Backorders: Tell the Customer Before They Ask

The customer ordered on the third. On the nineteenth they emailed to ask where it was. That email is how your business found out it had a problem, and by then the only things left to offer were an apology and a date nobody trusted.

What a backorder costs

A backorder is an order you accepted for stock you do not have yet. Taking it keeps the sale. What happens next decides whether it keeps the customer.

The costs are easy to miss because they land in different places. Support answers the same where is my order question three times. Some customers cancel and some never come back. The warehouse ships a partial order and pays for freight twice. None of it appears on a report headed backorders.

The stock delay itself is often forgivable. What customers punish is finding out late, or being given a date that turns out to be wrong.

Where the date comes from

Every backorder message contains a promise, and the promise is only as good as the information behind it.

Using the spread matters. A supplier who averages ten days and sometimes takes thirty should not produce a promise of ten days. Quote the date you will usually beat, which is the argument in quoting lead times you can keep.

Where AI earns its place

Estimating the date. Supplier by supplier, your receiving records show how late each one tends to run against its promise. Turning that into an expected arrival with a range is analysis that is tedious by hand and routine for software.

Sending the update before the question. When a purchase order slips, every customer waiting on that item can be told the same day, with the new date and their options. Drafting those messages in your tone, with the order details filled in, is the kind of job that does not get done manually when it involves forty customers.

Spotting trouble early. An item selling faster than its reorder plan assumed will go on backorder next week. Flagging it now gives purchasing a chance to act, and it is the other side of the item-level view in reading inventory turnover by item.

Sorting the queue when stock lands. Listing who has waited longest, who is a contract customer and which orders can ship complete is quick to assemble. Deciding the order is a different matter.

What to keep with a person

Allocation. When thirty units arrive against fifty on backorder, somebody decides who gets them. First come first served is fair and may cost you your largest account. That is a commercial judgment and it should have a name against it.

The promise itself, when the evidence is thin. If there is no confirmed purchase order, an automated message with a confident date is worse than an honest one saying you are waiting on the supplier. A model will happily produce a date. Make it show what the date is based on, and route the ones with no basis to a person.

Substitutions and cancellations. Offering an alternative product or a refund is a customer conversation, and for a business customer it may touch a contract.

The fix upstream

Count how many backorders were avoidable. Some come from a supplier failing. Many come from a reorder that was raised late, a count that was wrong, or a product page that showed stock you did not have.

Those are process problems with known fixes: accurate counts, as in cycle counting, and purchase orders raised on time, as in how purchase orders get raised. A customer you never had to apologise to is the cheapest outcome, and it protects the repeat business described in what your retention data already shows.

Frequently Asked Questions