Free Shipping and Fashion Returns: Measure the Whole Cohort

Free outbound shipping and free return shipping are separate promises. A fashion promotion can change who buys, which sizes sell and how many garments arrive in each parcel. If more items later come back, that sequence alone does not show that the shipping offer caused the returns. Build a review around the original orders and follow them through their return cycle before deciding what to change.

For a US apparel merchant, the useful output is a cohort sheet: a fixed group of orders, the policy each received, a consistent observation period and a reconciled outcome. Our holiday returns FAQ guide covers explaining the rules to customers. This guide covers measuring what happened afterward.

Keep two shipping policies and one order history

Save the outbound offer separately from the return-postage policy. An order might receive free standard delivery while the customer pays return postage, or pay outbound delivery while qualifying for a subsidized return label. Combining these situations into one “free shipping” group loses the distinction you need to investigate.

• Order ID and original line-item or variant ID; order, shipment and delivery dates.
• Outbound shipping charged and waived, service and destination; return-postage charge or subsidy recorded separately.
• Policy version, applicable return-window starting event and deadline, including approved exceptions.
• Original units shipped, returned units received, exchange replacements, refund or credit outcome, and event dates.
• Product, size, discount, first-time or repeat customer, and recorded return reason; leave missing information visible.

Shopify’s return-rule documentation says rule changes apply to future orders. It also distinguishes windows starting from delivery of each item or the last item in the order. Retain the rule that applied to the purchase instead of assigning today’s setting to every historical order. Check split deliveries and later exchanges individually.

Wait for comparable observation time

Define the cohort before inspecting the result. For example, select paid, shipped apparel orders placed during a specified campaign week, excluding test orders and pre-shipment cancellations. Keep a separate reconciliation of excluded orders. An unfulfilled order has no physical shipment to return, but its cancellation can still matter to the campaign’s economics.

Choose a review date that allows the applicable return windows and normal transit and processing delays to run. Report outstanding requests, parcels in transit and unresolved inspections alongside the final outcomes. Do not quietly count every open case as “kept.” Where outcomes remain unsettled, label the analysis provisional and schedule another review.

Consider a fictional 30-day window measured from delivery, with an additional 10-day operational observation allowance. Compare groups at the same 40-day age after delivery, while separately reporting exceptions and later events. Those numbers are illustrative, not Shopify defaults, a recommended policy or a guarantee that all cases finish by day 40. A group observed for seven days cannot fairly be called better than one observed for forty.

Count physical returns with a defined denominator

In an illustrative completed cohort, 200 orders shipped 400 garments. Thirty distinct orders had at least one physical return, totaling 50 garments. The order-level rate is 30 ÷ 200 = 15%; the unit-level rate is 50 ÷ 400 = 12.5%. Count an order once in the first numerator even if it produces several parcels. These measures describe this fictional cohort and are not industry benchmarks.

Keep exchange replacements linked to their original order and identify whether the numerator concerns original units or replacement units. For the example above, the 400-unit denominator contains original shipments only. Do not add replacement shipments to that denominator while leaving replacement returns out of the numerator. Report exchanges separately when the data cannot support a consistent combined measure.

Shopify’s current sales-report definitions distinguish physically returned quantity from reversed quantity, which also includes other order adjustments. Sales reversals can include cancellations and edits as well as returns. Inspect the field definition before using a dashboard percentage as a garment-return rate.

A calendar report and an original-order cohort answer different questions. Sales and reversals can appear on different processing dates, so this month’s negative entries can concern earlier purchases. Join events back to the original order rather than dividing all reversals recorded this month by all orders placed this month. Preserve the event-date report too; operations still need to know the workload arriving now.

A May 2025 Shopify Community reply about return reporting discusses historical changes to return processing and the distinction between goods and money movement. It provides context for older forum advice, not evidence that free shipping changes return behavior. Use current documentation and a reconciled sample of your own orders for today’s report.

Reconcile the dollars without counting a refund twice

Keep physical movement, sales adjustments and payment refunds in separate columns. Shopify’s discrepancy guidance explains that sales reports and payments reports serve different purposes, and date ranges or grouping can show only part of an order. Check a full return, partial return, exchange and refund without returned goods before trusting an aggregate.

For a simplified cash-refund example, suppose the same 200-order cohort collected $20,000 of merchandise receipts after discounts, refunded $3,000, and collected no shipping revenue. Retained merchandise receipts are $17,000. Assume $7,000 of product cost for retained goods, $400 of unrecoverable returned-inventory cost, $1,600 outbound fulfillment, $300 return postage, $250 inspection and handling, and $600 payment costs after applicable adjustments. All figures are fictional USD; taxes, acquisition and fixed overhead are excluded.

The resulting contribution before those excluded costs is $17,000 − $7,000 − $400 − $1,600 − $300 − $250 − $600 = $6,850, or $34.25 per original shipped order. Do not subtract the $3,000 refund again: it is already removed from retained receipts. This is a measurement example, not a profitability forecast or an accounting standard.

The example assumes resalable returned garments remain inventory and their recoverable cost is not also charged as retained-goods cost. Actual write-downs, replacement costs and handling need their own treatment. An exchange or store-credit case requires a separate reconciliation; issuing credit is not the same as sending cash back. Reconcile with your accounting records rather than forcing every resolution into the cash-refund example.

Investigate fit and customer mix before changing the offer

Break the cohort into a few groups chosen in advance: product family, size, new versus repeat buyer, or destination. Show the order counts and missing reasons beside each rate. A small size group with two returns can move sharply with one more case. Differences may reflect assortment, fit familiarity or acquisition channels, so describe them as investigation leads.

Review garment measurements, fit language, photographs and variant selection against recurring reasons. A “too small” label is a clue to inspect the information and product, not proof that the customer ordered carelessly. Record when product information changes so later cohorts are not treated as receiving an identical shopping experience.

AISellor: AI Retail Analytics lists product, order, inventory and customer analysis, alongside detecting missing images or descriptions. It can be a relevant starting point for investigating product-information gaps. Verify the underlying records yourself; the listing does not establish that it automatically constructs this mature-return cohort, supplies every logistics cost or proves a causal effect of shipping policy.

Decide using the whole campaign outcome

A shipping offer could alter both the number of purchases and which customers purchase. Comparing only buyers who redeemed it with buyers who did not creates a selected comparison. Keep the return-cohort view for operational diagnosis, and assess conversion and reconciled contribution across all eligible customers or visitors when evaluating the offer.

Where feasible, use a properly designed randomized policy test with stable eligibility, consistent measurement and enough observations for uncertainty to be assessed. Hold return-postage terms and product information constant if the intended question concerns outbound shipping alone. If several elements change together, the result concerns the package of changes, not one isolated incentive.

For an observational rollout, document dates, product availability, traffic changes and policy exceptions. Do not claim that removing free shipping will lower returns, or that a lower return rate necessarily improves the business. Review retained contribution, conversion, support problems and unresolved cases together, then choose a bounded next test. The deliverable is a traceable cohort decision, not a universal rule about fashion shoppers.

Sources checked November 15, 2026. All cohort counts, observation periods and financial amounts above are illustrative. The reused editorial hero depicts online shopping and delivery costs, not an actual fashion-return operation.