Summary. The store’s margin advantage is not only about lower fulfillment cost. A meaningful part of it is created at the moment of advice, because a purchase someone was helped to make is less likely to come back. Returns are now a very large number, they fall hardest on the digital channel, and they are the clearest place to see what capable store staff are worth. This paper puts the arithmetic in your hands, answers the objection that returns policy is a policy problem, and sets out an eight-week test that produces a number from your own data.

The size of the returns problem

The National Retail Federation, with Happy Returns, put total US retail returns at $849.9 billion in 2025, and estimated that 19.3 percent of online sales would be returned. The same research found 9 percent of all returns are fraudulent, that 82 percent of consumers treat free returns as an important consideration when shopping online, and that 45 percent think it is acceptable to bend the rules when returning something.¹

Nearly one in five online sales coming back is not a customer service statistic. It is a margin structure. Every one of those transactions carries outbound shipping, return shipping, handling, inspection, restocking, and frequently a markdown on goods that can no longer be sold as new.

The consumer figures in that research explain why the problem is not going to solve itself. Free returns are now a purchase criterion for the large majority of online shoppers, which means tightening the policy carries a demand cost. If the lever most retailers reach for first is expensive to pull, the alternative levers become worth understanding properly.

Why a store sale behaves differently

The gap between channels is not mysterious. It comes from the customer having handled the product, and from someone having helped them choose.

Three of the most common reasons for returning something are addressable in a conversation and not addressable on a product page: it did not fit, it was not what the buyer expected, or it was the wrong choice for the use they had in mind. An associate who asks what the customer needs it for, checks the fit properly, and sets expectations about how the product behaves is doing return prevention, whether anyone calls it that or not.

The strongest version is the one that feels like lost revenue at the time. An associate who says a product is not right for this customer has just prevented a return, protected the margin on a garment that would have come back shopworn, and made the next visit more likely. Trained associates do that. Untrained ones ring up whatever the customer brought to the register.

“Returns are a policy problem, not a training problem”

The natural place to attack a return rate is the returns policy: shorten the window, charge for the label, require the tags. It is faster than training and the effect is immediate. So why is this a capability argument?

Three reasons, and a concession.

Policy suppresses returns without preventing them. A customer who cannot return an unsuitable purchase still has an unsuitable purchase. What you have removed is the transaction, not the mismatch, and what remains is a customer who will be more careful about buying from you next time.

The demand cost is measurable and large. With 82 percent of online shoppers treating free returns as an important consideration, a policy change is a pricing change in disguise, and it is felt at the top of the funnel where it is hardest to attribute.¹

Policy does nothing for the store channel. Store returns are lower already, so tightening rules mostly disciplines the channel that is not creating the problem, while leaving the mechanism that makes store sales stick unexamined and unfunded.

The concession: fraud is a policy and systems problem, not a training one. With 9 percent of returns fraudulent and 45 percent of consumers relaxed about bending the rules, some of this genuinely does need rules and detection rather than better advice.¹ The two approaches address different portions of the same number and neither substitutes for the other.

Work out what it is worth on your own numbers

We will not invent a cost per return, because it varies enormously by category, weight and whether the item can be resold at full price. You already have the inputs.

1. Take your return rate for a category, split by channel.
2. Multiply by units sold to get returns per year.
3. Cost each return honestly: shipping both ways where applicable, handling and inspection labor, plus the average markdown taken on returned stock, plus the share written off entirely.
4. Ask what a one percentage point reduction in the return rate would be worth on that category.

That last figure is usually the moment the conversation changes, because a single point of return rate on a large category is frequently a larger number than the entire training budget being argued over.

Two refinements make the number harder to argue with. Cost the returned unit at the margin you actually recover, not at the original price, since a resold return is frequently a markdown. And separate the returns that come back saleable from those that do not, because the second group is where the real money is and it is usually the group that better advice would have prevented.

Where the return rate is really made

It is worth being specific about which conversations move this number, because “better service” is not actionable and four things are.

The use question. What is it for, and where will it be used. Most unsuitable purchases are unsuitable for a use the associate was never told about.

The fit check. In apparel and footwear this single behavior drives more return prevention than everything else combined, and it is the one most often skipped when the store is busy.

The honest limitation. Saying what a product will not do costs one sentence at the counter and prevents a disappointed customer at home. Associates will only do it if they have been told it is acceptable.

The alternative offered. Steering someone from the wrong item to the right one preserves the sale and removes the return. Steering them out of the category entirely preserves the relationship. Both need permission from the scorecard.

The other structural advantage

A customer standing in a store has already spent the expensive part: the decision to come. They traveled, parked, walked in. Their intent is higher than a browser session by construction, which is why in-store conversion is not comparable to online conversion and why the marginal value of a good interaction is so high.

That is what makes staff capability a channel-economics question rather than a service nicety. The traffic has been paid for by property, marketing and the customer’s own effort. Whether it converts, and at what basket size, is decided in a conversation lasting a few minutes.

What this asks of training

• Comparative knowledge, not feature lists. Preventing a wrong purchase requires knowing which product suits which need, which is a different body of knowledge from being able to describe one item.
• Fit and sizing as a discipline. In apparel and footwear this is the single largest driver of returns, and it is teachable in minutes with the right content.
• Expectation setting. How a fabric wears, how a finish ages, what the product will not do. Honest limits at the point of sale prevent disappointment at home.
• Explicit permission to steer away. If associates are measured only on units, they will not talk anyone out of a purchase. This is a targets question as much as a training one.
• The returns desk as a source of content. The reasons customers give when they bring things back are the specification for next month’s training, and almost nobody routes them anywhere useful.

How to test it in eight weeks

This is the rare training question with a clean natural experiment available.

1. Pick a category with a high return rate and a clear fit or suitability component.
2. Deliver a short module on comparison and fit to half your stores, matched by volume and market.
3. Hold the other half back for eight weeks.
4. Compare return rates between the two groups, not before and after, so the promotional calendar cannot explain the result.

Two practical notes on running it. Allow for the return window: if customers have sixty days, an eight-week comparison is measuring returns against sales made partly before the module landed, so either extend the window or compare returns as a share of the cohort of sales made after the start date. And decide in advance what result would cause you to stop, because a test with no failing condition is a rollout with extra steps.

If it works, you have a number in your own data rather than a vendor’s. If it does not, you have learned something more valuable than another case study.

The measurement side of running that test matters as much as the content, because a result you cannot attribute to a specific cohort is not a result.

Model the wider training case on your own figures.

References

1. National Retail Federation with Happy Returns, a UPS company, 2025 Retail Returns Landscape, October 15, 2025. Total US returns of $849.9 billion in 2025; an estimated 19.3 percent of online sales returned; 9 percent of all returns fraudulent; 82 percent of consumers say free returns are an important consideration when shopping online; 45 percent say it is acceptable to bend the rules when returning items. nrf.com