Refund Rate tells you what share of your ordered units came back as refunds. It's a useful early signal for listing quality, sizing or fit issues, and product defects. However, it's built differently from the refund figures in your Amazon financial reports, so it's worth understanding what it does and doesn't measure.
How Refund Rate is calculated
Refund Rate = Units Refunded ÷ Units Ordered
Units Refunded comes from Amazon's Data Kiosk (Sales & Traffic dataset). This is a different source than the Finance API we use for Accounting.
Units Ordered comes from the All Orders Report — the same source behind Product Sales, with cancellations already removed.
Refund Rate is available for Amazon Seller Central accounts only. It is not currently available for Vendor Central or Walmart.
Walmart and Vendor Central accounts show 0% rather than a blank, matching how Return Rate already behaves for those accounts. As with early date ranges, 0% there means the data isn't available, not that no refunds occurred.
What counts as a refund
A refund is a financial event: money returned to the customer. It does not require the customer to send anything back. Refund Rate includes:
Standard refunds where the item was physically returned
Concession and goodwill refunds where nothing was returned (damaged in transit, arrived late, customer service credits)
Because concession refunds are included, Refund Rate usually runs higher than Return Rate.
Timing: refunds and orders are counted in the same date range
This is the most important thing to understand about the metric. Refunds are counted on the date the refund occurs, and orders are counted on the date the order is placed. Both are pulled from the date range you've selected, but a refund processed today may relate to an order placed weeks earlier.
That means Refund Rate answers "how much refund activity are we absorbing during this period?" rather than "of the units we sold in this period, how many came back?" Those are different questions, and they diverge whenever order volume moves sharply.
Example: you sell 10,000 units in November and 3,000 in December. Some refunds against those November orders may process in December. December's Refund Rate divides a large refund count by a small order count, so it reads high — even if nothing about product quality changed. The reverse happens during a sales spike, where the rate reads artificially low.
Refund Rate is most reliable when you look at it over a longer window (28 days or more) and when order volume is relatively steady. Treat sharp month-over-month swings as a signal to check your order volume first, before concluding anything about product quality.
Refund Rate vs. Return Rate
Metric | Source | Definition | Attribution |
Refund Rate | Data Kiosk (Sales & Traffic) ÷ All Orders Report | Units refunded as a share of units ordered. Includes refunds with no physical return. | Refunds on the day the refund occurs; orders on the day the order is placed. |
Return Rate | Returns Report ÷ All Orders Report | Units physically returned as a share of units ordered. | Returns on the day of return; orders on the day the order is placed. |
Refunds (Accounting) | Finance API | Refunded dollars as they appear in your Amazon settlements. Used for P&L reporting. | Settlement timing, which may continue to adjust until the period closes. |
Why this may differ from Seller Central's Refund Rate
Amazon also reports a Refund Rate in Seller Central, and the two won't always match. Amazon calculates theirs entirely within the Sales & Traffic dataset, using its own units-ordered figure as the denominator. Kapoq uses the All Orders Report instead, which keeps Refund Rate and Return Rate on the same denominator so the two stay comparable to each other. Same refund count, different order count, so the percentages can differ slightly.
Why this won't tie exactly to your financial reports
Refund Rate is built on unit counts from Data Kiosk. The refund figures in Accounting are built on dollars from the Finance API. These are separate Amazon datasets with separate timing rules, so the two will not reconcile line for line; expect a modest difference over any given period.
Use Refund Rate to spot trends and compare products. Use Accounting for anything financial: P&L, reconciliation, or figures you're reporting to a client or finance team.
Data availability
There is no historical backfill, and the underlying data defaults to zero, so date ranges beginning before your refund data starts will show 0% rather than a blank. Treat 0% on an early period as "no data," not "no refunds," and avoid period-over-period comparisons that span the start of coverage. Coverage begins from August 5th, 2026 for most accounts, though a few have less history than others.
Summary/Takeaway
Refund Rate is units refunded ÷ units ordered, for Seller Central accounts only.
It includes refunds where nothing was physically returned, so it runs higher than Return Rate.
Refunds and orders are counted in the same window, but a refund today usually belongs to an older order. When order volume swings, the rate swings with it — check volume before reading anything into a spike.
It won't match the refund figures in Accounting, which come from the Finance API in dollars. Refund Rate is for trends; Accounting is for financials.
There's no backfill, and periods before your coverage starts display 0% rather than a blank. On an early date range, 0% means "no data," not "no refunds."