Reduce your return rate: 9 measures with a measurable effect

Take your shop's last fifty returns and write down the reason for each one, exactly as the customer gave it. If you have never done this, you will notice two things: for some, the reason is missing entirely, and for the rest, three or four phrases keep repeating. “Too small.” “Looks different from the picture.” “Ordered two sizes.” That is not customer whim. It is a list of places where your shop failed to answer a question.

The scale is well known. According to a Bitkom survey from April 2024, 67 % of German online shoppers who have ever returned something name a wrong size as the reason. 56 % name a faulty or damaged product, 41 % a difference from the description or picture, and 29 % say they deliberately ordered several variants. Size, expectation, selection orders: those are the three causes you can actually work on.

This article lists nine measures, sorted by cause. For each one, you get what it costs, how to measure the effect and where its limit is. The framework all of this sits in is the pillar article Returns process for online shops: step by step.

First: no return reasons, no measures

The first measure is not on the list because it is the prerequisite for all the others: capture the reason for every return, and do it in a structured way, not as free text. According to the EHI study 2025, only 29.5 % of surveyed merchants capture return reasons in a fully automated way; 34.4 % do it manually, and a fifth are not considering automation at all.

Offer five to seven fixed reasons in your returns portal, for example:

• too small
• too big
• not as expected
• damaged or faulty
• wrong item delivered
• ordered several variants
• other

Then you have a per-product analysis after one month. And then you know which of the following nine measures applies to you at all.

Cause 1: The size does not fit

Measure 1: A size chart per product, not per shop

A generic chart in the footer helps nobody. What helps: the measurements of the specific item (chest width, length, inside leg) right on the product page, in centimetres, with one sentence on fit (“runs narrow, choose the larger size if you are in between”). This costs one round of measuring per item and is the cheapest measure on this list.

Metric: share of returns with reason “too small” or “too big” for this product, before and after.

Measure 2: Fit notes from real returns

If a product comes back with “too small” more often than average, this sentence belongs on the product page: “This model is often ordered one size up.” That is not an admission, it is information the customer would otherwise learn through a return. The data for it comes from measure zero.

Limit: with very small quantities, the signal is too thin. Wait until a product has a double-digit number of returns before you turn it into a statement.

Measure 3: Size recommendations in support instead of a standard reply

Customers ask before buying: “I usually wear M, which size fits this jacket?” Replying “Please refer to our size chart” shifts the decision onto the customer and the risk into a return. Better: a concrete recommendation based on the item's measurements and what the customer told you.

This is how it works at Humane: if the measurements and fit notes are stored in the knowledge base, the agent answers such pre-purchase questions with a concrete recommendation in your brand's tone. And if you correct a reply (“for in-between sizes, always the larger one”), that becomes a rule for every following case.

Cause 2: The product does not match expectations

Measure 4: Photos that answer the customer's question

Four pictures on a white background show the product. They do not show how big it is in the hand, what the colour looks like in daylight, how the fabric drapes. Add at least one photo with a size reference per product, one in use and, for clothing, one on a model with height and size worn. If a colour regularly comes back as “not as expected”, the photo is usually the problem, not the product.

Metric: share of returns with reason “not as expected” per product.

Measure 5: Descriptions that also name downsides

“High-quality workmanship” is not information. “The fabric is firm and barely creases, but feels warm in summer” is. One sentence that honestly names a downside prevents orders from customers the product does not suit, and so reduces the return before it happens. That costs a few orders in the short term and saves considerably more in the long run.

Measure 6: Reviews with questions about fit

After the purchase, ask customers not just for stars but two concrete questions: “Does the product run small, true to size or large?” and “Does the colour match the picture?” These two answers, aggregated on the product page, are the most credible information a customer can get before buying.

Cause 3: The selection order

Measure 7: Do not punish the returns portal, steer it

Someone who orders two sizes to keep one is using your return policy as intended. You will not abolish that, but you can steer it. If the portal offers an exchange to the other size at the return request stage, before the refund option appears, some of the returns turn into exchanges. How that works in detail is in Exchange instead of refund.

Limit: customers who deliberately ordered two sizes do not need an exchange, they already have the right one. Here only measures 1 to 3 help, so that next time they order just one size from the start.

Measure 8: Return shipping costs with a sense of proportion

The temptation is strong to curb selection orders through return shipping fees. It works, but it works on all customers, including the ones who honestly return something once a year. If you are considering this lever, know the numbers first: what customers expect, what competitors do and what the law allows. The maths is in Who pays return shipping. A middle ground many merchants use: free returns above a minimum order value or for exchanges, paid below that.

Cause 4: Delivery and packaging

Measure 9: Take transport damage and delivery times seriously as return reasons

Two reasons from the Bitkom list have nothing to do with the product: 56 % of respondents have returned something because it was faulty or damaged, 13 % because the delivery came too late. Damaged goods are often a packaging problem, and late delivery is a communication problem: someone who needs a gift by Friday and receives it on Monday sends it back no matter how good the product is.

Both can be tackled with basic tools: shatterproof packaging for the three products that most often arrive damaged, and a delivery date instead of a range in the shipping email. The latter also reduces the delivery questions we described in Reducing WISMO requests.

What you can realistically expect

None of these measures halves your rate. Anyone who claims that does not know your shop. A fashion retailer will have far more returns than a household goods retailer even with perfect size charts, because fit cannot be checked on a screen. Benchmarks by industry are in Return rates in Germany 2026.

What you can expect: for products where you found and fixed a clear cause, the share of exactly that return reason drops visibly. That is the number to measure, not the overall rate. The overall rate only moves once you have run the procedure across your twenty best-selling products.

A worked example, as an assumption: 1,000 orders a month, 200 returns, 80 of them with reason “size”. If a size chart and fit note on the five most-returned items prevent a quarter of those 80, that is 20 fewer returns a month. With handling costs that, according to the EHI study, come to up to €10 per item for more than half of merchants, that is a modest but real amount, and the effort was one afternoon with a tape measure.

What automation cannot do here

A support agent can answer size questions, capture return reasons cleanly and suggest an exchange at the return request stage. It cannot reshoot your product photos, write your descriptions honestly or decide whether you introduce return shipping fees. The measures with the biggest effect are product work, not support work. Support can make them visible by telling you which product comes back how often and for which reason.

In short

Returns come from three causes: the size does not fit, the product does not match expectations, or the customer deliberately ordered a selection. Capture the reasons first, then you know which cause dominates for you. Work through your best-selling products and fix the one thing that triggers the return reason for each. Measure the share of that reason per product, not the overall rate. And leave selection orders alone as long as you can steer them towards an exchange.

If you want to know what size questions and return requests look like when an agent answers them with your product data: try it free for 14 days, no credit card required.