Calculate return rate per style and per size, using returned units divided by delivered units for the same order cohort. Split customer returns from RTO, then group return reasons into size, quality and expectation issues. Flag any SKU or size whose return rate is clearly above your own category average, provided it has enough volume to be meaningful.
Start with a clean return rate per SKU
The base metric is simple: returned units ÷ delivered units for the same group of orders. Do this at style level first (parent SKU), then drill into colour and size.
- Use order-date cohorts (for example, orders placed in a given month). Returns trickle in for weeks, so comparing this month's sales with this month's returns understates the rate.
- Use delivered units as the denominator, not ordered units. Cancelled and undelivered orders will otherwise dilute the figure.
- Set a minimum volume before you judge a SKU. A size with 8 units and 4 returns is a signal to watch, not a conclusion.
Pull the data from your marketplace sales and returns reports. Column names differ by marketplace and report version, so check the field definitions in each report before building formulas.
Separate customer returns from RTO
RTO (return to origin) and customer returns have different causes and different fixes. An RTO is a shipment that never reached the customer, so it points to address quality, courier performance or order confirmation. A customer return happens after delivery, so it points to product, fit or listing accuracy.
Tag each return as one of: customer return, RTO/undelivered, or other (such as exchange). Only customer returns should feed your size analysis. Mixing RTO into it will make a perfectly fitting style look like a sizing problem.
Break returns down by size and reason
A style-level return rate hides the real story. Build a size-wise table for each high-return style with delivered units, returned units and return rate per size.
Then group the return reasons. Marketplaces usually record a reason selected by the customer or logged at the warehouse. Map them into three buckets:
- Size and fit: too small, too large, does not fit, length or sleeve issues.
- Quality: defect, damaged, fabric, stitching, wrong item.
- Expectation: colour different from image, not as described, did not like.
Patterns to look for: returns bunched at one end of the size run (usually grading or a size chart problem), a single middle size with high returns (often a fit or pattern issue), or uniformly high returns across all sizes with a quality or expectation reason (not a size problem at all).
Set flags relative to your own catalogue
Do not borrow a universal benchmark. Return rates vary by category, price band, season and marketplace. Compare each SKU with your own category average or median instead.
- Calculate the median return rate for each category (kurtas, tees, dresses and so on).
- Flag styles at, say, 1.5 times the category median. This multiple is a working rule you choose, not an industry standard, so tune it to your data.
- Within flagged styles, flag any size whose rate is well above the style's other sizes.
- Check the share of size-and-fit reasons. If it is the majority of returns, treat it as a sizing issue first.
Diagnose the cause before changing anything
Once a size problem is visible, test the likely causes in order of cost:
- Size chart and listing: does the chart show garment measurements or body measurements? Is the model's height and worn size stated?
- Fabric and fit notes: a low-stretch fabric or a slim cut may need a "size up" note.
- Production variance: measure a few units from the batch against your spec sheet. Returns that cluster in one batch suggest a manufacturing issue.
- Customer comments and ratings: phrases like "runs small" confirm the data.
Fix the listing first, then production specs, and re-check the size-wise rate on the next cohort of orders.
Track costs and check what the marketplace charged
High-return styles cost more than the lost sale: forward and reverse logistics, fees on returned orders, and damaged or unsellable stock. What is charged on a return, and what is reversible, depends on your seller agreement and rate card, so confirm the latest terms there and in your settlement report.
Tools like MarketplaceMetrics by Fundle can reconcile sales, returns and settlements and check fees against agreed terms, so you can see which high-return styles also carry charges that look wrong. Myntra Excel reports are reviewed today; other marketplaces are a scoping discussion. Whatever you use, keep a list of returns that came back damaged, missing or wrong so you can raise recovery cases within the marketplace's timelines.
Worked example
Illustrative example (figures are made up for explanation only). A kurta style, KT-204, has 360 delivered units in a month's order cohort and 100 customer returns, a 27.8% return rate against an assumed category median of 18%.
| Size | Delivered | Returned | Return rate |
|---|---|---|---|
| S | 60 | 9 | 15.0% |
| M | 120 | 22 | 18.3% |
| L | 110 | 38 | 34.5% |
| XL | 70 | 31 | 44.3% |
S and M sit near the category median, while L and XL are far above it. Of the 69 returns in L and XL, 48 (about 70%) cite "too small". That points to a grading or size chart problem at the larger sizes, not a general quality issue. The action: measure L and XL units against the spec, update the chart with garment measurements, and add a "size up" note if needed. If the handling cost per return is an assumed ₹120, the 100 returns cost ₹12,000 in that month before any lost margin.
Questions and answers
What is a high return rate for an apparel SKU?
There is no universal number. Return rates differ by category, price band and marketplace. Compare each style with your own category median, and treat styles well above it, with enough volume, as high return. Check your marketplace reports for current definitions.
How do I know if returns are size-related?
Group return reasons into size and fit, quality and expectation. If most returns on a style cite too small, too large or fit, and the rate spikes in particular sizes, the cause is probably sizing. Uniformly high returns across all sizes point elsewhere.
Should I include RTO when calculating return rate by size?
No. RTO shipments never reach the customer, so they say nothing about fit. Keep RTO as a separate metric for address, courier and confirmation issues, and use only delivered-then-returned units for size analysis.
How much data do I need before acting on a size problem?
Enough that a few returns do not swing the percentage. Set your own minimum delivered units per size, and watch low-volume sizes over more than one cohort. Combine the numbers with customer comments before changing the size chart or production.