Blog · Support operations

What to do when a customer says they never got their order

A lot of brands don't have a clear way to handle these. Maybe there's no playbook yet, or the rules exist but they're fuzzy. Below is how I'd set them once, with a builder at the bottom that writes the SOP for you.

By Nathan Tone · September 13, 2026

Skip to the SOP builder ↓

It usually shows up a few days after tracking says delivered. "Hey, I never got my order." Sometimes there's a photo of an empty porch attached. You've already spent money getting this order to the customer, and now you have to decide whether to spend more: send another one, refund it, or start asking questions.

And nobody can prove whether a box got opened. So without a clear rule, the call ends up depending on whether the founder is still answering support, whether there's a playbook at all, how busy the day is, how the P&L is looking, and sometimes how well someone slept the night before. Two customers with the same problem can get very different answers from the same brand in the same week.

In my experience it works better to decide ahead of time. You pick an order value you're comfortable with. Under it, you send the replacement and don't ask any questions; over it, you do a bit of digging first. Customers who keep having the same problem get their own track. Once that's in your SOP, everyone who answers these, whether that's a person or AI, makes the same call.

Pick your number first

Everything else depends on that order value. There isn't a correct figure. It comes down to your stomach and your budget.

The simplest way I know to find it is to count your claims from the last few months and look at what a replacement actually cost you on average. Multiply the two and you've got a rough monthly cost for replacing every one without questions. If you can live with that, the number can sit pretty high. If you can't, bring it down until you can.

It helps to think of that monthly number as a marketing cost. It's a manageable investment in goodwill with your customers, and in the community, repeat orders, and lifetime value that come with it.

On the other side of the scale is what investigating costs. There's the back-and-forth over email and the time someone spends clicking around a carrier's site, and you're asking a customer who may well be telling the truth to prove it. On a cheap order, that usually adds up to more than the order is worth.

First claim, under your number

Send the replacement, or the refund if that's how you'd rather handle it, and don't ask them to prove anything. Let them know it's on the way and send tracking once it ships.

This is where you want most of your claims to land. Yes, it's generous, and it's usually the cheapest option as well. It also gives an honest customer a good reason to order from you again.

First claim, over your number

Above the line, a few questions are fair before a replacement goes out the door. I'd keep the tone curious, since plenty of these packages turn up with a neighbor or in a building's mail room.

1. Ask them to have a look around: anyone else in the house, the neighbors, the mail room, a side door.
2. Read through every carrier scan, including the ones before "delivered."
3. See whether the carrier took a delivery photo.
4. If the tracking looks strange, give it a couple of days for a late scan.

If none of that turns anything up, resolve it the way you would under the line, and think about filing a claim with the carrier. At least you've done a little homework before sending something expensive.

Whichever path a claim takes, tag the ticket when a free replacement or refund goes out. That way the next conversation with this customer starts with the context that they've already had one on the house.

When tracking stops moving

If tracking hasn't updated in days, the package is stuck somewhere between you and the customer, so skip the neighbor questions entirely.

Decide how many days with no scan counts as stuck. Once you hit it, get in touch and tell them what you're doing about it. Ideally they hear from you before they've had to write in at all.

When it's the second or third claim

A no-questions rule only works if someone notices when the same customer keeps using it. That means whoever answers needs to see the customer's history, and this one order won't tell them much. The tags from earlier are what make that history easy to spot.

If someone's had a few of these, whether it's missing packages or missing packages plus other kinds of claims, stay kind anyway. You can politely empathize with what they're dealing with and explain how it'll go from here: the next one gets looked into, or someone on your team reviews it. Accusing a customer of fraud in a support email helps nobody.

Won't a no-questions policy get abused?

Some people will take advantage of it. That's why the no-questions part only covers a first claim under a number you picked, and why repeat claims get their own process.

Questioning every claim has a cost too, though. You'd be treating all your honest customers like suspects to catch a handful of people, and those people tend to show up as a pattern after a claim or two anyway.

And the money that does go out under your number is the marketing spend from earlier: an amount you set on purpose, going into goodwill with the customers who stick around.

Writing it down

All of this comes down to making the decision once, before the email arrives. In our work, we put these rules in the SOP and have AI reason over the SOP, the customer's history, and their message to put together the action and the reply. It takes a lot of the pain and emotion out of a tense ticket, because nobody is deciding from scratch whether to believe someone.

You can get most of the way there without any AI by writing the rules down. Set your number and your preferences in the builder below and it'll write a first version of the SOP you can hand to your team. Then nobody has to play detective on a Friday afternoon.