Your payment approval rate is lying to you

In advertising, there’s a rule that took me too long to learn: the campaign that looks best is almost never the campaign that performs best.

A high click-through rate may feel like success and even show-up well in a board deck. But, while click-through is the metric that is easy to see, it is almost never the thing that matters the most. The real conversion happens later, buried deep in data nobody pulled into the original report.  

That lesson is being relearned across business right now. The signals companies once trusted to tell them what customers wanted are getting more difficult to read, which forces a hard question: where is the signal a business can actually rely on, the kind it owns outright?

The doorman you didn’t know you had

Rory Sutherland, the advertising executive turned behavioral commentator, talks about what he calls the doorman fallacy. A hotel decides the doorman's job is to open the door. A door, which for management reasons, can be opened by a machine for a fraction of the cost. So, the doorman goes. What the hotel did not measure, and therefore never valued, was everything else the doorman did: hailing taxis, remembering returning guests, deterring trouble, setting the tone of arrival.  

They optimized the one function they could see and lost everything they couldn’t.

Most eCommerce enterprises run their payments stack exactly the same way. They watch one number: approval rate. Blended, aggregate, reported monthly. It’s the door being opened. And like the door, it’s the smallest part of what payments data is actually telling them.

What’s actually inside a transaction

Think about what a single transaction carries with it. The method the customer chose. The market they paid from. The currency. The time of day. Whether they had tried and failed before. The issuer's decline reason, if it declined. Multiply that by millions of transactions and an enterprise is sitting on one of the richest customer behavior datasets it owns. Most read one summary figure off the top of it and file the rest as exhaust.

Here is what that actually costs.  

In our own recent survey of US consumers, nearly 1 in 3 said they’ll abandon a purchase rather than use a payment method they don’t prefer. That isn’t indifference at checkout, but rather recoverable revenue sitting on the table. A blended approval rate will never show it to you, because in aggregate, that cohort is a rounding error.

At the individual level, it’s a pattern. The same segment declines on card, again and again, but would convert on a bank transfer or a local method if you surfaced it. The data already says so. The aggregate metric is built, by default, to hide it.

Where scales makes this a revenue line, not an optimization detail

This is where the doorman fallacy stops being a cautionary tale and becomes a commercial reality where scale changes everything. A small merchant with a few thousand customers can perhaps afford to treat payments as a single number. A multinational enterprise can’t. Its cohorts are large enough to be material. A behavioral segment that’s a rounding error for a regional retailer is a meaningful book of revenue for a global one.

The arithmetic is unforgiving. A 0.5% shift in converted transactions converted sounds like noise. On $10 billion in annual processing volume, it is $50 million. The larger the business, the smaller the percentage that still amounts to something the CFO cares about.  

That’s precisely why this argument is strongest for the businesses most likely to dismiss it. At enterprise scale, the difference between reading your payments data and glancing at it, is far more than just an optimization detail. It’s a revenue line.

Intelligence in the infrastructure, not bolted on top

Reading data at that resolution is not a job for a human and a dashboard. The volume is too high and it moves too fast. It is a job for intelligence built into the infrastructure itself, analyzing each transaction as a decision rather than counting them all as one.  

This isn't about AI bolted on to the stack, but real intelligence in the place where money moves. AI-adaptive routing reads issuer behavior, method preference, and customer history in real time and routes each transaction for the outcome most likely to succeed. In Nuvei's data that can deliver up to a 5% lift in approval rates. Not by simply relying on tracking the blended number, but by measuring and acting on everything underneath it.

Payment method preference is only the clearest example. The same data reveals which markets a merchant is underperforming in, where cross-border routing is dragging approvals that local acquiring would lift, which retry patterns signal a recoverable transaction versus genuine fraud. Each is a decision point. Each is invisible in the monthly summary. Each is legible in the detail.

The harder question

This is what “every payment, everywhere” actually means in practice. Not reach as a slogan, but reach as a discipline of reading each transaction as a decision worth getting right. Reach is the promise. Reading the data underneath it is how the promise pays off.

The question for an enterprise isn’t what its approval rate is. It is whether anyone has looked beneath it.  

The hotel that replaced its doorman wasn’t wrong to ask what he did. It was wrong to accept the easiest possible answer. The discipline that taught me to distrust the easy metric is now being forced to relearn it.  

Advertising is hunting for signal it can trust. Payments data has been generating exactly that all along – for every business willing to ask the harder question of its own numbers. Not how many transactions converted. But which didn’t, and why, and what the data has been telling you about them the whole time.  

‍

Nuvei is the growth infrastructure for every payment, everywhere. Enterprises use Nuvei to accept 720+ payment methods, acquire locally in 50+ countries, and route every transaction through AI that lifts approval rates by up to 15%, all on one integration. One intelligent system, built to scale.