Returns are a product-data problem wearing a logistics costume
Most shops fight returns in the warehouse. The damage is done on the product page. How to measure return reason per SKU, find the ten products doing the harm, and fix them.
Photo: free stock photography (Unsplash licence) — see imprint
You cannot fix a number you have never broken down
Ask a shop owner about returns and you get one number: the rate. Ask them which products cause it and the room goes quiet. That single number is useless for doing anything about the problem, because it averages a catalogue in which most articles come back almost never and a handful come back constantly. The average tells you the building is on fire. It does not tell you which room.
So break it down by SKU and sort. In almost every catalogue we have looked at, the distribution is brutally uneven — a small number of articles produce a wildly disproportionate share of the returns, and the rest of the catalogue is fine. That list of ten or twenty offenders is the whole project. Everything else is a rounding error, and every generic 'reduce returns' initiative that does not start with that list is spending effort where there is nothing to win.
The return reason is the only field that matters, and yours is broken
Return reason per SKU is where the diagnosis lives. Too small is a sizing chart problem. Not as pictured is a photography problem. Not as described is a copy problem. Damaged is a packaging or carrier problem. Changed my mind is often a pricing or a competing-tab problem. Each one points at a different department and a different fix, and knowing the split for a specific article turns an argument into a work order.
Which is why it is depressing how bad the data usually is. The reason field is free text, so it contains 4,000 unique values including 'siehe Mail' and 'k.A.'. Or it is a dropdown with one option everybody picks. Or the warehouse staff select whatever is first in the list because they are paid to move parcels, not to fill in forms. Fix that before you buy any tooling: six to eight reasons, mutually exclusive, mandatory, chosen by the customer in the return portal rather than guessed by a picker three days later.
- Return rate per SKU, not per shop — the average hides everything actionable.
- Six to eight mutually exclusive reasons, mandatory, picked by the customer.
- Cost per return in euro — labour, shipping, inspection, write-off, refund fee.
- Return rate by variant, not just by parent — one size is usually the culprit.
- Rate by customer, so you can see the small group ordering three sizes to keep one.
The fix is almost always on the product page
A return is a customer telling you that what arrived did not match what they expected. Expectations are set entirely by your product page. That means the fix lives in the photo, the description, the dimensions table and the size chart — not in the warehouse, where the only thing left to decide is how efficiently to absorb a mistake that was already made three days earlier.
The interventions are unglamorous and they work. A photo with a hand or a common object in it for scale. Actual measurements instead of S, M and L. A note that says this style runs small, because it does and your reviews already say so. A second photo of the colour under daylight, because your studio lighting is lying about the beige. A material description honest enough to lose the order to someone who would have returned it anyway — that is a good trade, and it is the trade most people are unwilling to make.
Free returns is a pricing decision, not a service decision
Nothing is free. If you pay the return label, that cost is inside your price, spread across every customer including the ones who never send anything back. That is a legitimate choice — it is exactly how flat-rate insurance works, and in fashion it is close to unavoidable because your competitors made the decision for you. But make it as a pricing decision, with the arithmetic written down, instead of drifting into it because a consultant said it lifts conversion.
Do the sum for one article. If a product leaves 25 euro of margin and a return costs 15 euro all-in — label, handling, inspection, repackaging, the payment fee you do not get back — then a return rate of forty percent on that article means you are earning roughly nine euro per order sold, not 25. That is the number that decides whether it stays in the catalogue, gets a better photo, or gets a higher price. Not the conversion rate.
In B2B returns are rarer and far more expensive
B2B shop owners often wave the topic away: our rate is two percent, it is not a problem. The rate is not the point, the ticket is. A wrong pump in a consumer shop is a parcel and twelve euro. A wrong pump on a B2B order is a machine standing still, an engineer on site with nothing to fit, a customer-specific configuration nobody else will ever buy, and a phone call in which the words framework agreement appear. There is no right of withdrawal to hide behind, so every return is a negotiation.
The good news is that the B2B fix is even more clearly a data fix. Your buyer is not returning a jacket because the colour disappointed them; they are returning a part because the thread was wrong, the voltage was wrong, or the datasheet on your page was three revisions behind the manufacturer's. Complete, correct, current technical attributes — with the drawing and the datasheet attached — remove most B2B returns at source. That is a PIM project, and it is far cheaper than the returns it prevents.
| Return reason | What it actually means | Where the fix lives |
|---|---|---|
| Too small / too large | Your size chart is decorative | Real measurements, fit note, variant data |
| Not as pictured | Studio lighting or a scale illusion | Daylight shot, object for scale |
| Not as described | Copy oversold it, or attributes are wrong | Product data and honest copy |
| Damaged on arrival | Packaging or carrier, not the product | Warehouse — the one case that is logistics |
| Changed my mind | Price, delivery time, or a competing tab | Pricing and expectation setting |
- A shop-wide return rate is a vanity metric — only rate per SKU and per variant tells you what to fix.
- If your return reasons are free text, you do not have return data — you have anecdotes.
- Free returns is an insurance premium hidden in your price; decide it with arithmetic, not with a slogan.
- In B2B the rate is low and the per-case cost is brutal — the answer is correct technical data, not a returns portal.
Frequently asked questions
It depends so heavily on the category that a cross-industry benchmark is close to meaningless. Fashion with size variants sits in a completely different world from spare parts or consumables. The comparison worth making is against yourself: your rate this quarter versus last, and this SKU versus the rest of your own catalogue. That comparison is always fair and always actionable.
Do the arithmetic before the philosophy. Charging shifts a cost from your price to the returning customer, which reduces returns and also reduces orders — the question is the ratio, and it differs by category. In fashion, charging usually costs you more in conversion than it saves. In heavy or bulky goods it is often obviously right. Run it as a test on one category, not as a policy announcement.
Technically yes, if you have enough order history — and it is usually the wrong project. Knowing an order will probably come back does not let you do much: you cannot refuse it, and nudging the customer at checkout mostly just loses the sale. The same effort spent fixing the description and photos of your worst twenty SKUs prevents the returns instead of forecasting them.
The rate is low, the per-case cost is high, and the relationship damage is higher still. A wrong part means a customer's machine is idle, and that conversation reaches the people who sign your framework agreement. Because most B2B returns come from wrong or outdated technical attributes, the cheapest countermeasure is boring: keep the datasheets, dimensions and compatibility data current.
We do this for a living — Shopware, Node.js, React, ERP integration and automation for B2B.
Talk to an engineer