
A customer returns a pair of shoes in store after buying them online. The counter team records the return, the warehouse later receives the parcel, and finance sees a credit in the ERP. The payment provider still shows a pending refund. Each team has a record of the event, yet nobody can tell the customer when the money is likely to arrive.
For Australian and New Zealand retailers, the investment decision is where to automate refund reconciliation and which exceptions people should review. A returns portal can capture what the customer wants to send back. Reconciliation establishes whether the goods, credit and money reached the right places. That distinction matters when stores, fulfilment teams and finance work in different systems.
Start by following a handful of transactions through the whole process. Include an ordinary return, a refund without a physical return, and a case the finance team had to correct manually. Trace each from the original order to the payment provider and the stock or ledger record. The evidence from those examples will tell you where automation is likely to help.
Key Takeaways
- Track receipt, stock disposition, credit and payment as separate events.
- Match partial refunds to original lines and provider transactions.
- Check the original payment outcome before retrying an uncertain refund.
Separate the return, refund and credit
A return records the goods and the intended outcome. It can cover one line from an order, an exchange, or an item that is no longer fit to sell. A refund records money being sent back. A credit memo or equivalent accounting entry records the financial adjustment against the sale. A stock adjustment records where the item ended up and whether it can be sold again.
Keep those events separate in the design. A scanned parcel confirms receipt at a location. The inspection outcome may come later. A customer service gesture can trigger a refund with no parcel at all. A warehouse can receive an item and move it into quarantine while the business resolves a discrepancy in the return.
Build a short state map with operations and finance. For every stage, name who can advance it and what evidence is required. For example, a counter receipt may move the return to received, an inspection may determine its disposition, and the payment provider can confirm the payment refund. Give staff one place to see those states and the references that connect them.
If the customer still needs a way to request and track a return, start with the buying journey and authorisation process. OSE’s returns portal and RMA guide covers that earlier part of the workflow.
Decide which system can move money

Pick one authorised starting point for each refund type. An ordinary card refund might start in the commerce platform, with finance receiving the accounting entry and payment reference. A manual bank repayment might start with finance and need to be recorded in commerce afterwards. An account credit may follow a different process again. Document who can act in each case and how the remaining systems learn about the decision.
Adobe distinguishes online, offline and account credit workflows in its credit memo documentation. Test the actual payment integration and configuration before treating a saved memo as evidence that money moved. An offline adjustment and an online processor transaction need different reconciliation evidence.
Refund status also matters after submission. Shopify’s guidance on failed and pending refunds distinguishes payment-provider results from the record visible to staff. Design the recovery procedure for your actual provider. When the outcome is uncertain, check the existing transaction before attempting another payment action.
Give customer service labels that reflect the evidence. “Refund requested” describes a different event from “refund processed by the provider”. Customer receipt can depend on the payment method and bank. Agree what wording your support team can substantiate and where staff should look when a customer follows up.
Tie partial refunds back to the original lines
Use the original order reference, stable line identifiers, refund request reference, payment transaction and credit reference to connect records. Store the amounts and quantities as separate values. When possible, preserve the reason and the user or service responsible for each decision. This makes it easier to answer both “what did we pay?” and “which goods did this relate to?”
For an illustrative NZ retailer, suppose a customer bought two items on the same order and returned only one. The purchase included an order-level promotion. The refund calculation needs the approved allocation for the returned line, any freight adjustment, and the original currency. Finance should confirm which values flow into the credit record. Simply taking the item’s current shelf price may produce a different amount.
Ask what happens if the second item comes back later. The second refund should account for amounts already returned and use the remaining eligible quantities. A store return and a support refund can also arrive close together. Test the concurrent case so staff can see and resolve a conflict before both payment instructions are accepted.
Mixed tenders require their own acceptance cases. If an order used a card and a gift card, establish the refund allocation supported by the deployed configuration and approved policy. Staff need to see the split and the status of each component. Avoid collapsing different tenders into a single amount that cannot be matched to the underlying transactions.
Release stock only when the disposition is known

