The returns management workflow, and what fires each step

What a returns management workflow is

A returns management workflow is the documented sequence an e-commerce brand runs after checkout when a customer sends something back: request, authorization, transport, receipt, inspection, disposition, refund and inventory update, with a named owner and a completion test on every step. Every guide gives you that list and calls it the workflow.

The steps aren't the workflow. The workflow is what fires each step, what proves it finished and what happens when it forks, and a return that's going wrong is rarely inside a step. It sits between two of them with nothing waiting for it.

When we point Keeyu at a brand's returns and ask what is stuck, the answer looks like this: six returns, received five to fourteen days earlier, inventory never put back, refunds never released. On paper, every one had completed.

Under ASCM's SCOR Digital Standard, Return is a top level supply chain process with inputs and outputs like any other. Most brands still run it out of a shared inbox, on a channel now worth 16.9% of US retail sales.

The seven steps, and what starts each one

1. Request: the customer asks to send an item back, against a published policy and a window. Fired by the customer, through whatever channel they picked.
2. Authorization: the return is approved or refused and an RMA is issued. Fired by a person, or by a rule standing in for one.
3. Label and transport: a label is issued and the item starts moving. Fired by a carrier scan, in somebody else's system: USPS prices return labels per piece and per label scanned, so an unused label costs nothing, and nothing bills or alerts when the parcel is never posted.
4. Receipt: the parcel lands at the warehouse and is matched to its RMA. Fired by a receiving scan.
5. Inspection and grading: condition is assessed against a written standard. Fired by a warehouse queue, which is to say by whatever is in front of it.
6. Disposition: the unit is routed: restock, refurbish, liquidate, donate or dispose. Fired by the grade.
7. Settlement: refund, exchange or credit is issued and inventory is reconciled. Fired by the disposition decision, or by a clock, depending on the policy.

Read that back and count. Two of those seven triggers belong to somebody outside the business: the customer's request and the carrier's scan.

A workflow is its exit conditions, not its steps

A list of steps tells you the order of the work. A workflow tells you, for each step, what starts it, who owns it and what test proves it's finished, the three fields that also make an order management process legible. BPMN, the standards body notation for process diagrams, exists because a documented process needs more than an ordered list. Most brands have the list and none of the rest.

• Trigger: the event that starts the step, and whose it is: yours, the customer's or a carrier's.
• Owner: the named role accountable for the step completing, not the department. A step owned by operations is owned by nobody.
• Exit condition: the test that proves the step finished, written so a system can evaluate it. Not "inspect the item" but "this unit has a recorded grade and a disposition decision within 24 hours of receipt".

Some exit conditions carry a statutory deadline. Under the UK's Consumer Contracts Regulations 2013, regulation 34, a trader must reimburse within 14 days of receiving the goods back, or of the consumer's evidence of having sent them, whichever is earlier: a clock that can start in the customer's hands.

The branches nobody draws

The returns workflow diagrams brands actually publish run in a straight line: no decision point, no failure path, no loop back. A workflow with no branches is a list, and these five forks are where the money goes.

1. The parcel with no matching RMA: it sits on the dock as unexpected stock while the returns record still says in transit.
2. The unit that fails inspection: the grade rules out the outcome already promised, so the step routes to a decision, not to a refund.
3. The partial return: two of three items come back, so two settlements, two inventory events and one confused customer-facing status.
4. The exchange with no stock: the replacement is unavailable, so the outbound leg is a new order that can fail on its own terms, while the customer's UK 14 day window to send goods back runs regardless.
5. The label that's never used: the return is open in the portal, the parcel never moved, and because nothing bills or alerts, nothing chases it.

The approval branch isn't taste, either: a California policy more restrictive than a seven day refund must be posted conspicuously, or the seller is liable to a buyer returning within 30 days.

Where the workflow actually stalls

A return spends most of its life between steps, not in them. Every state above is a promise the brand has already broken and can't see: not late, not failed, not flagged, because no system believes the return is its own. The portal says approved. The carrier has no scan to report. The warehouse has stock it didn't expect. Finance has no refund request. Each system is telling the truth about its own leg, and the return is broken between them.

