Returns management for e-commerce: read the reason, work the exception, stop the next one.
At a luxury fashion e-commerce retailer, leadership said it out loud: “Mitigating returns is our number one opportunity.” The follow-up was manual, the reasons went unread, and the sizing signal never made it back to the product page. We built the AI that reads every return reason across 12,500 client profiles, encodes the team’s disposition rules, and catches the size-and-fit signal before the next box ships.
Official services partner of the platforms defining AI
NVIDIA
Anthropic
Returns management is a board-level priority with an inbox for a system.
From a standing AI program with a luxury fashion e-commerce retailer: the board named returns the number one opportunity, and the actual system was a queue of reason codes nobody had time to read.
“Mitigating returns is our number one opportunity.”
The board has said it plainly. Meanwhile the working system is a returns queue and good intentions — the AI that could read it all hasn’t been built yet, so the priority sits above a manual process.
Every return carries a reason nobody reads
Reason codes, return-form notes, support threads, stylist feedback — thousands of signals about fit, fabric, and expectation. Individually they’re anecdotes; at scale they’re the answer. No one has the hours to read them all.
The 48-hour window slips
The team’s own rule is follow-up within 48 hours of a return. The queue doesn’t know the rule — so the clients with the most at stake wait in line behind routine refunds, and the window closes quietly.
The recovery email re-shows the returned item
Recover-return outreach should offer the next size down or a similar piece in the client’s palette. Re-showing the item she just sent back isn’t recovery — it’s a second miss, from the brand that’s supposed to know her.
The designers don’t conform to standard size scales
A UK 10, a French 34, an Italian 38 — the same fit across three national conventions, and many designers don’t conform to standard size-scale conventions at all. The conversion lives in stylists’ heads, not in the store.
The sale dies at the doorstep
“Clients will just not even take it out of the packaging.” The delivery window is the last chance to secure the sale from the point of delivery — and nothing watches it.
The recover-return window closes in 48 hours.
What it looks like when returns management runs on a queue instead of encoded rules. Your numbers will differ — the first thing we build puts figures on yours.
From an anonymized engagement — a luxury fashion e-commerce retailer with ~12,500 client profiles under management
Read the reasons, encode the rules, move the fit finder signal upstream.
We build the AI that does the reading and the routing, on your book of business and your rules. Three steps, fixed order.
- 01
Read every return reason
The AI reads reason codes, return-form notes, support threads, and stylist feedback across the whole book — 12,500 client profiles, not a weekly sample — and turns them into a reasons report the ops team can actually act on.
- 02
Encode your disposition rules
The 48-hour follow-up. Recover-return outreach that offers the next size down or a similar item in the client’s palette — never the returned item. Prioritization for carts over $10,000 and fine jewelry, average unit retail thresholds, the lot. Your rules, not a vendor’s defaults.
- 03
Catch the fit signal before the next shipment
Kept-size history — “the most consistently purchased and kept size” — feeds size guidance upstream, across designers that don’t share a size scale. The wrong size stops shipping; the return never happens.
Returns management software sorts the queue. It doesn’t decide.
Returns management software is good at the mechanics — RMAs, labels, refund routing, status emails. What it doesn’t do is decide: which client gets a call instead of a credit, what to offer when the size was wrong, when to stop pushing and let a return go. Those are your rules, and they live in your team’s heads.
We’re an AI operating partner — we build and run the machine that reads the reasons and applies your rules, on top of whatever returns software you already have. How to reduce returns in ecommerce isn’t another dashboard; it’s the 48-hour follow-up that actually goes out and the fit signal that reaches the product page before the next order ships.
Mitigating returns is our number one opportunity.
Led by Peter Enestrom.
How to reduce returns in ecommerce, answered straight.
What e-commerce operators and CFOs ask before anything ships.
Everything that happens between “I’m sending it back” and the decision that follows — the label and the refund, yes, but also the reason it came back, the follow-up the client gets, and what you change so the next one doesn’t happen. Most tooling covers the first part. Our AI covers the rest: reading the reasons at scale, applying your disposition rules, and feeding the signal upstream.
By moving the decision upstream. Most ecommerce returns at a fashion retailer trace to size and fit — so the AI reads kept-size history, the most consistently purchased and kept size per client, and the size-scale quirks per designer, then puts that guidance in front of the buyer before the box ships. Prevention beats a smoother refund.
It sorts the queue: RMAs, return labels, refund routing, status updates. It stops at the judgment — which client earns a personal follow-up inside 48 hours, whether to offer the next size down or a similar piece in her palette, which carts over $10,000 get worked first. We encode that judgment as rules the AI applies; the software stays the plumbing.
A fit finder asks the shopper questions and guesses a size; ours doesn’t guess — it reads what the client already kept. When your designers run a UK 10, a French 34, and an Italian 38 for the same fit, the AI maps each client’s kept-size history onto each designer’s real sizing, so the recommendation is evidence, not a quiz. The return never happens because the wrong size never ships.
The ones where the relationship has the most at stake. The team’s rule: prioritization goes to carts over $10,000, with average unit retail thresholds and fine-jewelry priority behind it. The AI scores the queue against those rules every morning, so a high-value return never waits in line behind a routine refund.
Wherever they already live — your returns platform’s reason codes and notes, order and kept-item history, support threads, and stylist feedback. Nothing moves to a new system of record; the AI reads across what you have and writes its output where your team already works.
Start with one workflow.
Tell us where your team loses hours. We will come back with a straight answer on whether AI can help, what it would take, and what it would pay.