What to Do with Adjustments

In ecommerce, the sale is only the first state of the order. The durable systems are the ones that handle what happens after: returns, refunds, chargebacks, partial credits, and price adjustments. Those flows decide margin, trust, and whether your support queue becomes the real product.

Start with the scale#

According to the 2025 Retail Returns Landscape from NRF and Happy Returns, retailers expected 15.8% of annual sales to be returned, about $849.9 billion. Online, the return rate was 19.3%. Free returns were a major purchase consideration for 82% of consumers. Instant refund or exchange options made 76% more likely to choose a path. After a poor returns experience, 71% said they were less likely to shop with that retailer again.

Fraud sits inside the same pile. NRF reported that 9% of returns are fraudulent, and that 85% of retailers are using AI to detect or prevent return fraud. Younger shoppers also return more often: people 18 to 30 averaged 7.7 online returns in the prior year.

Adjustments are not a back-office annoyance. They are a core customer and finance workflow.

Separate honest recovery from abuse#

The Merchant Risk Council's 2026 Global eCommerce Payments and Fraud Report found that 62% of merchants saw an increase in first-party misuse disputes, and 57% saw an increase in refund or policy abuse. Payment fraud was estimated at 3.2% of annual e-commerce revenue globally.

Digital Commerce 360's summary of Chargebacks911's 2026 field report adds more color. Among merchants who saw a change in friendly fraud over three years, 73.7% said it worsened, rising to 83.4% among enterprise merchants. Overall, 74.4% called friendly fraud a significant concern. Thirty-eight percent said chargeback costs had influenced the prices of goods and services. About two-thirds either use AI fraud tools now (26.7%) or plan to (37%).

If your policy only optimizes for saying yes, abuse rises and honest customers eventually pay through higher prices or stricter rules. If your policy only optimizes for saying no, conversion and loyalty fall. The work is classification.

A practical operating model#

I would handle adjustments in layers.

1. Make the happy path fast#

For low-risk, in-policy returns, refund or exchange quickly. NRF's consumer data is unambiguous about immediacy. Speed for honest cases is not softness. It is how you keep support away from work a system can finish.

2. Keep evidence with the order#

Carrier scans, delivery confirmation, chat transcripts, product identifiers, photos, and prior adjustment history should sit on the order record. Chargebacks and friendly-fraud disputes are evidence contests. If the evidence is scattered, you lose by default.

3. Score risk before you score tone#

Use signals that are hard to fake: velocity of returns, mismatched product identifiers, high-risk geographies or payment methods, repeated claim patterns, and BNPL exposure where relevant. Chargebacks911 noted that about 40% of merchants who accept BNPL believe it can increase chargeback exposure.

4. Offer graduated remedies#

Full refund, exchange, store credit, partial credit, and paid return shipping are different tools. Match the remedy to confidence and customer history instead of one blunt policy.

5. Close the loop into catalog and ops#

Repeated returns on the same SKU are a product data problem: sizing, photos, materials, expectations. Repeated chargebacks on the same path are a payments and fulfillment problem. Adjustments should create tickets for merchandising and logistics, not only for support.

6. Use AI as a reviewer, not as an unaccountable judge#

AI is already widely used for return-fraud detection. Keep a human review path for edge cases, and log why a claim was approved, challenged, or denied. An opaque denial creates the next chargeback.

Policy is product#

Retailers told NRF that top priorities included increasing online sales and reducing return rates, while nearly two-thirds planned to update returns processes. The tension is permanent. Customers want easy. Finance wants control. Engineering sits in the middle with state machines, evidence, and clear ownership.

Treat adjustments as a first-class product surface. Define states, owners, evidence, and remedies. Measure refund abuse, friendly fraud, recovery time, and repeat purchase after a return. The brands that handle the second half of the order well will keep the customers the front of the funnel spent money to win.