Delivered, but not received

The scan says delivered. The customer says nothing arrived. Somebody in that exchange is wrong, and you can't tell which one from your desk.

This is the single most common claim category in ecommerce, and it's the one that decides whether a merchant thinks protection works.

Why the scan proves less than it looks like it proves

A delivered scan means a driver pressed a button. It doesn't mean the parcel reached the right address, that it was still there an hour later, or that the driver pressed the button at the doorstep rather than three streets earlier while running behind.

Carriers mark parcels delivered when they're handed to a neighbour, left behind a bin, put in a lobby, or occasionally when they're lost and the driver would rather not have the conversation. Photo proof helps and it's far from conclusive: a photo of a box on a porch says nothing about who picked it up ninety seconds later.

So the scan is evidence. It isn't proof, and treating it as proof is how merchants end up calling honest customers liars.

The three people making this claim

They look identical in your inbox.

The first genuinely lost the parcel to theft or a misdelivery. They're the large majority, they're annoyed, and they'll forgive you quickly if you make it right.

The second has the parcel. A partner brought it in, it's behind the plant pot, a neighbour signed for it. They aren't lying, they just haven't looked properly yet. A short, non-accusatory nudge resolves a surprising number of these before they become claims at all.

The third is defrauding you, and they know exactly how this category works, which is why they chose it.

Any policy that treats all three the same is wrong twice: it insults the first group or it funds the third.

Waiting periods are doing real work

Asking a customer to wait a few days after a delivered scan before filing feels like friction, and it's the single most effective thing in the process. Parcels marked delivered turn up. They turn up in porches, with neighbours, and at addresses one digit out.

The waiting period isn't there to discourage claims. It's there so the genuine claim gets filed by someone who has already checked, which makes it faster to approve when it arrives.

What separates the third group from the first

Not the story. The story is always the same, because there's only one story to tell.

What separates them is pattern. How many times has this customer done this. Do the addresses match. Is the account new. Does the claim cluster with others on the same street or the same card. None of that is visible in a single claim, and all of it is visible across a network of merchants who share signals rather than each guessing alone.

This is why fraud scoring belongs underneath the claims process rather than beside it. A first-time customer with a genuine loss and a repeat abuser send identical emails. The difference is entirely in what you know that isn't in the email.

The decision most merchants get wrong

Faced with a claim they can't verify, most merchants either approve everything to avoid the argument, or tighten until honest customers give up.

Blanket approval is expensive and it teaches a small group exactly how to exploit you. Blanket suspicion is more expensive, because the cost lands on the customers you wanted to keep and it arrives as a review rather than a line item.

The workable answer is to be fast and generous with the clear cases, and to have something other than instinct for the rest.

---

Written by the Guide Team. We publish what we learn running this for merchants, not what ranks.