Livestream returns and refund rates in China: how to protect margin on Douyin and Tmall
Here's the short version: livestream returns in China are a margin problem you solve before the stream, not after it. Protect margin on Douyin and Tmall by pricing returns into the offer, cutting the impulse mismatches that cause them (sizing, shade, expectations), controlling who you stream with, and treating the refund flow as policy design rather than leakage. Chasing buyers to keep them from returning is the expensive way to lose.
Every foreign brand I've seen go live in China has the same quiet moment. The GMV screenshot looks great, then finance nets out the refunds and the room goes silent. Livestream selling works by compressing consideration into a few minutes of host energy and a countdown. That is exactly why it returns more than a quiet product page does. You bought conversion with urgency, and urgency has a return policy.
What the rules actually give the buyer
China's Consumer Protection Law provides a seven-day no-reason return on most goods bought online, and that applies to goods bought through livestreams as well. Platforms layer their own rules on top, and those have moved. In April 2025, Taobao, JD.com, Pinduoduo, Douyin and Kuaishou all announced changes to their after-sales policies, widely reported as scaling back the automatic "refund only" (tuikuan jin) option that let buyers get money back without sending the item. Reports at the time said individual sellers can still agree to a no-return refund case by case. Rules like this get revised, so check the current seller-centre terms on each platform before you build a model on them.
The point for margin: the legal right is fixed, the platform mechanics are not, and your plan has to survive both.
Where livestream returns come from
I won't quote you a single return rate, because I haven't found a reliable, current Douyin-wide figure. Douyin itself publicly disputed a viral claim of a 90 percent return rate in jewellery, saying its category rate was at the industry average, while sellers kept reporting returns as a persistent problem. That is the honest state of the data: returns are high and costly, and nobody publishes a clean number. Build from your own category data instead.
The causes are more useful than the averages:
| Cause | How it shows up | Fix before you go live |
|---|---|---|
| Impulse mismatch | Buyer bought the host's energy, not the product | Show the unglamorous parts: size on a real body, shade in daylight, actual dimensions |
| Sizing and fit | Apparel and footwear return in volume | Honest size chart pinned in the room, model height and size stated aloud |
| Oversold promises | The host claimed more than the label says | Approved script, claims checked against what you can legally say |
| Return-shipping insurance abuse | Empty parcels or swapped items sent back | Inspect on receipt, document, and use platform dispute channels |
| Fulfilment slips | Late or wrong shipment, then a cancel | Stock and warehouse locked before the stream, not after |
Return-shipping insurance is a price, not a perk
Most Chinese platforms let sellers fund return-shipping insurance (yunfei xian), which pays the buyer's return postage. Buyers expect it, and it reduces the friction that otherwise stops a purchase. Research on livestream supply chains suggests it can cut handling costs in high-return settings, though its value depends on how generous the subsidy is. It is also a known target for fraud. Model it as a per-order cost in your unit economics from day one, the same way you model a platform commission.
Controlling margin without making returns painful
The instinct from outside China is to make returns hard. I've covered why that backfires: after-sales in China is public and searchable, and a clumsy refund becomes a post that sits next to your product for months. So the margin work happens upstream:
- Price the return in. Take a conservative return rate for your category, apply it to every live SKU, and decide whether the stream still clears your floor. If it doesn't, the discount is too deep.
- Pick SKUs for live deliberately. Low-variance products (consumables, standard sizes) return less than high-variance ones. Put the fit-sensitive hero items in a controlled format.
- Choose the host for fit, not follower count. A big room selling on hype can return more than a smaller one that explains the product. Read what foreign brands get wrong on Douyin live before booking anyone.
- Make the return flow resaleable. Agree packaging, inspection and restock steps with your warehouse so returned stock comes back as sellable inventory, not write-off.
- Watch the store score. Returns and complaints feed back into platform visibility, so a stream that boosts GMV and damages the rating can cost more than it earned.
Douyin versus Tmall: different problems
On Douyin the return risk is mostly impulse and host claims. On Tmall, livestreams sit inside a store where the buyer already compares reviews and prices, so you get fewer impulse returns but more expectation-gap returns driven by the detail page and pricing history. Same discipline, different leaks. Track returns per channel and per stream, never as one blended number.
Bottom line
Livestream GMV is a vanity number until you net out returns, shipping insurance and the cost of handling parcels coming back. Price them in before the stream, choose products and hosts that reduce mismatch, and keep the refund experience clean. Brands that do this keep the margin; brands that fight the returns lose the buyers and the margin.
If you're planning a China livestream launch and want a second read on the unit economics, reach out.