False declines: what they cost and how to find them

Fraud is visible. It arrives as a chargeback with a reason code, a fee and a line in the monthly report. A false decline is invisible. A genuine customer tries to pay, is refused, and either tries something else or leaves. Nothing in your dashboard says that the refusal was wrong, and the customer rarely tells you.

This article defines false declines and where they come from, sets out what credible studies say they cost, shows how to find them in your own decline data, and lists the fixes that work without letting more fraud through. For which decline codes to retry, see our article on soft and hard declines; this one is about telling a wrong decline from a right one.

What is a false decline?

A false decline, also called a false positive, is a legitimate payment refused because it looked risky. It can happen in three places, and each needs a different response.

Where it happens	What you see	Who can change the outcome
Your own fraud screening, before authorization	A block that never reached the card network	You, by changing rules or allow lists
Manual review	An order held and then cancelled by your team	You, by changing review criteria
The issuer, at authorization	A decline code such as 05, 59 or 62	The issuer; you can only improve what you send and where

Payment providers draw the same line in their data. Stripe separates issuer declines from payments its fraud tool blocked, which show a network status of “not sent to network”. Adyen distinguishes its own risk refusals (FRAUD, when pre-authorization risk checks reach a score of 100 or more) from Issuer Suspected Fraud, when the issuer itself reported the payment as suspected fraud. If your reporting mixes the two, you cannot fix either.

What do false declines cost?

Estimates vary because nobody can see the purchase that did not happen, but the figures available put the cost in the same range as fraud losses.

• Datos Insights, formerly Aite-Novarica, puts the global average false decline rate at 1.51% of e-commerce sales, according to a 2026 Chargebacks911 summary.
• The Merchant Risk Council’s 2025 report, as cited in the same summary, found a global order rejection rate of 5.0% of orders, down from 5.8% a year earlier. Rejection is not the same as false decline, but it is the pool false declines come from.
• In the same MRC report, 14% of merchants reported false positive rates above 10%, while 22% reported 2% or less. About 19% of manually screened orders were declined after review.
• For comparison, the MRC’s 2026 report puts revenue lost to payment fraud at 3.2% of annual e-commerce revenue globally.

Two cautions. First, as Chargebacks911 points out, widely quoted global figures in the hundreds of billions measure the gross value of transactions falsely declined, not lost profit. Second, these are averages across very different merchants. The useful number is your own, and the next sections show how to estimate it.

The cost also goes beyond the order. A wrongly declined customer may buy elsewhere, and a first-time customer who is refused never becomes a repeat customer at all. For a merchant that pays to acquire customers through advertising, a false decline on a first purchase wastes the acquisition cost as well as the margin.

Why are false declines so hard to see?

Because most declines carry no useful reason. Stripe’s documentation says card issuers categorise most declines as generic, making the exact reason unclear, and that issuers discuss the specifics only with their cardholders. A Visa 05, do not honor, can mean a genuine fraud suspicion, an unstated limit, or a risk model that disliked a foreign merchant. The do not honor code is where most false declines hide.

Cross-border payments make it worse. Stripe notes that customers whose cards were issued in a different country from the merchant’s account may see more declines, and that some issuers do not allow purchases from certain countries at all. For a merchant acquiring from Europe and selling into Mexico, Kazakhstan or the UAE, a large share of the generic declines may be issuers treating a foreign merchant with suspicion rather than customers doing anything wrong.

How do you find false declines in your data?

You cannot label each decline as true or false. You can estimate the share with signals that genuine customers leave behind.

1. Split blocks from issuer declines. Count your own fraud-tool blocks and manual review rejections separately from issuer responses.
2. Group issuer codes by meaning. Use Stripe’s list of network codes as the reference: Visa 05 do not honor, 59 suspected fraud, 62 restricted card (card invalid in this region or country), 1A additional customer authentication required; Mastercard 83 suspected fraud and 05 do not honor, each with an advice code.
3. Work on payment intents, not attempts. One customer trying three times is one intent. Our guide on how to calculate approval rate explains why.
4. Look for the signals of a genuine customer. Then measure how often each group of declines shows them.

