Product returns can reveal faulty products, delivery problems, unclear promises, or purchases that did not suit the buyer. To learn from a creator campaign, connect each returned item to its original order, code the reason, and compare similar purchases after equal time to return. Raw return totals cannot tell you whether a creator caused the problem.
The useful outcome of influencer campaign return analysis is a specific change to the product, fulfillment process, product page, or creator brief. The method below is a recommended review process. It is not a platform rule or an industry benchmark.
Count physical returns separately from money adjustments
Start with the distinction in Shopify's sales-report documentation. Sales reversals include returns, cancellations, edits and other negative adjustments. A refund can happen without any product coming back.
Shopify separately defines Returned quantity as physically returned units. Its physical-return-specific Returned quantity rate divides those units by ordered quantity. Return line item reason records the selected reason for each line item.
For this review, use an explicitly named custom measure:
Delivered-unit return rate =
physically returned units from a defined delivery cohort
/ original units delivered in that cohortThat denominator differs from Shopify's ordered-quantity measure. Do not compare the two as though they were identical.
Keep return requests awaiting receipt in a separate pending count. Also separate cancellations before delivery, refund-only adjustments, and exchanges. Count an original unit once when it physically comes back, even if an exchange follows. Track the replacement separately so it does not inflate the original campaign's delivered volume.
Join records using order ID, line-item ID and quantity. A partial return should affect only the units returned. For the broader accounting work, use commerce back-office reconciliation before drawing conclusions from mismatched totals.
Give comparable buyers equal time to return
A cohort is a group of purchases selected by the same rule. Define the creator cohort using your recorded campaign code or link attribution. Keep ambiguous attribution separate.
Modash's ecommerce guide describes codes, landing pages and UTM links as tracking inputs. Those inputs can identify purchases for review. They do not prove that the creator caused a later return.
Compare the creator cohort with other purchases of the same SKU and variant, under similar conditions:
- Delivery period and time elapsed since delivery.
- Price, discount and return policy.
- Customer region, shipping service and fulfillment location.
- New versus returning customers.
- Product batch, where recorded.
Use a consistent observation window that covers the merchant's applicable return period plus a reasonable allowance for return transit and processing. Label younger cohorts provisional. Keep outstanding requests visible rather than treating them as retained purchases.
Shopify records sales and reversals on their respective dates. Dividing this month's returns by this month's sales can mix different buyers. Reconnect the returned units to their original cohort first.
If you cannot match every condition, list the differences. A campaign that sold mostly one variant should not be judged against a store average dominated by another.
Use a return-reason coding sheet
Create one row per returned line item and reason, with a quantity column. Split the row when units have different reasons. Preserve the customer's selected reason and support notes, then add your review code in a separate field.
Record these fields together:
| Field group | What to keep |
|---|---|
| Purchase | Order and line-item IDs, SKU, variant, batch, delivered quantity |
| Campaign | Creator identifier, attribution method, post URL and publication date |
| Timing | Delivery, return request and receipt dates; observation cutoff |
| Outcome | Physical return quantity, exchange or refund-only status |
| Evidence | Original reason, relevant support detail, inspection finding |
| Review | Primary code, secondary issue, confidence, action owner |
Use one primary code per returned unit so the totals reconcile. A secondary issue can explain overlap without counting the unit twice.
| Code | Use when evidence indicates | First check |
|---|---|---|
| P: product | Fault, damage or performance below the supported specification | Inspection, batch and other-channel complaints |
| F: fulfillment | Wrong item, missing part or delivery timing problem | Pick record, packing evidence and delivery history |
| E: expectation | Buyer expected a feature, fit or use that the product does not provide | Post, brief, product page and manufacturer facts |
| N: preference | Buyer no longer wants it or prefers something else, with no identified fault or misleading promise | Original reason and any supporting detail |
| U: unresolved | Reason is missing, vague or contradicted by other evidence | Support clarification or inspection |
These codes route investigation; they do not assign blame. Damage may originate in manufacturing, packing or transit. A size complaint may reflect preference, an inaccurate chart, or unclear fit advice. Keep it unresolved until the available evidence supports a narrower explanation.
Treat preference returns as part of the observed baseline. There is no universal percentage that makes them acceptable, and a familiar reason still deserves review if it changes sharply.
Check the promised use against the actual product
A concrete manufacturer fact helps separate a limitation from a fault. YETI's Rambler FAQ says tumblers with a MagSlider lid can spill when tipped in a backpack. The lid does not make that tumbler leakproof.
For a hypothetical complaint about liquid escaping in a bag, inspect the exact model and lid before coding a defect. Then check whether the campaign or merchant page suggested bag-safe transport. The manufacturer fact alone does not establish who created the expectation. This example does not describe an actual YETI campaign or customer return.
If the mismatch came from the destination page, align the landing page with the promised product. If it came from a brand-supplied brief, the brand owns that correction too.
For U.S. advertising, the FTC's advertising guidance requires evidence for claims before an ad runs. A money-back guarantee does not replace that evidence. Necessary qualifying disclosures must be noticeable and understandable. A low return rate therefore cannot establish that a product claim is supported.
Read the cohort comparison before judging the creator
The following data is entirely hypothetical, for an unnamed merchant selling one tumbler SKU. Each cohort contains 1,000 original delivered units. Both have reached a 45-day observation cutoff under the same hypothetical 30-day return policy, with no pending requests. Price, delivery period and customer mix are assumed comparable for this example.
The reference cohort has no recorded creator attribution. That does not prove its buyers never encountered creator content.
| Primary reason | Creator cohort | Reference cohort |
|---|---|---|
| Product | 20 | 20 |
| Fulfillment | 15 | 10 |
| Expectation | 50 | 10 |
| Preference | 25 | 30 |
| Unresolved | 10 | 10 |
| Total returned units | 120 | 80 |
| Delivered-unit return rate | 12% | 8% |
The raw gap is four percentage points. Product-coded returns are equal at 2% of delivered units. Expectation-coded returns are 5% versus 1%, which identifies the first investigation to run.
It does not prove a creator made false claims. The campaign could have reached buyers with different intended uses. The merchant page could have omitted a limitation. Attribution could be incomplete. Read the expectation-coded records alongside the actual content before assigning a cause.
Also keep denominators visible. Expectation returns are 50 of 120 returns in the creator cohort, but 50 of 1,000 delivered units. Those percentages answer different questions.
Turn the finding into one tracked change
Assign the next step according to the evidence:
- A defect appears across channels or one batch: send the records to product quality.
- Delivery or picking problems cluster: send them to fulfillment operations.
- A repeated expectation conflicts with the content: correct the claim and supporting brief or page.
- Reasons remain unclear: review more records before changing creator selection.
Keep commission settlement separate. Use return-adjusted affiliate reconciliation for payment handling under the agreed terms.
For the hypothetical expectation gap, record the precise claim to review, the owner, and the publication date of any correction. Compare the next equally mature cohort using the same coding rules. Your next action is to review those 50 expectation-coded units and identify which expectation the evidence supports changing.
