Chargeback ratio is one of the most watched metrics in payment fraud operations, and also one of the most misunderstood. Teams that are new to the number often calibrate their anxiety around the card network thresholds — 1% chargeback-to-transaction ratio for Visa's standard monitoring program — and conclude that anything below 1% is fine. That's the wrong frame. What's "normal" varies significantly by vertical, and what's "safe" depends on your trajectory as much as your current position.
This piece puts current chargeback benchmark ranges in context for the verticals where we see most of our fraud work: general e-commerce, travel and accommodation, and BNPL. We're not publishing these as authoritative industry statistics — we can't. What we can do is give you practitioner reference ranges based on what risk operations teams in these verticals actually operate against, and explain why the numbers differ so much by category.
How Chargeback Ratio Is Calculated (and Why Definitions Matter)
Before benchmarks mean anything, the formula has to be consistent. The card networks define chargeback ratio as: total chargebacks in the current month divided by total transactions in the prior month. That's not the same as chargebacks divided by current-month transactions, and it's not the same as a rolling 30-day window. These differences matter when you're close to a threshold because your denominator is slightly different from what you'd compute naturally.
Some teams also track "fraud-to-sales ratio" separately, which measures the dollar value of fraud chargebacks against gross sales volume. This metric is more sensitive to high-ticket fraud because a single $1,200 chargeback weighs more than twelve $100 chargebacks in fraud-to-sales terms but identically in transaction-count terms. Both metrics are relevant; card network monitoring programs tend to trigger off transaction-count ratio, but your internal risk ops posture should be watching both.
General E-Commerce: The Reference Baseline
Healthy general e-commerce merchants — meaning those with active fraud controls, reasonable delivery confirmation practices, and clean dispute history — tend to run chargeback ratios in the 0.3% to 0.6% range. This assumes a mix of product categories and customer acquisition methods. Pure digital goods (game credits, software licenses, streaming subscriptions) tend to run higher, in the 0.6% to 1.0% range, because there's no physical delivery to confirm and chargebacks are easy to file.
E-commerce merchants above 0.65% on a consistent basis are approaching the Visa VFMP standard tier trigger threshold. The zone between 0.65% and 1.0% is where most remediation conversations happen — you're not yet in the formal monitoring program, but you're accumulating chargeback volume that can tip you over in a bad month. Merchants above 1.0% have typically already received a monitoring program notification.
The floor is harder to establish than the ceiling. A low-fraud e-commerce merchant with good behavioral scoring and strong authentication can run at 0.1% to 0.2%. That range is achievable but not the norm — it typically requires active real-time scoring on every transaction, not just post-hoc rules.
Travel and Accommodation: Structurally Higher
Travel is one of the most chargeback-intensive verticals in card-not-present processing, and the reasons are structural rather than purely fraud-related. Travel purchases are high-value, often made weeks or months before service delivery, and cover categories (airfare, hotel deposits, package tours) where "service not as described" and "service not rendered" disputes are genuinely common and often legitimate.
Fraud chargebacks in travel tend to concentrate around account takeover at the booking stage and stolen card use for high-value reservations. But the dispute volume from non-fraud categories — cancellations, trip interruptions, airline flight changes — is high enough that even well-run travel merchants often see blended chargeback ratios of 0.7% to 1.2%.
This is one of the verticals where fraud-to-sales ratio matters more than transaction count ratio. A travel merchant processing $10 million in monthly bookings at a 0.8% transaction chargeback ratio might have fraud chargebacks representing only 0.3% of sales by value, with the remaining 0.5% coming from service disputes that the network doesn't classify as fraud. Keeping those dispute categories clean in your analytics is how you avoid falsely inflating your fraud signal or misdirecting remediation effort.
BNPL: A Different Measurement Problem
Buy-now-pay-later products have a chargeback dynamic that doesn't map cleanly onto card network metrics because the primary fraud vector in BNPL is application fraud at account creation — not disputed card transactions. A BNPL provider's fraud exposure shows up as first-payment default, charge-off after the first installment, and synthetic identity-driven credit abuse rather than the classical chargeback workflow.
That said, BNPL providers who operate as Visa or Mastercard issuers or who route through card rails for some transactions do generate chargeback exposure in the conventional sense. For those transactions, the benchmark range is typically 0.5% to 0.9%, skewing higher than general e-commerce because BNPL attracts higher-risk consumer segments by design — the product exists specifically to extend credit access to customers who might not qualify for traditional credit cards.
The more meaningful metric for BNPL risk ops is first-payment default rate: the percentage of approved accounts that miss their first installment payment. Healthy BNPL programs targeting established credit segments run this at under 2%. Programs with looser underwriting or aggressive customer acquisition in thin-file segments can see first-payment default rates of 4% to 8%, which is where the economics of the product start to break down.
Why Being Below Threshold Isn't the Same as Being in Good Shape
The most important thing about these benchmarks is what they don't tell you. A chargeback ratio of 0.5% in a vertical where healthy peers run at 0.2% is a signal worth investigating, even though 0.5% is comfortably below every monitoring program threshold. Conversely, a ratio of 0.9% in the travel vertical is elevated but not anomalous for the category.
Context-calibrated benchmarks also need to account for trajectory. A merchant who ran at 0.3% for twelve months and is now at 0.55% has a more urgent problem than a merchant who has run at 0.55% for three years with no upward trend. The card networks' monitoring programs respond to the current number; your internal risk ops should be responding to the rate of change.
We're not saying that keeping your ratio below the monitoring thresholds is a bad goal — it's a fine floor. But the teams that run consistently low fraud rates are generally not managing to a threshold. They're managing to a model quality standard: what does our precision look like at our current operating threshold, and is it getting better or worse? The chargeback ratio is an outcome metric that reflects the answer to that question, not a target to optimize directly.
The Measurement Infrastructure Question
Accurate chargeback benchmarking requires the ability to categorize disputes by reason code and trace them back to the original transaction's fraud score. This sounds obvious but is operationally non-trivial. Many PSPs and e-commerce platforms receive chargeback notifications that arrive without reliable linkage to the original transaction record, especially for transactions processed through third-party payment gateways with limited data pass-through.
At Txnworks we maintain a transaction log that preserves the original risk score alongside dispute outcomes when they come in through the API. This lets customers run analyses like "what was the risk score distribution of transactions that resulted in reason code 10.4 (card-absent fraud) chargebacks over the past 90 days?" — which is the kind of feedback that shows you whether your model is actually catching the right patterns or missing a systematic fraud type. Without that linkage, benchmarking is backward-looking accounting. With it, it becomes forward-looking model calibration.