Returns are a sizing problem before they are a logistics problem
Reverse logistics gets the budget because returns are visible there. The decision that caused most of them happened on the product page, where nobody was measuring.
Returns show up in operations, so they get solved in operations. Faster processing, better reverse logistics, a smoother refund flow.
All of that is worth doing and none of it reduces the number of returns, because the decision that caused most of them was made on the product page, before the order existed.
Where the decision actually happens
A shopper who cannot tell whether something will fit has three options. Buy nothing. Buy one and hope. Or buy two sizes and send one back.
The third is a rational response to uncertainty, and it is a return that was planned at the moment of purchase. Anything that reduces the uncertainty reduces those returns at the source. Anything downstream of checkout is handling them more efficiently.
The distinction matters commercially because the two have different ceilings. Better reverse logistics makes each return cheaper. Better fit confidence means the return does not happen, which also keeps the item sellable and skips the inspection, the restock and the refund entirely.
Why size charts do not close the gap
Every product page already has a size chart, and returns persist anyway, for reasons that are worth being precise about.
A chart gives garment measurements. The shopper needs to know how it will sit on them, which is a different question requiring their own measurements, which most people do not have to hand and will not stop to take.
Sizing is inconsistent between brands and often within them, so prior experience transfers badly. A shopper who is one size in one label has learned, correctly, not to trust that anywhere else.
And fit preference is not a measurement at all. Two people with identical dimensions want different things from the same garment, and no chart encodes that.
What raises confidence enough to change the order
The useful bar is not accuracy in the abstract. It is whether the shopper is confident enough to order one size instead of two.
Seeing it on a body shaped like theirs does more than a number, because it answers the actual question — how it sits — rather than a proxy for it. It works on the product page where the decision is, not in a separate tool the shopper has to go and find.
And it should be honest about uncertainty. A tool that renders every garment as a confident perfect fit teaches shoppers to ignore it after the first mismatch. Saying it is between sizes is more useful than picking one.
The number worth watching
Most teams track return rate, which moves slowly and is confounded by everything. Two better ones.
Multi-size orders: how often does one customer buy the same item in two sizes in a single order? That is planned-return behaviour, measured directly, and it is the number that should fall first.
Return reason, specifically “did not fit”, as a share of all returns rather than a raw count. If total returns drop but the fit share is unchanged, something else moved and it was not the sizing work.
Both are already in your order data. Neither needs new instrumentation, and knowing them before you buy anything tells you whether this is your problem at all — because if fit is not a meaningful share of your returns, a try-on tool will not help you and the honest answer is to look elsewhere.
We built the virtual try-on platform for the first case. The demo page shows how it sits on a product page and where it declines to guess, and there is a live version to try.
If the product pages themselves are the constraint — catalogue structure, variant handling, page speed — that is a storefront question first, and worth separating from the fit question before choosing what to build.