Why Payment Approval Rate Can Be a Dangerous KPI

The first time I watched it happen, the number went up and the business got worse.

Approval rate is the cleanest-looking metric in payments. Out of every hundred payment attempts a business sends to the banks, how many come back approved. It fits in a box on a dashboard. It goes up and the room claps. It goes down and someone gets a call before lunch.

It is also one of the easiest numbers in this industry to move in ways that quietly cost money.

Picture two bouncers working the same door on the same night. The first waves everyone through. His count is spectacular — the room is full by ten. By midnight there are three fights, a broken table, and a conversation with the licensing authority. The second bouncer turns away anyone whose shoes he doesn’t like. No fights, no broken furniture, and a nearly empty room full of people who will drink somewhere else next Friday.

Both of them can hand you a number. Neither number tells you whether the night went well.

That’s approval rate. It counts what got through the door. It says nothing about who got through, who was turned away, or what happened to either group afterwards.

Let’s imagine

A company sells an online subscription. Nothing exotic, a €40 monthly charge. Every month it sends about 10,000 payment attempts to the banks — new sign-ups and renewals together.

8,600 come back approved. That’s an 86% approval rate. Leadership wants 90% by the end of the quarter, and someone puts it on a slide.

There are two easy ways to get there, and both of them are bad.

Route one: open the door wider. Loosen the fraud rules, stop screening the riskier traffic so hard, route more attempts through whichever bank connection says yes most often. Approval climbs to 91%. That’s roughly 500 more approved transactions a month — €20,000 that wasn’t there before, and the slide is green.

Then, six weeks later, the disputes start arriving. A chargeback is a customer’s bank forcibly pulling the money back — usually because the card was stolen, sometimes because the customer simply doesn’t recognise the charge. Say 120 of those new approvals turn into chargebacks. The business loses the €40, loses the service it already delivered, and pays a dispute fee on top. That’s most of the €20,000 gone before anyone has counted the staff hours spent fighting them.

And that’s the mild version. The card networks watch the ratio of chargebacks to sales. Cross their threshold and the business lands in a monitoring programme: higher fees, mandatory remediation, and a real chance that the acquirer — the bank that actually holds the merchant account and moves the money — decides this relationship isn’t worth the trouble. Losing your acquirer is not a bad month. It’s an existential event, and it is nowhere on the approval-rate chart.

Route two: shut the door tighter. Decline anything that looks even slightly unusual. Chargebacks fall, the risk team looks disciplined, and the acquirer is happy.

But of those 1,400 declined attempts, how many were fraud? Maybe a couple of hundred. The rest were a business traveller whose card triggered a location rule, a customer whose bank rejected the charge because the merchant looked unfamiliar, someone whose card expired last week and who would have happily entered the new one if anyone had asked.

Nobody asked. They got a red screen that said “declined” with no explanation — because the merchant genuinely doesn’t get one — and they went and bought the same thing somewhere else. That cost never shows up as a loss. It shows up as an absence, and absences don’t get dashboards.

A decline is not a data point, it’s a moment in someone’s day

This is the part the headline number can’t carry: declines are not interchangeable.

Some are what the industry calls soft — temporary, fixable. Insufficient funds this morning, fine this afternoon. A bank’s own fraud model twitching at a merchant it hasn’t seen before. A routing hiccup on one connection that a second connection wouldn’t have blinked at. These transactions are recoverable, and a retry a few hours later with a different setup often just works.

Others are hard. The card is cancelled, the account is closed, the card was reported stolen. Retrying those isn’t persistence, it’s noise — and repeatedly hammering a dead card is one of the fastest ways to make a bank trust you less.

A raw approval rate flattens all of that into one percentage. Two businesses can sit at 86% where one is losing recoverable customers by the thousand and the other is running about as well as its market allows. The number can’t distinguish them. It was never built to.

Then there’s what happens after the decline, which almost nobody measures. Does the customer get a clear path to try again — an obvious retry, an offer to use a different card or a payment method that fits where they live? Or do they get a dead end?

For a subscription business this is where the money leaks quietest. A renewal fails. The customer isn’t at their laptop; there’s no red screen, no moment of friction, nothing to react to. They simply stop being a customer, and they find out weeks later, if at all. From the inside it looks like ordinary churn. It was a payment problem wearing churn’s clothes.

The timing trap

Here’s the structural reason this keeps happening to good teams.

Approval rate is measured today. Its consequences land in six to eight weeks.

So the approval win is booked in one month, and the chargebacks, the fee increases and the risk conversations arrive in another. Often they arrive on a different dashboard, owned by a different team, discussed in a different meeting. The line connecting them exists, but nobody in the room is holding both ends of it.

In high-risk verticals the gap is wider still, because the swings are bigger, the fraud pressure is heavier, and the margin for error with acquirers is thin. That’s exactly where the pressure to show a good number is most intense — and where a good number bought carelessly does the most damage.

Better questions than “what’s our approval rate”

Not a checklist. Four questions I’d rather see on a slide than a single percentage.

What does our approval rate look like after chargebacks, refunds and fees are subtracted — approved revenue we actually kept, per attempt? That one number kills most of the gaming, because both bouncers score badly on it.

Why are our declines happening? Split them by reason and by market. The shape of that split is a to-do list. The percentage isn’t.

What share of recoverable declines do we actually recover, and how? If nobody knows, the honest answer is usually “less than we think.”

And how many customers hit a decline and never came back? Track that cohort for ninety days. It’s the most uncomfortable chart in the business and the most useful one.

I’ve watched a KPI look excellent while the company underneath it was slowly getting poorer, and the reason was never that someone was careless. It’s that the metric was easy, the timeline was long, and the cost arrived in a different room than the win.

If you’re running payments somewhere and your approval rate is up but the revenue doesn’t feel like it moved, I’d genuinely like to hear about it — that gap is the interesting part, and it usually has a specific, findable cause. Come tell me the ugly version of the number.