Chargebacks and Their True Cost

Of all the ways an order can go wrong, a chargeback is the most expensive per event, and the one most likely to be recorded as a single number — the reversed order value — when it costs considerably more than that.

Why it costs more than a refund

A refund and a chargeback both remove revenue. Everything else about them differs.

With a refund, you usually get the goods back. They return to stock, possibly needing a markdown, but the cost of goods is largely recovered.

With a chargeback, the customer keeps the product. The cost of goods is gone permanently. The outbound shipping is gone. The packaging is gone. And a dispute fee is added, charged by the provider regardless of whether you ultimately win.

That is why the example totals $158.95 on a $92 order. The order value is less than sixty percent of what the event actually cost.

The recovery arithmetic

Set against a contribution margin of $46.25 per good order, one chargeback needs 3.4 successful sales to recover.

At a one percent dispute rate, that means roughly three and a half percent of your successful orders exist purely to cover disputes. For a store running a six percent net margin, that is more than half of everything it keeps.

This is the framing that makes disputes worth attention. Discussed as a rate they sound trivial — one percent, or a fraction of one. Converted into orders that had to be sold and shipped to break even, they are not.

Where disputes come from

Three broad categories, with different fixes.

Genuine fraud. A stolen card used on your store. The cost lands entirely on you, and prevention is a checkout and verification problem.

Friendly fraud. The cardholder made the purchase and disputes it anyway — sometimes deliberately, often because they did not recognise the charge. This is the largest category for most merchants and the most preventable.

Service disputes. The parcel never arrived, arrived damaged, or was not what the customer expected, and they went to their bank instead of to you. Every one of these is a support failure before it is a payments failure.

The mix matters because the interventions differ completely. A store assuming fraud when its problem is unrecognised descriptors will spend on fraud tooling and see no improvement.

The cheapest preventions

Clear billing descriptors. The single highest-return fix for friendly fraud. If the name on the statement does not match the store the customer remembers, a share of them will dispute rather than investigate. Changing the descriptor to something recognisable costs nothing.

Visible, easy contact. Disputes are what customers do when they cannot reach you. A prominent support route and quick responses convert would-be chargebacks into refunds, which cost less than half as much.

Delivery confirmation with signature on higher-value orders. It is the evidence that wins item-not-received disputes, and without it those cases are largely unwinnable.

Order confirmation and dispatch notifications. They remind the customer who you are before the statement does.

Address verification and fraud screening for genuine fraud, applied proportionately — over-aggressive screening declines good orders, and a declined order costs the whole margin.

Whether to fight

Representment takes time, and the fee is charged either way, so the decision is economic rather than moral.

For low-value orders where evidence is thin, the labour usually exceeds the recovery. For higher-value orders with delivery confirmation, signature and clear communication records, the odds are considerably better and the amount justifies the effort.

The practical rule most stores land on: fight where you have delivery evidence and the order value is meaningful, and do not fight where you would be arguing without documentation. Then spend the time saved on the preventions above, which have a much better return than winning cases individually.

Where the cost belongs in the accounts

Chargebacks are usually recorded as a single deduction in a payout and left there, which loses most of their analytical value.

The cost should be attributed back to the originating order, and through it to the product, the channel and the region. Booked as a general payment expense, disputes become a line that moves for no visible reason. Attributed, they become a pattern: this product generates disputes at four times the store rate, this acquisition channel’s customers dispute far more often, this market has a systematic problem with delivery confirmation.

Those findings have specific fixes and none of them survive aggregation.

The join is available — gateway dispute records carry a transaction ID that maps back to the order — but it has to be made deliberately, because the dispute arrives weeks or months after the sale on a completely different report. It is the same attribution discipline that late refunds and carrier adjustments require, and it fails for the same reason: nobody does it by accident.

The timing that distorts reporting

Disputes can be raised long after the sale — commonly up to ninety days, sometimes longer.

That has two consequences worth building into how you read the numbers. Recent periods understate the dispute rate, because the window has not closed. And when a dispute lands, its cost belongs against the original order rather than the month it arrived in, or the product and channel that generated it look identical to those that did not.

A dispute rate calculated on last month’s orders will always look better than the same cohort will look in three months. Judge the trend on cohorts that have fully matured, and treat anything inside the last quarter as provisional.