Signal after a decline	What it suggests
Same card approved within 24 hours, after the customer called the bank or authenticated	The first decline was probably false
Customer paid with another card or a local method in the same session	Genuine intent; the first decline may or may not have been false
Customer has months of approved, undisputed payments	A fraud decline on this customer is suspect
Declines cluster on one issuer, one country or one acquirer	A routing or relationship problem, not customer risk
Approved after 3-D Secure where the unauthenticated attempt failed	The issuer wanted more confidence, not a different customer

The share of declined intents followed by a successful payment from the same customer is your recovered rate. The share followed by nothing, from customers with good history, is the best proxy you have for lost customers.

Worked example: estimating false declines

Example figures, for illustration only.

An online electronics store reviews one month of card declines on 40,000 payment intents. It has 1,200 fraud-tool blocks and 4,800 issuer declines.

Group	Declined intents	Later approved for the same customer	Customer with 6+ months of good history
Own fraud-tool blocks	1,200	300	420
Issuer 05 do not honor	2,600	780	650
Issuer 59 and Mastercard 83 suspected fraud	900	90	180
Issuer 51 and 61 funds and limits	1,000	350	n/a
Other issuer codes	300	30	n/a

The fraud-tool row is the most actionable. A quarter of blocked customers paid later anyway and 35% had a long, clean history. Those rules are blocking good customers, and changing them is within the merchant’s control. The 05 row is the largest in absolute terms: 780 of 2,600 succeeded later, and the declines cluster on cross-border cards, which points to acquiring rather than customer risk. The suspected fraud row looks mostly genuine: few later approvals and few long-standing customers. Leave those declines alone.

At an average order of USD 120, the 420 blocked good customers alone represent about USD 50,400 of orders a month, before counting any repeat business.

How do you reduce false declines?

Match the fix to where the decline happened.

• Tune your own rules first. Review the rules that block the most payments, compare blocked customers’ history with approved ones, and use allow lists for customers with long, clean records. Stripe’s documentation describes adding a wrongly blocked customer to an allow list so later payments are not blocked by the same rules.
• Authenticate instead of blocking. For borderline payments, request 3-D Secure rather than refusing. Stripe says 3-D Secure can lower decline rates in countries that support it, and an issuer that returns code 1A is asking for exactly that.
• Send better data. Stripe recommends collecting the CVC and postal code to avoid declines from suspected fraud. Issuers decline what they cannot recognise.
• Use network tokens for saved cards. Stripe’s guide says network tokens help deliver higher authorization rates. Our article on network tokens and PSP tokens explains the trade-offs.
• Acquire locally where the volume justifies it. Stripe suggests processing locally in larger markets where cross-border cards see higher declines. Our article on how many PSPs you need per market covers how to weigh a local provider against its cost.
• Make the charge recognisable. A clear soft descriptor with a name the customer knows helps them recognise the charge on their statement instead of calling the bank about it.
• Retry and cascade by code. Generic and technical declines can be retried on another route; suspected fraud should not be. Our article on cascading and decline recovery sets out the rules.

What are the common mistakes?

• Treating every decline as the customer’s problem. Generic declines concentrated on one acquirer or one issuing country are a routing problem.
• Loosening fraud rules blindly. It swaps false declines for chargebacks. Change the specific rules that misfire.
• Cascading suspected fraud. Sending code 59 to a second acquirer until one approves is what card testing looks like.
• Measuring on attempts. Retries inflate the decline count and make the false decline problem look bigger or smaller than it is.
• Never asking the customer. A short message after a decline, suggesting another method or contacting the bank, recovers payments and tells you which declines were wrong.

How does Omno help find and reduce false declines?

Omno’s routing-decision log records the response code for every attempt, and the dashboard shows approval and decline reasons by provider, method, country and customer group, so generic declines can be traced to an issuing country or route. Risk controls can restrict providers by customer verification level and customer group, and card blocking stops cards linked to fraud before any provider sees them, which keeps blocks targeted instead of broad. The risk controls page lists the settings.