The product return rate formula, and the half you can fix

The product return rate formula

The product return rate is the share of what you sold that came back: divide units returned by units sold over the same period, then multiply by 100. The division is the easy part. The denominator is a decision, and the number it produces averages the returns you agreed to accept with the returns your operation caused. Only one of the two is yours to fix. The pattern I keep meeting, a composite rather than one brand, is a monthly rate that sits flat for two quarters while the share coming back as wrong item or damaged climbs inside it. Split the same numerator by reason code and the conclusion reverses, with no change to the arithmetic.

The formula scales down unchanged to one SKU, category or return reason, provided you narrow both halves of the fraction together and both still describe the same thing. Unlike most customer satisfaction metrics, this one already sits in your order data, and it's worth getting right: NRF and Happy Returns put 19.3% of online sales returned in 2025.

A worked calculation, start to finish

Take a store that sold 12,000 units in March and received 960 units back. That's a product return rate of 8% for March: eight of every hundred units that left came back. Say it as a sentence, not a bare number, because a bare number invites the next person to compare it to something that is not comparable.

Now notice what just happened. In one division you made three choices and nobody asked you to make any of them: what counts as a unit, which period each half of the fraction was drawn from, and what counts as a return. The 960 came back in March. The 12,000 were sold in March. Those aren't the same population of orders. The next three sections are those three choices, in order.

Three denominators, three different numbers

There's not one return rate. There are three standard ones, and they don't agree.

• Item-based: returned items over items sold. It counts things and answers a warehouse question.
• Revenue-based: the revenue of returned items over revenue sold. It counts money and answers a finance question.
• Profit-contribution-based: margin lost to returns over margin earned. It answers the only question that decides whether a return population is worth acting on.

El Kihal, Nurullayev, Schulze and Skiera, in the Journal of Retailing in 2021, ran all three across more than 8 million transactions at sixteen retailers and found the rates they produce differ on average by 24.3%. The second half of that finding is the useful half: whichever method you use, the rate moves similarly over time. Comparing your number to another retailer's is unsafe. Comparing it to your own last month is sound.

The pages ranking for this term publish item-based, order-based, shipped-based and revenue-based formulas between them, each as the formula, and none acknowledges the others exist. The USPS Office of Inspector General, quoting NRF, reports 17.6 percent by revenue of US online purchases returned by 2023. The qualifier is the point. Pick one method, write it down, and never change it mid-series.

The period problem nobody mentions

A return arrives weeks after the order that caused it, so the standard monthly calculation divides one month's returns by a different month's sales. How long that lag runs is set by the returns policy you publish, and in some markets by statute on top of it. The UK's Consumer Contracts Regulations 2013 are the clearest published floor: a consumer may cancel without giving a reason, then has a further 14 days to send the goods back, with the refund due within 14 days of receipt. Whatever window you publish, weeks pass between a sale and the return that belongs to it.

Attribute each return to the order month rather than the month the warehouse processed it and raised the RMA in shipping. Accept that the most recent month or two is incomplete. A cohort rate is right and late. A calendar rate is fast and wrong, and a brand reporting the fast one to a board should say so. The tell is a return rate that appears to be falling in the current month: usually the window hasn't closed.

The number your platform hands you is not this number

Shopify's default order reports don't give you a return rate. They give you a reversed quantity rate, documented as the percentage of items removed from orders compared to the total items ordered in the period, where a reversal covers refunds, returns, cancellations, or edits. Read that as your return rate and it's inflated by three things that aren't returns.

• Return rate: units returned over units sold.
• Refund rate: the returns that ended in money going back out. Not every return does.
• Reversed quantity rate: Shopify's default, which bundles cancellations and order edits in with returns.
• Exchange-adjusted return rate: returns net of items swapped for another variant. A different question about the same event.

None of the four is wrong and only one is the return rate, so check what your report counts before you quote it. A cancellation isn't a return: under the FTC's Mail, Internet, or Telephone Order Merchandise Rule, a shipment the seller can't make in the promised window is canceled and refunded. Shopify's returns documentation then treats a return and a restock as separate events chosen at processing.

One number, two populations

Every return in that numerator belongs to one of two groups, and the two have different owners.