The first thing that notices is the customer, by opening a ticket. By then the workflow has failed, and a reply moves no parcel and releases no money: a helpdesk is a system for replying about problems, not resolving them. One in five orders hits an operational break after checkout, and in returns those breaks hide longest.

The scale isn't in dispute. NRF puts 19.3% of online sales returned in 2025, against $849.9 billion of total retail returns, and, a year earlier, 68% of retailers told NRF and Happy Returns they were prioritizing a returns upgrade within six months. Those are reported rates, not targets.

What to automate, and what to leave to a person

Automation isn't a strategy, it's a trigger: it fires the next step when the previous one completes, so it compounds a healthy workflow and does nothing for a stalled one. The steps that hurt never started, which is the limit to understand before buying customer service automation.

Automate outright the deterministic steps with an unambiguous exit condition: label issue, status notification, inventory update on a recorded grade, refund release once disposition is set. A refund that can't go back by the original payment method is due within seven working days, and the replacement leg of an exchange is a shipment with its own advertised window.

Automate with a threshold on approvals inside policy, below a value ceiling, on accounts with no flags. Above it, route to a person: NRF puts 9% of returns in the fraudulent column, so a blanket auto approve isn't a saving.

Never automate the judgment calls. Goodwill outside policy, a flagged account, a high value unit, anything where the right answer is to bend the rule. Automating a decision you wouldn't put in writing is how a policy becomes an incident.

Automating the steps is the easy half, and everyone does it. The hard half is noticing that a step didn't happen.

Running the workflow as an operation, not a diagram

At most brands nothing fires until the customer chases it, and then it lands in the helpdesk. That's the old category, and it structurally can't resolve this: a helpdesk answers the customer, it doesn't match the parcel on the dock or release the refund. The new category is proactive e-commerce operations: detect the exit condition that was never met, decide against the promise the brand made, act, and tell the customer before they ask.

Keeyu isn't a returns portal, not a returns management platform, not a workflow automation tool, not a WMS, not a 3PL, not a carrier and not a helpdesk. We don't print labels, receive parcels or grade stock. We work on the step that didn't complete. Every order is a promise, and a return is the last chance to keep one.

Close the gaps between the steps

A diagram can't tell you that step four never fired on this return. Keeyu watches every return against the promise made at checkout, catches the ones whose exit condition was never met, the unmatched parcel on the dock, the refund approved in one system and never released in another, and acts on them before a ticket arrives. That's proactive e-commerce operations in practice. Book a Keeyu demo and bring your stuck returns.

• ASCM. SCOR Digital Standard. Return as one of the top level supply chain processes, split into Source Return and Deliver Return with categories for defective, MRO and excess product.
• US Census Bureau. Quarterly Retail E-Commerce Sales. E-commerce at 16.9% of total US retail sales in the first quarter of 2026.
• USPS. Business Return Services. Return label pricing is charged per piece and per label scanned, so a label that is never used costs the merchant nothing.
• Object Management Group. Business Process Model and Notation 2.0. The standards body notation for business process diagrams.
• UK legislation. Consumer Contracts Regulations 2013, regulation 34. Reimbursement within 14 days of the trader receiving the goods back, or of the consumer supplying evidence of having sent them, whichever is earlier.
• UK legislation. Consumer Contracts Regulations 2013, regulation 35. The consumer's own 14 day window to send the goods back, and who bears the direct cost.
• California Legislative Information. California Civil Code section 1723. A policy more restrictive than a full cash refund within seven days must be posted conspicuously, or the seller is liable to a buyer returning within 30 days.
• Cornell Legal Information Institute. 16 CFR 435.2. A prompt refund is due within seven working days where it cannot be made by the original payment method.
• Federal Trade Commission, business guide to the Mail, Internet, or Telephone Order Merchandise Rule. The advertised shipping window, or 30 days where none was advertised, which applies to the replacement leg of an exchange.
• National Retail Federation. 2025 Retail Returns Landscape. An estimated 19.3% of online sales returned in 2025, and 9% of returns reported as fraudulent.
• National Retail Federation. Consumers expected to return nearly $850 billion in merchandise in 2025. The $849.9 billion total returns figure.
• National Retail Federation and Happy Returns. 2024 retail returns total $890 billion. 68% of retailers prioritizing an upgrade to their returns capability within six months.