Chargeback vs refund cost, and why cheaper is not always better

"Just refund it, a chargeback costs more" is the most repeated piece of advice in this category. It is often correct.

It is also stated without any of the numbers that decide whether it is correct in your case, and those numbers are ones you already have.

Three outcomes, three different bills

The same disputed order can end three ways, and the costs are not shaped alike.

You refund before anything is filed. You lose the order value and the goods, if they shipped. Your processing fees may or may not come back to you, which depends on your processor and is worth confirming rather than assuming. No dispute fee. Nothing enters your dispute count.

You refund on a pre-chargeback alert. Same as above, plus the alert fee. Published per-alert prices sit somewhere around $15 to $29 depending on the network, and you pay it whether or not that particular cardholder would ever have gone on to file.

It becomes a chargeback and you fight it. You risk the full order value, you pay a fee of $15 to $25 or more, and you spend 20 to 45 minutes assembling the evidence packet. Then, win or lose, the dispute sits in your ratio numerator. That last part is the one people leave out of the comparison, and winning does not remove it.

The inputs, and where each one comes from

You need six numbers, and none of them are industry averages.

Order value, from the order. Recoverable value, which is usually zero once shipped and roughly full cost if you intercept in time. Alert fee, from your provider's rate card.

Dispute fee. From your processor's schedule, not from a blog. Check whether it is returned when you win, because that changes the answer.

Recovery rate. Your own, measured as recovered dollars over disputed dollars. Not wins over cases fought, which flatters you the moment you start declining the hard ones. This is why win rates mislead.

Ratio headroom. Your dispute count over settled transactions, against the threshold for your region. Visa's current VAMP fact sheet puts the excessive-merchant threshold at 150 basis points for the United States, Canada, Europe and Asia Pacific as of April 1, 2026, with CEMEA on a different figure. Check the live fact sheet rather than trusting that sentence.

A worked example, with the assumptions stated

This is one example with numbers I have chosen to make the mechanics visible. It is not a benchmark and it is not typical of anything. Substitute your own.

Assume a $200 order, delivered, nothing recoverable. A $20 dispute fee charged whether you win or lose. Thirty minutes of evidence work. A recovery rate of 40%, taken from your own last quarter.

Fighting it costs you the 60% you do not expect to recover, $120, plus the $20 fee, so $140 and half an hour of somebody's time.

Refunding it proactively costs the full $200.

Deflecting it on a $19 alert costs $219.

On those assumptions, fighting is the cheapest option by a wide margin, and the deflection is the most expensive thing on the list. That result is not an accident of the numbers I picked. Fighting beats refunding in pure dollars whenever your recovery rate exceeds the dispute fee divided by the order value, which here is $20 over $200, or 10%. Most merchants clear that bar comfortably.

So a purely financial model will tell you to fight nearly everything. Run the model and you will get that answer, over and over.

What the model does not know

It does not know how close you are to a threshold.

The gap between fighting at $140 and deflecting at $219 is about $79. On these assumptions, that $79 is the price of keeping one dispute out of your numerator. Whether it is worth paying depends entirely on a number the calculator has no access to: how much headroom you have left.

With a ratio comfortably under the line, $79 to avoid a count you can afford is money set on fire. Sitting near 150 basis points with an acquirer already asking questions, it is cheap, because the alternative is escalating fines, mandated remediation, and eventually not being able to process Visa at all.

This is why the refund-versus-fight question resolves differently for two merchants with identical order economics.

There is a fourth option that beats all three, and it is the one nobody models because it does not produce a dispute to measure. If the customer contacts you before they call their bank, a refund costs the order value and nothing else: no fee, no alert charge, no ratio entry, no evidence work. Everything that makes that conversation more likely, a descriptor they recognise, delivery updates that arrive, self-service that works, is what preventing avoidable disputes means in practice. Deflecting a dispute before it is filed is the fallback for the ones who skip you and go straight to the issuer.

The honest part

The recovery rate is doing most of the work in this model and it is the input most likely to be wrong. If you have been declining your weak cases, your measured rate is optimistic, and the model will tell you to fight things you will lose.

Proactive refunding also has a cost I cannot quantify for you or for anyone. Some share of the customers you refund would never have escalated, and you never learn which ones, because the counterfactual does not exist. You can bound it. You cannot measure it.

Ratio headroom has no honest dollar price either. VAMP fines are quantifiable, but the actual downside is account risk, and account risk is not a number. Anyone handing you a calculator that prices it has made something up.

Fee treatment matters more than it looks. If your processor returns the dispute fee on a win, the breakeven shifts and fighting looks better still. Get that detail from your merchant agreement before you build anything on it.

Build the sheet this afternoon. Four columns: order value, your own recovery rate, your dispute fee, and your alert fee if you have one. Then two formulas, expected cost of fighting and cost of refunding.

Run last quarter's disputes through it and find where the crossover falls in your catalogue. Then look at your ratio against the current threshold and decide what you are willing to pay per dispute to stay away from it. That second decision is a judgement call, and it should be made deliberately rather than one alert at a time on a Friday evening.