In a controlled trial we ran with a DTC apparel brand, roughly three in ten units came back as returns. The revenue dashboard looked healthy the entire time. Contribution was flat. The gap between those two sentences lived almost entirely in returns, and in where the reporting chose to put them.
Most stacks book returns as an operations cost: a reverse-logistics line, a warehouse metric, a quarterly headache. That framing is not wrong, but it hides the more expensive truth. Returns are customer behavior, they concentrate in specific customers, and until they are netted into customer-level value, every "best customer" list in the building is partly fiction.
Revenue counts the outbound trip
A revenue dashboard records what leaves the warehouse. It does not subtract what comes back, or the two shipping legs, or the inspection and restocking, or the markdown when the item can't be resold at full price. In most categories the gap between gross and net is a rounding story. In apparel, at three in ten units, it is the story.
Watch what that does to two customers who look identical on a revenue report:
| Customer A | Customer B | |
|---|---|---|
| Orders over 12 months | 3 | 4 |
| Gross revenue | $560 | $560 |
| Units returned | 1 of 9 | 9 of 14 |
| Refunds and return handling | $74 | $438 |
| Net contribution at margin | +$128 | Negative $41 |
Customer B is a bracketer: three sizes ordered, two sent back, every order. On the revenue leaderboard, A and B are the same customer. One of them is funding your growth; the other is charging you for the privilege of shipping boxes both directions.
Returns concentrate, which makes them a signal
If returns were spread evenly across customers, netting them would shrink everyone's value by the same fraction and change no decisions. They are not spread evenly. Return behavior clusters hard in a minority of customers, bracketers and serial refunders, and it is stable over time: a customer's past return rate is one of the strongest predictors of their future one.
That is what makes returns a selection signal rather than a cost line. When we scored the trial brand's roughly 120,000 customers by 12-month forward value, we netted expected returns into every score, and in apparel the netting is the whole story. Customers who looked prime on gross revenue fell out of the top decile; quieter customers who kept what they bought rose into it. The pool that eventually justified investment, the one the trial's offer built +$5.20 of measured forward value per customer on, was selected net, not gross. Selected gross, part of that budget lands on Customer B.
The same logic runs upstream into acquisition. Lookalike audiences built from a gross-revenue "best customers" seed teach the ad platform to find more bracketers. The waste is not just in the retention budget; it compounds at the top of the funnel.
The strongest objection
"Returns are a fit problem. Better size guides and product pages fix returns; a scoring model doesn't."
Partly true, and worth doing. Better fit content lowers the file-wide return rate, and in apparel that is real money. But it addresses the rate, not the allocation. Even in a catalog with perfect size guidance, behavior-driven returners exist, wardrobers, chronic bracketers, buy-to-photograph customers, and they will absorb whatever marketing you aim by gross revenue. Fixing fit changes how many units come back. Netting returns into customer value changes which customers you fund. They are different levers, and only the second one is a targeting decision you can change this quarter without touching product.
What netting changes operationally
Once every customer-level number is net of expected returns, a set of decisions gets sharper immediately: who gets the winback offer and who is suppressed from it, who qualifies for free exchanges versus store-credit-only, which customers seed your lookalike audiences, and whose "VIP" status quietly expires. None of that requires a new channel or a new tool. It requires the value number underneath the decisions to stop lying about returns.
The verdict
In apparel, a customer value number that is gross of returns is not an approximation. At three in ten units, it is a different number about a different customer. Returns are not the cost of doing business; they are information about which business is worth doing.
- Pull your top 100 customers by gross revenue, net out their returns at margin, and count how many fall out of the list. That count is your exposure.
- Make every per-customer value figure in your stack net of returns and discounts. If a tool can't, rank outside the tool.
- Treat a customer's return rate as a targeting input, not just an ops metric: suppress, gate, and seed audiences on net value.
The free diagnostic scores your customers on forward value net of returns and discounts, and shows you exactly who moves when the netting is done honestly. Client details and figures here are altered for confidentiality; the direction of every result is as run.
Take it with you. The Returns Self-Audit is a free field guide and worksheet: the four customer profiles, the fraud-versus-abuse signals, and a fill-in sheet to compute net contribution after returns on your own numbers. It runs in an afternoon, no access required.