Your approval rate is one number covering five decisions

Two payment providers can look at the same month of transactions and report approval rates several points apart. Neither is wrong. They are simply counting different things.

On a hundred million dollars of annual card volume, a few points is millions in what appears to have cleared. Yet the number gets treated as settled fact. It is quoted in quarterly reviews, compared across providers in an RFP, and tracked on a dashboard as though it were a mere temperature reading.

An approval rate is an average of at least five separate decisions, each with its own failure mode and its own fix. Averaging them tells you how much revenue you recovered and very little about where the rest went.

Two providers, same data, different answers

Most comparisons break before they begin, because the denominator moves.

Some providers count every authorization message sent to an issuer, including the failed attempts that preceded a success. Others count the order: if it eventually went through, it counts once, as approved. A single transaction that clears on the fourth attempt is a perfect result under the second method and a 25 percent result under the first. Same customer, same sale, same four messages on the network.

Then there is what gets left out. Transactions blocked by fraud screening before they reach the issuer may or may not sit in the denominator. Authentication drop-offs, where a customer abandons at a 3D Secure prompt and no authorization request is ever generated, usually do not.

None of these conventions is dishonest on its own. But put two of them side by side and call it a comparison, and you are measuring accounting method rather than payment performance.

The customers the metric cannot see

Before any of that, there is a group the approval rate cannot see at all: the customers who never attempted a payment.

According to a recent Nuvei survey, close to one in three said they would abandon a purchase rather than pay with a method they do not prefer. Nearly two in five said card acceptance on its own is not enough for them. Those customers generate no authorization request. They never enter the denominator. They leave behind a checkout with an intact approval rate and no revenue.

This is why approval rate gains and flat conversion can appear in the same report. The metric describes the health of the transactions a merchant managed to start.

Five decisions, one number

A card payment moves through a sequence of decisions, and a failure at any one of them arrives on the dashboard wearing the same label.

Authentication comes first. A transaction pushed into a full 3D Secure challenge when an exemption would have applied loses customers at the prompt, and those losses are routinely attributed to the issuer instead. Nuvei Optimize addresses this with exemptions management, applying exemptions where the rules allow, and with soft decline handling, which initiates a 3DS flow automatically when an issuer requires authentication rather than treating the soft decline as the end of the conversation.

Routing comes next, and this is where the geography of a transaction starts to show up in the numbers. A French card processed cross-border through an international scheme behaves differently from the same card routed through Cartes Bancaires. The pattern repeats with Bancontact in Belgium and with debit routing in the United States. Direct scheme connections and local scheme routing exist for that reason: to match a transaction to the rail with the strongest record for that card, in that market, at that moment.

Then comes the authorization message itself. Issuers approve what they can read. Stale card-on-file credentials, thin data, an amount the account cannot cover in full: each produces a decline that has nothing to do with a customer’s willingness or ability to pay. Account updater keeps stored credentials current. Partial approval captures the balance available rather than surrendering the sale.

Recovery comes last, as it should.

The most common thing I see when a merchant sends over a month of decline data is that the failures cluster. They are rarely spread evenly across reason codes. One issuer, one market, one card type, or one authentication rule is usually doing most of the damage, and the headline rate had been averaging that cluster into invisibility for months.

Retries are a cost line, not a strategy

Retry logic is the easiest lever to pull and the most expensive to lean on.

Every attempt carries a processing fee whether or not it succeeds. Beyond that, the card networks now price retry behavior directly. Mastercard’s Transaction Processing Excellence program sets thresholds on declined attempts against the same card at the same merchant, with fees applied to attempts beyond them, and uses Merchant Advice Codes 03 and 21 to signal cards that should not be retried at all. Visa operates comparable limits across a rolling 30-day window. Retrying past those points costs money on top of the attempt itself.

These costs are hard to see, because failed attempts do not appear in a settlement file. They accumulate in the background and arrive at month end as a single line, detached from the transactions that produced them.

A provider running heavy retry volume can hold a strong headline number while the recovery is funded out of a merchant’s margin. There is a second cost too. Undifferentiated retrying spends the attempts a card is allowed, so when a genuinely recoverable failure arrives, the window has already closed.

The question worth asking is not whether a provider retries. It is whether the retry knows anything.

A soft decline for insufficient funds is a timing problem and responds to a scheduled reattempt. A CVV mismatch is a data problem. A bank outage is an infrastructure problem and calls for a different bank, not a different day. Nuvei Optimize separates these: auth messaging adjusts how a retry is presented to the issuer, CVV reattempts revalidate credentials, bank cascading moves the transaction to a backup bank rather than hammering the first one, and payment recovery offers the customer an alternative method when the card is not going to work at all.

Across those layers, Nuvei Optimize is built to lift approval rates by up to five points. That uplift comes from the sequence rather than from any single module, which is why the modules are activated selectively. Merchants switch on the ones that match where their transactions are actually failing.

What to ask instead

The useful diagnostic is not the headline rate. It is the decline distribution: what share of failures came from authentication, from routing, from data quality, from insufficient funds, from issuer risk models. That breakdown is actionable in a way an average never is, because each category has a different owner and a different fix. Real-time insight into where transactions fail, alongside direct issuer relationships that can escalate a pattern rather than absorb it, turns the metric from a scorecard into a working diagnosis.

Which changes the questions worth putting to a provider before comparing anything:

• Is the reported rate counting attempts or counting orders?
• What sits outside the denominator, and where do authentication drop-offs land?
• What does the decline distribution look like by reason code, by market, and by issuer?
• How many attempts, on average, sit behind one recovered transaction?
• Where do retry fees and network penalties appear on the invoice, and can they be reconciled against the reported rate?

The answers describe the payment stack. The headline number describes the reporting.

One name for every failure

A single percentage is a comfortable thing to report. It rolls up cleanly, it moves in a direction, and it fits on a slide. It also gives every failure the same name.

Somewhere inside that number is a French cardholder who was routed cross-border, a subscriber whose stored card expired eleven months ago, and a shopper who saw an authentication prompt and closed the tab. All three are counted as one thing: a decline. Until they are counted as three, none of them gets fixed.

Turn approval data into revenue

A headline approval rate can tell you whether performance moved. It cannot tell you why. To improve revenue sustainably, merchants need visibility into the full path to payment: the customers who drop out before authorisation, the authentication flows creating friction, the routes underperforming by market or card type, and the declines that are genuinely recoverable.

Nuvei helps merchants move beyond a single blended percentage. With Nuvei Optimize, teams can identify where transactions are failing and apply the right intervention, whether that means managing exemptions, improving authorization messaging, routing locally, keeping stored credentials current, or recovering payments without relying on indiscriminate retries.

The result is a payment strategy built around the reasons transactions fail, not just the rate at which they do.

Speak to Nuvei to understand the decline patterns in your own payment data and build a more efficient path from checkout to conversion.