As of November 5, 2026, estimates compiled for the coming peak put worldwide online holiday returns above $181 billion in the 2025 season, roughly 14% of purchases and up about 10% year on year, while buy-now-pay-later spending is expected to reach a record $21.3 billion this season. The two numbers describe one behaviour: shoppers buy more variants than they keep. That turns review analysis into a pre-season task rather than a January post-mortem, because most avoidable returns are decided on the product page, months before the parcel comes back.
Most returns are a page problem
The return reason codes tell a narrow story: wrong size, not as described, changed mind. Only the third is uncontrollable. The first two are page problems, and both are visible in reviews long before they appear in a returns report. A cluster of complaints that a jacket runs small is not an opinion, it is a specification error with a cost attached.
Treating it as a review problem changes who acts and when. Instead of the warehouse absorbing returns in January, the content team edits the size chart in October, and the merchandising team checks whether the photo shows the true colour. Both are cheap, and both need to happen before the traffic arrives.
Discount depth encourages variant buying
Deep discounts push shoppers to buy two sizes or two colours and return one. That is rational behaviour, and it is profitable only if the returned unit resells at a price that covers handling. When discounts reach 30%, the margin on the resold unit may not cover the round trip, which is why variant buying needs to be modelled rather than welcomed.
What review analysis should surface
Four patterns are worth extracting weekly, on the SKUs that will carry the season. Repeated fit complaints, photo-versus-reality complaints, unanswered questions that appear inside reviews, and complaint clusters concentrated in one variant of a listing. The first three are page edits; the fourth is often a supplier issue hiding behind a shared page.
| Return reason | Review signal | Pre-season action |
|---|---|---|
| Wrong size | Repeated runs small or runs large phrasing | Publish measured garment dimensions |
| Not as described | Colour or material differs from photo | Add unedited photos and material detail |
| Incompatible | Questions about fit with other products | Add a compatibility table |
| Changed mind | Price or expectation comments | Clarify value and delivery terms |
Work the table from the top. Size complaints are the cheapest to fix and usually the largest single category. Description complaints cost more because they often require new photography, so they need to be scheduled earlier. Compatibility complaints matter most in electronics and home, where a single wrong assumption produces both a return and a negative review.
Sentiment is not the metric
A rising star rating is a poor guide to return risk, because happy customers write more than returning ones. What matters is the volume of a specific complaint, not the average score. A product with 4.6 stars and forty mentions of a tight collar will return more than one with 4.2 stars and no recurring theme.
BNPL raises the basket and the exposure
Buy-now-pay-later lifts average order value and, with it, the size of the parcel that may come back. The mechanism is commitment: paying in four instalments feels smaller at checkout than paying once, so shoppers add a variant. The answer is not to switch the option off, since customers expect it on higher-ticket gifts, but to state delivery and return terms where the option appears.
There is a measurement task attached. Return rate by payment method is rarely reported next to average order value, which means the BNPL uplift is measured on the way in and never on the way back. Splitting the two makes the trade-off visible, and it usually shows that BNPL is still worth it, just less dramatic than the checkout number suggests.
A pre-peak checklist
The window for page fixes closes about two weeks before the peak week, because marketplaces need time to re-index and ad systems need time to re-learn. After that, the leverage moves to the returns process: clearer instructions, faster refunds, packaging that survives a round trip.
FAQ
Why treat returns as a review problem rather than a logistics one?
Most avoidable returns start as a mismatch between the page and the product: size, colour, material, compatibility. Those mismatches appear in reviews months earlier as complaints, so reading them turns a reverse-logistics cost into an editing task.
What signals in reviews actually predict returns?
Repeated phrasing about fit and scale, complaints that a photo looked different, and questions inside reviews that the page never answers. A cluster across variants usually points to a page-level error.
Does buy-now-pay-later make returns worse?
It raises average order value, and larger baskets tend to return more. The mitigation is to state delivery and return terms where the option appears, so the decision is informed rather than removed.
How late is too late to fix product pages?
About two weeks before peak week, because marketplaces need time to re-index. After that, put the effort into the returns process: clearer instructions, faster refunds, better packaging.
