The 9.9 Returns Wave: Build Your RMA Workflow Before Orders Ship

If you set up one thing before your 9.9 orders leave the warehouse, make it a returns workflow: one RMA record per returned item, created the moment the customer raises the request, carrying order ID, reason code, courier tracking, refund status and a restock decision. That single record is the difference between a controlled mid-September returns wave and three weeks of WhatsApp threads, unexplained stock variance and a 10.10 buy based on numbers you cannot trust. It takes an afternoon to build. After the parcels ship, the same work takes a fortnight and you have already lost the data.

Why does the returns wave hit two weeks after the sale, not two days?

Most Singapore SMEs brace for the wrong date. Orders placed on 9 September fulfil over 10–13 September, deliver over 12–16 September, and only then does the return window open. Shopee and Lazada return windows typically run to a week or more from delivery, and customers use nearly all of it. TikTok Shop buyers tend to raise issues faster but escalate harder. Direct web store customers are slowest of all because they have to email you.

The practical effect: your returns peak lands roughly 18–30 September. That is the exact fortnight when your finance person is closing Q3, your ops lead is placing the China restock for 11.11, and your customer service is already fielding 10.10 pre-sale questions. Returns are not a small problem in that window — they are a scheduling collision. Every hour spent hunting for a parcel that a customer says they posted back is an hour not spent on the next peak.

What does an unmanaged returns wave actually cost?

Four costs, in rough order of size for a mid-sized Singapore seller:

• Stock you own but cannot sell. Returned units sitting unbooked in a corner are invisible to your reorder logic. You reorder stock you already have, or you mark an SKU out of stock during 10.10 while forty units sit in a bin.
• Margin you never see. Marketplace return handling fees, reverse logistics charges and refunded commission all land in the payout statement, not the order record. A 9.9 that looked profitable on gross merchandise value can be six to twelve points thinner after the September returns settle.
• Support time. A return handled by conversation rather than by record costs 15–25 minutes of staff time. Two hundred returns is a week of somebody's month.
• Rating damage. Slow refunds affect marketplace seller ratings, which affect 11.11 visibility. The penalty arrives at the worst possible moment.

What should you set up before the orders ship?

Keep it deliberately small. The goal is a record, not a system.

1. One RMA record per returned item. Not per order — per item. A three-item order with one faulty unit is one RMA. Fields: RMA number, order ID, channel, SKU, quantity, reason code, date raised, courier and tracking, date received, condition on inspection, restock decision, refund amount, refund date.
2. Five reason codes. No more. Wrong item sent, damaged in transit, faulty on arrival, not as described, change of mind. Fifteen codes guarantee nobody uses them consistently; five get used. Everything else goes in a free-text note.
3. A named owner and a stated turnaround. One person owns the RMA queue daily until 30 September. Publish an internal target — inspect within 24 hours of receipt, refund within 48 — and measure against it.
4. A restock decision gate. Every received item gets marked resellable, refurbish, or write-off, and resellable units are booked back into stock the same day. This is the single step most SMEs skip, and it is the one that corrupts your 10.10 stock position.
5. A credit note trigger. Whenever a refund is issued against a tax invoice you have already raised, your finance workflow needs a corresponding credit note with the original invoice referenced. Set the trigger now as a workflow step so the paperwork is generated as returns happen rather than reconstructed at Q3 close. Your accountant decides the treatment; your job is making sure the event is captured. If you are moving to InvoiceNow, credit notes are part of the document set — worth confirming your setup handles them before the volume arrives.

How do you automate this without buying new software?

You almost certainly do not need a returns platform for one peak season. What you need is one authoritative table and three flows into it.

Intake. Replace the reply-to-this-WhatsApp approach with a short form — order number, SKU, reason, photo upload — that writes directly to your RMA table and returns an RMA number to the customer. A form takes an hour to build and removes the entire category of returns that exist only in someone's chat history.

Status comms. Three automated WhatsApp or email messages, keyed off status changes in the table: RMA received, here is your return label and instructions; we have received your parcel and are inspecting; refund issued, expect it in X working days. Those three messages kill roughly 70% of your inbound where-is-my-refund traffic.

Marketplace sync. Marketplace-initiated returns will not flow into your table by themselves. Assign a daily 15-minute check to pull open return requests from each seller centre into the same table, so you have a single view across Shopee, Lazada, TikTok Shop and your own store. Manual is fine at this volume — a consistent daily manual pull beats an integration you will not finish before Thursday.

What should the returns data tell you before 10.10?

By 30 September you should be able to answer four questions from the table alone, without opening a spreadsheet reconciliation:

• Return rate by SKU. Anything materially above your baseline is either a listing problem or a product problem. Fix the listing before 10.10; pull the SKU from the 11.11 buy if it is the product.
• Return rate by channel. A channel returning at double the rate of the others is telling you something about its buyer profile or your listing there.
• Reason code mix. Heavy on wrong item sent means a picking problem that will get worse at 11.11 volume. Heavy on not as described means a content fix that costs nothing.
• True landed margin. Gross revenue, minus platform fees, minus returns, minus reverse logistics. This is the number your 11.11 stock commitment should be based on — not 9.9 gross.

Does this apply if you do not sell on marketplaces?

The structure holds for any Singapore SME that ships goods or delivers scheduled work. A distributor handles credit returns from trade customers. An F&B group handles supplier rejections and short deliveries. A servicing company handles callbacks and warranty rework — which is functionally the same record with a technician instead of a courier. In each case the failure mode is identical: the event exists in conversation rather than in a record, so nobody can count it, and the same problem recurs next quarter with no evidence trail to argue from.

Frequently asked questions

When will the 9.9 returns wave actually peak?

Expect the bulk between 18 and 30 September, with marketplace returns front-loaded and direct-store returns trailing into early October. Plan owner coverage through 5 October rather than assuming it ends with the month.

Should returned stock be restocked before 10.10?

Resellable units should be inspected and booked back in the same day they are received, not batched to a month-end exercise. Unbooked returns are the most common reason a 10.10 stock position is wrong, and they cause both phantom stockouts and duplicate reordering.

Is it too late to set this up if orders have already shipped?

No. The intake form and the RMA table are still worth building on 9 or 10 September, because almost no returns will have been raised yet. What you lose by waiting is the label and messaging templates going out with the first parcels — recoverable. What you cannot recover is returns that were only ever handled in a chat thread.

Digital Perpetual builds returns and reverse-logistics workflows for Singapore SMEs across marketplace, direct and B2B channels. If your 9.9 returns are already arriving faster than you can record them, that is the conversation to have this week — not in November.