Use Cases· 5 min read

How Online Stores Cut Returns With Pre-Purchase Answers

Strip out the faulty items and most returns trace back to a question nobody answered before checkout. Using an AI chatbot to resolve sizing, scale, and compatibility doubts while the shopper is still deciding.


Look at your returns reasons for a month. Strip out the genuinely faulty items and the changed-my-minds, and what's left is almost always the same thing wearing different clothes: the customer didn't know something before they bought it.

It didn't fit. The colour was different. It was smaller than it looked. It didn't work with the thing they already owned. The fabric wasn't what they pictured.

Every one of those is a question the shopper had on the product page, at eleven at night, with nobody to ask.

Returns are a support failure with a shipping label

Retailers tend to treat returns as a logistics problem. Reverse shipping, restocking, refurbishment. That's the cost, but it isn't the cause. The cause is unanswered pre-purchase uncertainty, and it's expensive twice: once in the return, and once in the sales you never made from the shoppers who had the same doubt and simply left.

That second group is invisible. You count the returns. You don't count the people who wanted to know whether the jacket runs small, couldn't find out, and closed the tab.

The four questions that drive most of it

Fit and sizing. The biggest single category in apparel and footwear. "I'm normally a medium in [brand], what should I order here?" A bot working from your size guide and any published fit notes answers that in seconds. If your product pages include fit feedback (runs small, generous cut, true to size), it can use that too.

Scale. Underrated and common in homeware, furniture, art, and accessories. Photos flatten scale. A shopper genuinely can't tell whether a vase is 15cm or 40cm without reading the spec, and most don't. A bot that says "it's 32cm tall, roughly the height of a wine bottle and a half" prevents a return that was otherwise certain.

Compatibility. Electronics, parts, accessories, refills. Will this fit my model? Is this the right cable? Does it work with the older version? These are factual, documented, and infuriating to look up, and getting them wrong guarantees a return.

Material and finish. How does it feel, is it stiff, is it see-through, is the wood real or veneer, will the colour match the photo. Your product copy probably answers half of this. A bot can surface the half that's buried.

What the conversation looks like

Shopper: Is this jumper thick enough for winter or is it more of an autumn thing? Bot: It's a midweight lambswool at 12-gauge. Warm enough for autumn and mild winter days, but most customers layer it under a coat in properly cold weather. If you want something heavier, the [chunky knit] is a 7-gauge and noticeably thicker.

That shopper either buys with accurate expectations, or buys the other one, or doesn't buy. All three outcomes are better than a return.

Note what the bot did there: it was honest about a limitation and offered an alternative. That's not a lost sale. That's the mechanic that makes the whole thing work.

Honesty is the feature

There's an instinct to configure a shopping assistant to be relentlessly positive. Resist it.

A bot that tells every shopper the item is perfect for them will convert marginally better and return catastrophically worse. Configure it to answer accurately from your content, to say when something isn't suitable, and to suggest alternatives when it isn't.

The maths generally favours this. A £60 item returned costs you outbound shipping, return shipping, handling, and often a resale at a discount. Preventing one return is worth several marginal conversions.

It fixes your content as a side effect

Here's the useful second-order effect. Once a bot is answering product questions, you can see what it's being asked, and, more importantly, what it couldn't answer.

That list is a content roadmap. If forty people asked whether a product is machine washable and the bot had to say it didn't know, that's a line missing from your product template. Fix it once and it improves the product page, the bot, and your search visibility simultaneously.

Run this for a season and the top unanswered question is often embarrassingly basic. Dimensions, usually. A field that exists in the PIM and is not rendering on the storefront. One template fix, and the bot stops being asked.

Where it also earns its keep

Beyond returns, the same bot handles the routine load: where's my order, what's your returns window, do you ship to [country], how long does delivery take, do you do gift wrapping. These are the tickets that consume your support team's day and require no judgement at all.

And it handles them in the shopper's own language, which for any store selling internationally is the difference between a conversion and a bounce.

What you'll need to configure

Point it at your product pages, size guides, spec sheets, and policy pages. Upload anything that lives as a PDF. Care instructions, compatibility charts, assembly guides.

Then set one instruction that matters more than the rest: answer only from the product content, and if the answer isn't there, say so and offer to check rather than guessing. A confident wrong answer about compatibility creates exactly the return you were trying to prevent.

Checking the payoff

Return rate by reason code, before and after, for the categories where the bot is most active. That's the number that pays for it.

Then look at conversion on the product pages where shoppers engage with the bot versus those where they don't. In most stores the engaged group converts higher, because the doubt that was quietly killing the purchase got resolved.

Frequently asked questions

Can a chatbot actually reduce returns?
Indirectly but meaningfully. A large share of returns are 'didn't fit' or 'not what I expected'. Both of which trace back to a question the shopper had before buying and couldn't get answered. Answering it at the product page removes the mismatch before it becomes a parcel coming back.
What product questions cause the most returns?
Sizing and fit, material and texture, compatibility with something the customer already owns, and scale. Whether an item is bigger or smaller than the photos suggest. All four are answerable from content you already have.
Won't answering honestly lose me sales?
It loses you the sales that were going to come back anyway, minus the shipping both ways and the restocking labour. A slightly lower conversion rate with a much lower return rate is usually the better business.
Does it need to connect to my product catalogue?
It works from whatever content you point it at. Product pages, size guides, spec sheets, policy pages. The richer those are, the better the answers, which is why the exercise often improves the underlying content too.

Keep reading

How Online Stores Cut Returns With Pre-Purchase Answers · SpideyChat