Cancellation rate: normalizing order statuses by channel
What cancellation rate divides by, where unpaid orders go, how failed delivery differs from cancellation: a standard status set to compare every channel.
By The DataxLand team· · 5 min read
"What's our cancellation rate this month?" sounds like a question with one answer. In practice, three people in the same company can give three numbers, because each counts "orders" and "cancelled" differently. Shopee, TikTok Shop, your website and your order management software also name statuses differently, so comparing cancellation rates across channels is often comparing apples with oranges.
This post explains why order statuses need to be normalized before you compute any rate, gives a standard status set that works for every channel, and points out three places where counts usually go wrong: unpaid orders, failed deliveries and returns.
Example numbers come from DataxLand's sample data: a simulated scented-candle shop selling on Shopee, TikTok Shop (two of the largest marketplaces in Vietnam and South-East Asia) and its own website, 1 Aug – 29 Oct 2026.
Three different "cancellation rates"
| How it's counted | Formula | Problem |
|---|---|---|
| Cancelled over successful orders | Cancelled orders ÷ non-cancelled orders | The denominator leaves out the cancelled orders themselves, so the rate is inflated |
| Cancelled over all orders created | Cancelled orders ÷ all orders | The standard way; comparable across channels and months |
| Cancellations lumped with returns and failed deliveries | (Cancelled + returned + failed delivery) ÷ all orders | Mixes three problems with different causes and costs |
DataxLand uses the second: cancellation rate = cancelled orders ÷ all orders created, including cancelled ones. In the sample data, the three-month cancellation rate is 7.42%, while the number of valid (non-cancelled) orders is 2,309. That is about 185 cancelled orders out of nearly 2,500 created.
Return rate is computed separately, on valid orders: an order delivered and then sent back is a completely different problem from an order cancelled before delivery (see the post on return rate and its real cost).
A standard status set
Each channel has dozens of status names. To compare them, group them into a small set:
| Standard status | Meaning | Counted as a valid order? |
|---|---|---|
pending | Created, awaiting confirmation or payment | Yes (until cancelled) |
confirmed | Confirmed, awaiting pickup | Yes |
shipping | In transit | Yes |
delivered | Delivered to the buyer | Yes |
completed | Completed, return window closed | Yes |
cancelled | Cancelled before delivery | No |
returned | Delivered, then returned or refunded | Yes, with the refund deducted from net revenue |
failed_delivery | Delivery failed, goods came back | Yes, tracked separately |
The mapping is a two-column table per channel: source status → standard status. An illustration (the source names in an export may
differ by marketplace and over time): "To confirm" → pending, "Shipping" → shipping, "Cancelled" → cancelled,
"Return/Refund" → returned.
Three places counts go wrong
1. Unpaid orders
An order with a prepaid payment method that the buyer hasn't paid yet is in limbo: not cancelled, but not certain to become a real order. There are two reasonable options:
- Map it to
pending: the order counts until the marketplace auto-cancels it for non-payment. Today's order count is slightly high, but corrects itself when the order is cancelled. - Map it to
cancelledif your channel policy treats unpaid orders as not real.
The only mistake is not mapping it at all. The sample data contains an UNPAID status. If you leave it out of the mapping, those
orders have no standard status, and DataxLand blocks the build rather than publishing a wrong order count.
2. Failed delivery is not cancellation
A failed delivery (the buyer refuses or never collects the parcel) has already cost shipping both ways and kept goods on the road for days. Lumping it into cancellations inflates the cancellation rate and hides a real cost. Track the failed-delivery rate separately, by channel and by delivery area.
3. Partially cancelled orders
An order has three products and the seller cancels one line because it's out of stock. Counted per order, the order is still valid; counted per line, one line is cancelled. Decide once and write it down: order cancellation rate is per order, out-of-stock rate is per line.
Reading cancellation rate by cause
| Common cause | Sign in the numbers | What to do |
|---|---|---|
| Buyer changes their mind before delivery | Cancellations concentrated in pending, late-night orders | Shorten confirmation time, send an order confirmation message |
| Out of stock, seller cancels | Cancellations concentrated in a few SKUs | Sync inventory across channels |
| Payment overdue | Cancellations concentrated in e-wallet and bank-transfer orders | Send payment reminders, review the default payment method |
| Fake or test orders | Many cancellations from the same address, same time window | Filter them out before judging a campaign |
Split cancellation rate by channel, by SKU and by payment method. An average of 7% can be 3% on the website and 12% on a channel with many unpaid orders.
Step by step: normalizing statuses for a multi-channel shop
- Export every distinct status value from each channel over the last three months.
- Assign each value to a standard status from the table above; note the values you're unsure about.
- Decide how to handle unpaid orders, failed deliveries and partial cancellations, and write them down as rules.
- Recompute the cancellation rate the second way for the last three months, and compare with the number you've been using.
- Whenever a marketplace adds a new status, add a row to the mapping before computing any numbers.
How DataxLand does this
When it reads your data, Si proposes a status mapping: each pair of channel and source status is assigned a standard status. You edit it on the Statuses tab of the Review proposal screen; yellow cells are where Si is unsure, and every row needs a standard status before approval. After every build, the data checks run again: if any order lacks a valid standard status, the build is blocked and dashboards keep showing the previously published version.
The result: cancellation rate, order count and return rate use the same definition for every channel, so Shopee, TikTok Shop and your website are directly comparable. See the returns control use case and the retail operations template.
FAQ
What cancellation rate is high?
There is no universal threshold across industries and channels. Compare your own rate over time and across channels, after normalizing statuses. A channel with twice the cancellation rate of another is a signal to look at causes.
Do cancelled orders affect profit?
An order cancelled before delivery creates no revenue, but may already have cost ad spend and tied up stock. A failed delivery also costs shipping, so it hits profit more visibly.
Why does the marketplace's order count differ from my report?
Usually because they count differently: the marketplace may count every order created, while your report counts non-cancelled orders; or the day cutoff uses a different time zone. Check the definition before checking the data.
What happens when a marketplace adds a new status?
A new status that isn't in the mapping is caught on the next build. You assign it a standard status, approve again and build.
Read next: Return rate and its real cost · A profit dashboard: 8 metrics · What is a semantic layer?