Resolving Wrong, Missing, and Damaged Items: A Resolution Playbook

A customer sends a photo of a dented box and one line: “this is broken, fix it.” Behind that message are questions that need answering first — was the wrong item shipped, was something missing from the box, or was it genuinely damaged in transit? Each scenario calls for a different resolution path, but the customer isn’t going to sort that out for you. They just want it fixed.

Wrong items, missing items, and damaged items are among the most common — and most argument-prone — post-purchase issues in cross-border e-commerce. This playbook covers how to turn “customer says it’s broken” into “replacement shipped or refund issued” through a repeatable process with clear ownership, instead of improvising every time.

Sort the issue type first — the evidence bar is different for each

The first step is classifying the complaint correctly, because what counts as proof — and what happens next — differs by type.

All three often end up resolved the same way — a replacement — but wrong-item and missing-item cases are usually clear-cut with little dispute, while damage cases involve carrier liability and need a higher evidence bar. Classifying first, then collecting evidence, saves a lot of back-and-forth.

Step one: let AI support take the first message and start collecting evidence immediately

A customer’s first message is usually incomplete — “it’s broken,” with no detail on whether the outer box or the product itself was damaged, and no photo attached. The slow part of resolution usually isn’t the judgment call — it’s getting all the necessary details in one pass instead of five.

Here’s what AI support can do at this stage:

• Ask the right follow-up questions based on issue type (wrong / missing / damaged) — order number, condition of what arrived, packaging state, photos;
• Guide the customer through what to photograph and from which angle (a close-up of the damage plus a shot of the full outer packaging);
• Pull the order’s history from the cross-border CRM to check for prior similar complaints, saving the agent a manual search;
• Recognize signals like an upset customer or a high-value order and hand off to a human agent promptly, rather than continuing to loop through questions on its own.

The goal here isn’t for AI to make the call — it’s to get the evidence as complete as possible on the first message, so the agent picking it up isn’t the one asking for a photo that could’ve been requested earlier.

Step two: build an evidence checklist by issue type

A complete evidence set cuts down on friction in the judgment and fulfillment steps that follow:

• Photos of the item in question from the customer (at least 2, showing the damage or wrong-item detail)
• Photos of the full outer packaging (key for determining whether damage happened in transit)
• Original order details (item purchased, quantity, SKU)
• Packing/fulfillment record (to check whether an item was actually left out)
• Carrier delivery record and tracking history (to check for anomalies in transit)
• Prior conversation history with this customer (repeat complaint or first occurrence)

Step three: who decides — and where AI’s role stops

Judgment splits into two paths. For wrong-item and missing-item cases with clear-cut fault, if the knowledge base already has a defined rule (for example, “wrong item ships a correct replacement automatically; whether the wrong item is returned depends on its value”), AI support can give the customer a preliminary answer based on that rule so they know roughly what to expect.

But damage cases involving carrier liability, higher-value orders, or any decision that involves issuing a refund or replacing an expensive item should go through human approval. AI support’s job in this step is to hand the agent clean evidence and full order context — not to make the final call itself.

Step four: fulfillment — keep the action and the record in sync

Once a decision is made, a few things matter in execution:

1. Generate a separate order for the replacement rather than reusing the original order number, to avoid confusion during shipping and finance reconciliation;
2. Decide upfront whether the original item needs to be returned, and tell the customer before shipping the replacement — asking for it back afterward creates an unnecessary second round of contact;
3. Write the resolution back to the customer record, so the next time this customer reaches out, the agent or AI support can immediately see “replacement issued for damage” instead of asking the same questions again;
4. Confirm status with the customer, even a short line like “replacement is on its way, expected in X business days” measurably cuts down on “where is it” follow-ups.

Step five: turn recurring issues into a faster resolution path

Looking back over a stretch of wrong-item, missing-item, and damage cases usually reveals patterns — a particular warehouse with an elevated wrong-item rate, a fragile product category with recurring damage, or a shipping lane with a cluster of damage complaints. These findings belong to the supply chain and logistics side and are worth routing to those teams directly.

At the same time, the judgment agents build up while handling these cases — for example, “for this kind of damage description, ask for this specific photo angle first, it speeds up the decision” — can, once flagged by an agent and confirmed by the owner in the review queue, become part of what AI support knows how to do. Every change like this stays traceable, can be tested, and can be rolled back with one click if it doesn’t work out — it never takes effect automatically, and it never bypasses approval to touch a replacement or refund decision directly.

There’s no shortcut for wrong-item, missing-item, and damage resolution, but there is a repeatable structure: classify first, then collect evidence; let AI support gather the full picture on the first message; keep human judgment on anything involving liability or money; log every action at fulfillment; and feed recurring patterns back into the knowledge base. Get that loop running, and what the customer experiences isn’t a slow investigation — it’s a fast answer.