Data Quality
Part of Product performance analytics
Linking returns to product-level performance
Link recorded physical returns, reasons and product-value reversals to original order lines using comparable follow-up windows.
Link each recorded physical return to its original order line and product variant where records allow. Compare returned units with units originally ordered in the same cohort and follow-up window. Keep refunds, cancellations and edits separate: they can change sales without showing that goods came back.
Measure the question you have
| Question | Measure | Boundary |
|---|---|---|
| How often were goods returned? | Recorded physically returned units divided by original ordered units | Use comparable order cohorts and return-observation windows |
| How much product value was reversed? | Relevant product-value adjustments linked to those orders | Keep tax, delivery and unrelated fee changes separate |
| What reasons were recorded? | Returned units by line-item reason | Show missing or unlinked reasons |
For a consistent physical-return rate, divide linked returned units by units originally ordered for the same cohort and follow-up window. Label this as your calculation, not a Shopify measure. Treat refunds, cancellations and edits separately, and check the underlying records before treating every recorded return as a verified received item.
Pros and Cons of Using Shopify’s Built-in Return Reports
- ProsProvides pre-aggregated return data - Tracks refund status separately from return processing - Includes reason codes where recorded - Integrates with order and inventory systems
- ConsDoes not always link returns to original line items - Refunds without returns are counted as reversals - Missing reason data may skew analysis - Reversal timing may lag behind original sale
Follow the original order line
Select orders placed in a stated period and retain their line IDs, product or variant keys, quantities and qualifying status. Follow returns and financial adjustments through a stated cut-off. A trading-period sales report instead shows reversals when processed, which may be later than the original order. Label the view before comparing products.
Record the original item, returned quantity, any exchange item and refund outcome separately. Shopify allows a return to be processed before its refund and distinguishes a refund without a return.
An exchange can return one product while supplying another; it should not simply erase the original order line. If a return cannot be linked confidently to an order, keep it in an exception group rather than assigning it to a likely product.
Compare cohorts only after the same return-observation time has elapsed, or mark newer rates provisional. Show ordered units and returned units beside percentages, especially for low-volume products.
Linking Returns to Original Order Lines: Step-by-Step
- Extract original order lines with line IDs and variant keysUse Shopify Sales Report filtered by date range and export line-level data
- Identify returns linked to the same order lines within follow-up windowMatch return records using order ID and line item ID in Return Management
- Flag unlinked or ambiguous returns for reviewAssign to exception group; do not auto-assign to products
- Calculate return rate per product cohortDivide linked returned units by original ordered units; label as custom calculation
Interpret reasons and costs
Keep return reasons linked to the relevant item or order line where the records allow, and make missing reason coverage visible. A refund without a return does not record a return reason in Shopify's return workflow. A cluster of size-related reasons can prompt a check of the goods and size information; it does not establish the cause.
Show physical returns beside relevant reversal value without merging the measures. A resaleable returned item may have a different cost outcome from a damaged one, and postage or handling costs require their own records. If those costs are missing, report the physical and sales measures while leaving contribution unconfirmed.
