At a glance
Customers and B2B Companies ranked by refund value over the rolling 90-day window. Helps Operations spot serial returners, fraud-pattern customers, and damaged-shipment-cluster B2B accounts. The card sits between “this is a normal returns operation” and “this is a possible fraud or chronic dispute” depending on context, the same customer at #1 means very different things in DTC vs B2B.
Calculation
Calculated automatically from your Adobe Commerce data. See the At a glance summary above for what the metric tracks and the worked example below for a typical reading.Worked example
A B2B+DTC homewares merchant on Adobe Commerce 2.4.7 with B2B Companies enabled. Snapshot 13 May 26, 90-day rolling. Top 10 refunding customers/companies:
Insight pattern:
- B2B-dominated top of the list (5 of top 10). This is normal for a B2B-leaning merchant; high-volume accounts produce high-value returns even at low rates.
- Three fraud watch flags. Customers at ranks 3, 5, and 9 have refund rates above 60% with multiple Credit Memos. That is a fraud pattern: they buy, return, repeat. Often reselling on secondary markets, sometimes “wardrobing” (wear once, return). Action: add to a watchlist and require manual approval on next order.
- Serial returners that are NOT fraud (ranks 7 and 10). Refund rates of 26-27% with consistent buying behaviour signal a customer who routinely buys multiple items to choose from, then returns the rest. This is “try-before-you-buy” behaviour, common in apparel. Customer service decision: keep them (they are net positive) or block them (they are net negative once handling cost is included). Compute net contribution:
LTV × gross_margin − handling_cost × refund_count. For most DTC merchants, a 25% refund rate at $400+ of net LTV is still profitable. - Westfield Schools at #2, refund rate 11.6%. Higher than the B2B norm of 2-5%. Drilling into Top Refunded Products by-customer view: 14 of 16 returned line items are uniform line items affected by the supplier sizing error. Not their fault; not fraud. Sales should reach out with the corrected sizing and a goodwill credit to retain the relationship.
- Cross-link with Refund Value: the top-10 customers here account for 284,600 90-day refund value, or 24%. Pareto-typical; the long tail is healthy normal-customer returns.
- Action queue for Operations: (a) flag the 3 fraud watches for next-order manual review, (b) Sales call to Westfield Schools with sizing correction, (c) audit Riverside Hospital’s 4 returns to ensure they were genuine damage claims (B2B fraud is rare but exists).
Sibling cards merchants should reference together
Reconciling against the vendor’s own dashboard
Where to look in Adobe Commerce Admin: Adobe Commerce does not have a native “Top Refunding Customers” report. Closest equivalents:Reports > Customers > Customers by Number of Orders is the inverse view (top buyers, not top refunders). No native top-refunders ranking.To verify any single customer’s refund value:
Customers > All Customers > [customer] > Orders tab. Manually count Credit Memos against that customer’s orders. Tedious but exact.For B2B Companies:
Customers > Companies > [company] > Orders tab. Same logic at the Company level.Why our number may legitimately differ from a manual Admin count:
Cross-connector reconciliation (when these connectors are connected for this merchant):
Known limitations / merchant FAQs
A guest customer is at #1, can I see their order history? Click through on the row to the customer email’s order list. Adobe Commerce stores guest orders bycustomer_email and customer_is_guest=1. The Vortex IQ workspace renders an aggregated profile by email, including all guest orders matched.
Why is my B2B Company at #1, isn’t that bad?
Not necessarily. High-volume B2B accounts have high absolute refund value because they have high absolute order value. The refund-rate column normalises: an account with 24k of refunds is at 8.6%, well within B2B norms. The customers to worry about are the ones with refund rate >40%.
A customer’s customer_id changed, why is their history split?
Customer ID is stable in Adobe Commerce; it doesn’t change. What can change is customer_email. Registered customers are aggregated by customer_id, so an email change doesn’t split them. Guests are aggregated by email, so a guest changing emails does split. There is no clean way to merge guest profiles in Adobe.
Should I block fraud-watch customers from buying?
Use a graduated response: (1) flag for manual review on next order; (2) require pre-payment via wire transfer (no chargeback risk); (3) blacklist if the pattern repeats. Adobe Commerce supports per-customer blocking via Customers > All Customers > [customer] > Account Information > “Disable account”. Use it sparingly; false positives create CS load.
A wholesale account showed up here for one bulk return, isn’t that misleading?
The card is a snapshot. A single 284k LTV account is not a problem. The diagnostic column (refund rate) gives context. Look at it together with the count column; one whale return at 8% rate is benign, four returns at the same rate suggest a pattern.
Why isn’t fraud detection automatic?
Because the false-positive rate of pure-statistical fraud detection in B2B and certain DTC verticals is too high. A fashion DTC merchant who accepts try-before-you-buy will see legitimate 25-40% refund rates for repeat customers; flagging those automatically would alienate good buyers. The card surfaces the data; the merchant applies context.
My multi-store, can I see top refunders per Store View?
Yes. Apply the Store View filter to the card. Useful when, e.g., the US site has different refund norms than the UK site (US allows 90-day returns vs UK statutory 14-day; rates structurally differ).
Can I export this list to Klaviyo as a suppression segment?
Yes via the connector to Klaviyo. The fraud-watch flag becomes a Klaviyo profile property; you can build a segment “exclude where vortex_fraud_watch = true” and use it on acquisition flows.
Why does the same customer appear in two rows?
Three causes: (1) guest using two emails; (2) one email registered, the other not; (3) data quality issue (typo in customer_email). The Vortex IQ workspace supports merge-by-email for cases (2) and (3); case (1) cannot be resolved without manual intervention.