a new engagement, because until it exists there is nothing to argue from except the platform's own reporting.