• Returns the customer chose: two sizes ordered and one kept, a color that read differently on a screen, a gift that missed. Merchandising and policy outcomes, and a cost of selling online.
• Returns your operation caused: the wrong variant picked, an item that arrived broken, a parcel that arrived after the date you promised. In Shopify's Admin API the return reason enum carries ten values, and the failure-side ones are spelled DEFECTIVE, NOT_AS_DESCRIBED and WRONG_ITEM. A late delivery has no value of its own, so it's filed under a value that doesn't name it, which is why the operation-caused share is undercounted.

Run the formula twice, once with each population as the numerator and the same denominator both times. The first rate is a merchandising and policy number. The second is your post-purchase failure rate, already collected, free to compute, and denominated in refunds you have already paid rather than in anyone's survey response. ASCM's SCOR Digital Standard already models return as a top-level supply chain process in its own right.

The two populations move independently, which is why a flat headline rate can hide a rising failure share, and why on-time delivery belongs in a returns conversation at all.

What the failure half is actually telling you

The operation-caused rate counts promises that broke after checkout, and every one was visible in the operation before the customer felt it. A WRONG_ITEM return was a pick error, visible at pick confirmation. A parcel that stops moving between carrier scans sits there for days before anyone raises a ticket. Keeyu was built on that gap.

A helpdesk can apologize for the wrong variant, authorize the return and process the refund. What it can't do is stop the pick error, and a return it handled perfectly still lands in the numerator. That's not a tooling gap, it's what a reply-shaped system is for. Acting on the break instead of replying about it is a different category: post-purchase operations, which is what we mean by proactive e-commerce operations.

Keeyu isn't an analytics platform, not a BI tool, not a returns portal, not a chatbot, not a carrier and not an OMS. It doesn't calculate your return rate and won't appear in your reporting. If the job this week is the report, build it, and build the split above. Every order is a promise, and Keeyu keeps the promise. The operation-caused rate counts the ones that did not.

Start with the half you can fix

Splitting the number by cause takes an afternoon. What comes back isn't a metric, it's a list: the wrong variants, the damaged units, the parcels that missed the date you promised, every one an operational failure nobody acted on before the customer felt it. Keeyu watches for those breaks in your order and carrier data, decides what to do about each, and acts before the customer knows. If the half you caused is bigger than you would like, book a demo.

• National Retail Federation and Happy Returns. 2025 Retail Returns Landscape. Returns at 19.3% of online sales in 2025, the one benchmark figure used in this article.
• El Kihal, S., Nurullayev, D., Schulze, C. and Skiera, B. A Comparison of Return Rate Calculation Methods, Evidence from 16 Retailers. Journal of Retailing 97(4), 676 to 696, 2021, DOI 10.1016/j.jretai.2021.04.001. Three standard methods across more than 8 million transactions at sixteen retailers, differing on average by 24.3%, with rates that still move similarly over time.
• USPS Office of Inspector General. Sending It Back, reverse logistics and the US Postal Service, RISC-RI-24-005, May 2024. NRF's estimate that 17.6 percent of US online purchases by revenue were returned by 2023.
• UK Consumer Contracts (Information, Cancellation and Additional Charges) Regulations 2013. Regulation 29. A consumer may cancel a distance contract without giving a reason.
• UK Consumer Contracts (Information, Cancellation and Additional Charges) Regulations 2013. Regulation 35. The consumer must send the goods back no later than 14 days after telling the trader.
• UK Consumer Contracts (Information, Cancellation and Additional Charges) Regulations 2013. Regulation 34. The trader reimburses within 14 days of receiving the goods back.
• Shopify Help Center. Order reports. Reversed quantity rate is the percentage of items removed from orders compared to the total items ordered in the period, and a reversal covers refunds, returns, cancellations, or edits.
• Cornell Legal Information Institute. 16 CFR 435.2, the FTC's Mail, Internet, or Telephone Order Merchandise Rule. A shipment the seller cannot make in the stated window becomes a cancellation with a prompt refund.
• Shopify Help Center. Creating and managing returns. The restock decision is made when the return is processed, so a return and a restock are separate events.
• Shopify Admin GraphQL API. ReturnReason enum. The ten values a Shopify return can carry, including DEFECTIVE, NOT_AS_DESCRIBED and WRONG_ITEM, with no value for a late delivery.
• Association for Supply Chain Management. SCOR Digital Standard. Return is a top-level supply chain process in SCOR.