Customers don’t dispute late orders because they forgot buying them. They dispute them because a shop that took the money and went quiet looks exactly like a scam — and under VAMP, that belief lands in your fraud ratio.

Peak planning tends to split into two workstreams that never speak to each other. Risk tightens rules, tunes models and staffs the review queue. Operations negotiates carrier capacity, stages inventory and sets cutoff dates. They report to different people, they are measured on different numbers, and they meet once a quarter.

That division made sense while Visa measured fraud and disputes separately. It stopped making sense this year.

What VAMP actually counts

The Visa Acquirer Monitoring Program replaced VDMP and VFMP, which tracked disputes and fraud as two programmes with two thresholds. VAMP collapses them into one ratio:

Fraud reports (TC40) plus all disputes (TC15), divided by settled transactions (TC05). Count-based, not value-based, and card-not-present only.

On 1 April 2026 the threshold for merchants in the US, Canada, the EU and Asia-Pacific dropped from 2.2% to 1.5%. Formal monitoring begins at 1,500 combined events per month. Fines run at $8 per violating transaction in any month above threshold — and because a transaction reported as fraud that later becomes a dispute lands in both buckets, a single order can cost $16.

Sit with the structure for a moment. There is one number. It has no column for this one was our carrier’s fault.

The delay is not what causes the dispute

Start with the carriers, because the baseline is worse than most merchants assume.

Outside peak, roughly 6–9% of parcels arrive late. During the 2024 holiday window that rose to 7–10%, and the forecast for the following peak was 10–12%. Express services offer less protection than you would hope: next-day and two-day air ran at 8–10% delays during peak. Paying for speed buys you a better average, not an exemption.

But a late parcel, on its own, does not produce a chargeback. Plenty of orders arrive late and nothing happens. What produces the chargeback is what the customer starts to believe while they wait.

The promised date passes. Tracking has not moved in four days, or never generated a scan at all. The support email has not been answered. And somewhere in there, the customer stops asking where is my order and starts asking a different question: was there ever an order?

They are not confused about what they bought. They remember the purchase precisely — the product, the price, the moment they clicked. What has changed is their theory of who they bought it from. They go back to the site and it looks thinner than it did a week ago. The reviews seem generic. The contact page is a form. They search the brand name with the word “scam” and the autocomplete obliges, because autocomplete always obliges.

So they call their bank and report it as fraud. Not as a tactic, and not as a lie. As the most accurate description they can give of what they now sincerely believe happened to them.

Here is the part worth sitting with. From the outside, a real business drowning in peak volume and a fake storefront are behaviourally identical. Both took the money. Both missed the date. Both stopped answering. The customer has no instrument that distinguishes an overwhelmed warehouse from a criminal one — except what you tell them, and by this point you have told them nothing for nine days.

Their issuer files a TC40. If the dispute proceeds, a TC15 follows. Visa classifies the result as first-party misuse: a cardholder claiming an unauthorised transaction on a purchase they made themselves. It puts friendly fraud at around 20% of all fraudulent disputes globally, and up to 30% for high-volume online merchants.

Miscategorised from Visa’s side. Sincere from the customer’s side. And from VAMP’s side, simply a number, sitting in the same ratio as the stolen cards.

Why your fraud team cannot fix this

Every lever that reduces a fraud ratio operates before authorisation: tighter rules, step-up authentication, more declines, a lower risk appetite. This dispute did not originate at authorisation. It originated in a warehouse, eleven days later, and was completed by silence.

There is a second-order effect that catches people out, and it runs backwards from intuition.

The ratio’s denominator is settled transactions. If you respond to a rising VAMP ratio by tightening and declining more, you shrink the denominator — while the numerator keeps arriving from orders placed weeks ago that are still working their way through delivery and dispute cycles. Tightening in December makes the January ratio worse, not better.

What to do with the weeks you have

The useful asymmetry here: once a parcel is late, you cannot make it early. You can stop looking like a scam at any point, and it costs almost nothing.

Publish cutoffs you will actually hit. Not the carrier’s stated transit time — yours, with peak slippage built in. If air freight runs 8–10% late at peak, a cutoff that assumes on-time performance is a promise you will break for roughly one order in ten.

Say it before they notice. On the product page, at checkout, in the confirmation. A customer told their order ships in nine days who receives it in nine days is satisfied. A customer promised three who receives it in nine is a dispute forming. Same parcel, same date, opposite outcome — you are managing the gap between expectation and reality, and expectation is the half you control.

Treat every update as proof you exist. This is the real function of a delay notification during peak, and it is why an honest “your order is running four days behind, here is the new date” outperforms silence so dramatically. It is not customer service. It is evidence that there is a company at the other end of the transaction. Visa’s own guidance to merchants is to send delivery notifications and encourage customers to come to you before going to their issuer.

Make the human reachable, and answer fast. An unanswered support ticket during a delay is the single strongest confirmation of the customer’s developing theory.

Talk to your acquirer about deflection. Disputes resolved through RDR or CDRN are excluded from the VAMP ratio, as are certain fraud reports qualifying under Compelling Evidence 3.0. If you are near threshold, the exclusions matter as much as the prevention.

And one structural note. Everything above is about preventing a belief from forming. Once it has formed, the only thing that reliably dissolves it is contact with a real person who knows about the order — which remains available long after the parcel is late, and works a great deal better than a carrier scan presented to an issuer in February.

This close to Black Friday, the fraud roadmap is mostly set. The logistics plan probably still has room.

Which is fortunate, because as of April they are scored on the same number — and so far only one of those teams has been told.