The Hidden Cost of Returns and How to Fix It

A customer orders a £50 item, decides it is not what they expected, and sends it back. You issue the refund and move on. On paper, that return cost you £50.

In reality it cost considerably more. There was the outbound postage you already paid (£4-6), the return label (£5-8, usually at your expense), and the time someone spent receiving, inspecting and relisting the item (£3-10 in labour). Add a slice of the advertising spend that won the customer in the first place, the customer service time spent on the back-and-forth, and the days of cash tied up in stock that was supposed to be revenue by now, and the £50 refund has quietly become a £120-150 hit.

Most sellers never see that number. The refund is the visible part, sitting neatly in the payments dashboard. Everything else is scattered across courier invoices, payroll and ad accounts, which is why returns are one of the most consistently underestimated costs in ecommerce.

What a Return Actually Costs

Work through the margin maths and it gets uncomfortable quickly. Say you sell at a 35% margin. When an order is refunded in full, you lose the profit and carry every one of those extra costs on top. Depending on category and price point, the true cost of a return usually lands somewhere between 150% and 300% of the refund value. On typical ecommerce margins of 30-40%, one return can wipe out the profit from three or four successful sales.

Scale that up. If your average order is £50 and your return rate is 10%, somewhere between 3% and 5% of total revenue is being consumed by returns, quietly, every month. Rates vary a lot by category. Fashion and footwear can run at 20-40%, electronics at 5-15%, home goods at 8-20%. But the direction is the same everywhere: a business that gets its return rate down from 12% to 8% has just recovered 1-2% of total revenue, and it drops straight to the bottom line because no extra sales were needed to earn it. Depending on your scale, that is £5,000-20,000 a year in recovered profit.

Why Customers Send Things Back

Before anything can be fixed, it helps to be honest about why returns happen, because the reasons are not all equal.

The biggest and most fixable category is expectation mismatch. The colour looked different in the photos, the size ran small, the material felt cheaper than the listing suggested. The customer is not being difficult; the listing sold them something slightly different from what arrived.

Then there is transit damage, which better packaging and carrier choices reduce but never eliminate. There are genuine changes of mind, largely out of your hands. There are wrong items shipped, entirely your own doing and thankfully rare if picking and packing are solid. And in some categories there is deliberate abuse: customers who order three sizes intending to return two, or who wear an item once and send it back.

Each of these needs a different response, which is why "reduce returns" as a single goal tends to go nowhere.

Fixing Returns at the Source

The cheapest return is the one that never happens, and most prevention work happens in the listing.

Accurate descriptions do more than anything else. Exact measurements, honest material descriptions, real colour names and photos from several angles, ideally with something for scale, close the gap between what customers imagine and what turns up. For anything where fit matters, a proper size guide built from real customer feedback earns its keep many times over. "Customers between these measurements usually take a medium" prevents more returns than any policy change.

Packaging is the other early win. Spending an extra 50p to £1 per parcel on protection feels like a cost until you notice it preventing even 2-3% of damage returns, at which point it pays for itself immediately.

One counterintuitive move worth knowing: a longer return window often lowers the return rate. Give customers 60 days instead of 30 and the urgency disappears. They live with the product, get used to it, and keep it. The pressure of a closing deadline is part of what pushes borderline customers to send things back.

Taking the Cost Out of Processing

Some returns will happen no matter how good the listings are, so the other half of the job is making each one cheaper to handle. Worth saying plainly: none of this part needs AI. It is straightforward workflow automation, and it works.

A self-service returns portal lets customers start their own return, choose a reason and print their own label without emailing anyone. That alone removes £2-5 of customer service time per return. From there the workflow carries on without manual steps. The warehouse gets notified, the item is inspected against a short checklist, the refund fires automatically once inspection is approved, and stock levels update in your inventory system the moment the item is restocked. No refunds waiting three days for someone to press a button, and no oversold SKUs because a returned unit was sitting uncounted on a shelf.

For a store processing a few hundred returns a month, this typically halves the labour cost per return and takes days off the refund cycle, which customers notice.

Where AI Earns Its Place

The processing side is rules. The analysis side is where AI does work that a person realistically never gets round to.

Every return carries data: a reason code, often a free-text comment, a product, a customer, a timeline. Individually they tell you nothing. Across hundreds of returns, patterns emerge that nobody skim-reading a returns report on a Friday afternoon will ever catch. An AI layer reading that data flags that one SKU's "doesn't match description" returns have doubled since you moved to a new supplier batch, which usually means the product changed and the photos didn't. It notices that one size in one style comes back far more often than the rest of the range, pointing to a grading or labelling problem upstream.

The same analysis catches abuse. Serial returners spread across multiple accounts at one address, return timings that fit wear-and-return behaviour, reason codes that make no sense for the product category. Sellers who actively analyse their returns data typically cut their return rate by 10-20% within a few months, not by tightening policy but by fixing the specific problems the data points at.

Where to Start

Which Tools Can Do This?

Dedicated returns platforms such as Loop and ReturnGO handle the customer-facing side: portals, labels, exchange flows and refund triggers. Power Automate (part of Microsoft 365), Make and Zapier connect that returns data to your inventory, finance and reporting systems. For the analysis layer, AI models such as OpenAI, Claude and Gemini plug into most platforms for reason-code classification and pattern detection, and custom API integrations cover more complex setups.

If you'd rather have someone design, build and manage the whole thing, that's what Fulcrum Three does. We'll work out what returns actually cost your business and where the biggest savings sit.