Agree which state makes a returned product available for sale. An unopened item may be ready after inspection. Damaged stock may need repair, disposal or a supplier claim. Each path can affect the warehouse record without changing the amount refunded. Keep the physical destination and sellable quantity explicit.
Adobe’s guide to issuing a credit memo describes return-to-stock options and the effect of configuration. Verify how those settings interact with warehouse receipts and inventory synchronisation in your implementation. If two systems independently add the same item back, the storefront can advertise stock that exists only once.
For returns accepted at a branch, decide whether the item belongs to that branch, goes to a distribution centre or waits for inspection. Record physical receipt separately from the movement to its final location. The team checking availability should know which stock is actually eligible to be sold from each location. OSE’s inventory accuracy guide covers the wider store, warehouse and ERP responsibilities.
Test a refund with no stock change and a received return with no immediate refund. These two cases expose integrations that treat inventory and payment as one event. Include a return to a different location from the original shipment and verify that the physical count and the customer-facing stock figure match the agreed outcome.
Reconcile the payment event and the payout

Finance needs a way to match the customer refund to its provider transaction and, where relevant, to settlement. The order screen is useful for customer service, but the payout can group many payments, refunds and other adjustments. Design the daily reconciliation around the identifiers available from the actual provider.
For example, Shopify documents the transaction breakdown and payout export in its payout details guidance. Decide how provider records will be imported or checked and who resolves discrepancies. A pending provider result and a confirmed refund missing from an expected payout are different exceptions.
Keep customer refund amounts, processor fees and settlement adjustments in separate fields. Use a consistent currency and document which conversion information is available when order and settlement currencies differ. Finance should approve the mapping using real transaction examples, including refunds that appear in a later payout period than the sale.
Agree a practical close process. List unresolved refund events, unmatched credits and differences between provider transactions and settlement. Assign someone to investigate each class. A daily report is useful only if an owner can trace a difference to the original order and resolve it without re-entering the refund.
Make retries safe and exceptions visible
Integration messages can arrive twice or in a different order from the events they describe. Give each refund instruction a stable identifier and define how retries find the same request. Keep the evidence from the provider even if the ERP update fails. Repairing a missing credit should be possible without repeating the payment.
Build a small exception queue with staff who will use it. Useful entries include “provider pending”, “credit missing”, “received without matching order” and “stock destination unresolved”. Show the transaction reference, evidence and next permitted action. Avoid making staff choose between doing nothing and reprocessing the entire return.
For a payment timeout, run a test where the provider has accepted the refund but the integration never receives confirmation. Staff should see that the state is uncertain and know how to check the provider reference. Once the result is known, reconcile the downstream records. OSE’s ERP integration failure guide explains the broader controls for recovery and duplicate processing.
Pilot the process before expanding automation
Choose one channel, payment method and return location for the first pilot. Define what a successful result looks like in commerce, payments, inventory and accounting. Use test transactions with expected values and outcomes agreed by the people who run each system. Include full and partial returns, a delayed payment result and a repeat message.
Use live pilot evidence to decide what to automate next. Track time to reconcile, unmatched credits, manual interventions and duplicate payment incidents. Compare similar cases and count the exceptions alongside the successful transactions. That helps separate improvements caused by better data from differences in the mix of returns.
You may find the business needs a clearer workflow and an exception report before it needs a new platform. If connector gaps are the issue, scope those against the cases that produce the most rework. Bring an anonymised order, return, provider transaction and credit record to an OSE integration discovery discussion. Those examples help define which systems should act and what evidence the implementation must preserve.
Frequently asked questions
Use the payment provider’s transaction status to establish what happened to the payment. The credit is evidence of an accounting adjustment. Link the two records and give staff a way to investigate when only one is present.
Base restocking on the physical receipt, inspection outcome and destination under your operating policy. Test that commerce and warehouse settings do not both add the same quantity back. A refund without returned goods needs a separate path.
Check the provider record and confirm the status before retrying. Preserve the original request reference and give staff a clear escalation route. Customer updates should reflect the result the team can verify.
Match it to the original order lines, allocated discounts, currency and prior refunds. Check the expected stock disposition and the actual payment amount independently. Verify the credit and payment references before closing the case.
Start with a routine refund path whose inputs and outcomes are consistent. Test pending responses, retries and partial quantities before expanding. Keep a staffed review route for exceptions where the system cannot safely determine the next action.