Benchmarks (Operator Field Manual)
- Auth rate: Visa's 2025 analysis reports overall U.S. approval rates below 87%, using VisaNet data from April 2023 to March 2025. No public source breaks that down by channel or for small merchants. Compare card-not-present (online and phone) and card-present (in-person) baselines; ask your processor for peers
- Dispute programs: U.S. Visa VAMP excessive starts at 1.5% with 1,500 combined fraud reports and disputes, when your acquirer (the bank behind your processor) isn't Above Standard or Excessive; Mastercard ECM at 100 to 299 chargebacks this month AND 1.50 to 2.99% of last month's Mastercard sales; HECM is 300+ AND 3.00%+. Your processor can set a stricter limit. Get it in writing
- Fraud: MRC's global merchant survey: 3.5% of accepted orders later found fraudulent. Fed research: in-person fraud value on non-prepaid U.S. debit cards at big covered issuers, on dual-message networks, ran 0.142% in 2023. They measure different things
- Refunds: No published all-merchant norm. Read refunds and disputes together
- Segment everything: CP vs. CNP, BIN (the card number's opening digits), geo; trend over 4-8 weeks beats snapshots
Key Fact: Your processor can act before a network threshold. Ask: "What's my limit, how's it calculated, and when will you contact me?" See Network Programs for regional conditions, acquirer tiers and fees.
You're here to grade your numbers. Use anchors, not absolutes. Compare to yourself over time and by segment (card-present vs card-not-present, method, BIN, geo). Your baseline matters more than industry averages.
Sources checked: 21 October 2026. Where no figure is shown, use your own cohort baseline; the published figures below apply only to their stated populations.
How to read these numbers
- Segment everything. CP vs CNP, card brand, BIN/issuer, country, device. Aggregate numbers hide problems.
- Trend beats snapshot. Direction over 4-8 weeks matters more than any single week.
- Read metrics together. Auth, fraud, chargebacks, refunds, and alerts are interconnected.
- Match the source's population. A global merchant survey isn't a US domestic target. Check channel, geography and MCC (your business category code) before comparing.
- Date your thresholds. "Last verified" on any number you publish or operationalize.
Authorization Rate Benchmarks
NETWORK Visa reports overall U.S. approvals below 87% in its 2025 analysis, using VisaNet data from April 2023 to March 2025. Visa doesn't give a channel split or merchant-size breakdown for that figure, so don't grade either channel against it. It isn't your target.
Card-Not-Present (CNP)
NOT PUBLISHED Universal CNP performance bands.
- Pull decline codes by issuer, device and country
- Check fraud rules, 3DS failures (the extra checkout verification) and abandonment
- Test tokens and retries; watch fraud and disputes too
Card-Present (CP)
NOT PUBLISHED Universal retail approval bands.
- Check terminals and connectivity when approvals drop
- Separate chip, contactless and keyed entry
By Payment Method
| Method | Published comparison point | What to check |
|---|---|---|
| Cards (CNP) | Not published | BIN and issuer segments |
| Cards (CP) | Not published | Entry mode and terminal |
| Apple Pay/Google Pay | Not published | Each wallet against matched card-entry traffic |
| PayPal | VENDOR 95% suggested target, authorize-now, capture-later integration, updated 8 July 2026 | This isn't an observed average or ranking |
| ACH | Not published | Settlement and returns, separately from card approvals |
By Issuer Geography
NOT PUBLISHED Regional ranges or cross-border gap.
- Compare countries and issuers within the same channel
- Separate authentication failures from issuer declines
- Ask: "How do I compare with similar merchants here?"
Fraud Rate Benchmarks
By Transaction Type
| Type | Published context | What to check |
|---|---|---|
| Global e-commerce, across payment methods | REPORTED 3.5% of accepted orders later found fraudulent. Merchant survey, MRC 2026 report, Figure 27, p.36 | Accepted orders, not dollars lost |
| CP, U.S. non-prepaid debit | REPORTED 14.2 bps (0.142%) on networks like Visa and Mastercard; 5.1 bps (0.051%) on PIN networks like Star, NYCE and Pulse, by value in 2023. Kansas City Fed research briefing, 25 February 2026, Chart 3 | Large banks only; losses to all parties, not just merchants |
| Digital goods | Not published | Product and delivery method |
| Subscriptions | Not published | Initial versus recurring payments |
MRC surveyed global merchants in November and December 2025. Its trimmed average covers the preceding 12 months. It isn't a U.S. card target. Fed coverage: debit cards from banks with $10 billion or more in assets. Neither sets your target.
By Industry (CNP)
NOT PUBLISHED Industry ranges or ranking.
- Retail and high-ticket: separate products; read dollars beside order counts
- Digital goods: check access and delivery evidence
- Travel: separate booking date from service date
- Gaming and crypto: separate payment fraud from account abuse, trading and transfer losses
Fraud Detection Metrics
| Metric | Published context | What to check |
|---|---|---|
| False positive rate | Not published | Confirmed legitimate fraud blocks / All fraud blocks; keep unknowns visible. Use Fraud Metrics |
| Catch rate | Not published | Confirmed fraud missed, by rule and segment |
| Manual review rate | REPORTED 21% of orders screened manually, average. MRC merchant survey, Figure 31, p.40 | Queue size, staffing and outcomes |
| Review-to-block rate | REPORTED 18% of manually screened orders subsequently declined, average. MRC merchant survey, Figure 31, p.40 | A decline isn't proof of fraud |
Figure 31 has 564 respondents in the same wave. These aren't cutoffs. Check fraud type, existing tools and good customers blocked before tightening rules.
Chargeback Benchmarks
Chargeback Ratio Thresholds
NOT PUBLISHED Universal healthy, danger or crisis bands.
- Watch your trend against your processor's written limit
- Start remediation when you're approaching it
- Got a warning? Confirm its metric and deadline; use the emergency playbook
Network Program Thresholds
Visa counts monthly card-not-present VisaNet fraud reports plus disputes over settled transactions, subject to exclusions. Its merchant test applies when the acquirer isn't Above Standard or Excessive. Mastercard uses this month's chargebacks over last month's transactions.
| Network | Published monthly threshold |
|---|---|
| Visa VAMP, U.S. | NETWORK At least 1.5% and 1,500 combined fraud reports and disputes, since 1 April 2026. Visa public fact sheet |
| Mastercard ECM | REPORTED Acquirer documentation: 100-299 chargebacks AND 1.50-2.99%; HECM: 300+ AND 3.00%+. Braintree ECP guide, thresholds effective October 2019 |
| American Express | NETWORK US fee after three straight months over 1% (a location's chargebacks minus reversals, against its sales minus refunds). Fraud Full Recourse counts fraud by value against your Amex sales in every region: 0.9% AND USD 25,000 in a month brings a program notice, and three straight months under either line after it avoid the lower-tier penalties; at 1.8% AND USD 50,000 penalties start at the notice. Conditions and ECI detail: Amex's published chargeback and fraud lines. Source: Amex Merchant Rules, October 2026, sections 9.9, 9.11.4 and 12.4.4 |
| Discover | Get Discover's current numeric limit and its measurement terms from your acquirer in writing |
Program owner: Network Programs.
Amex runs a closed loop and prints its lines in its Merchant Rules (October 2026): a US chargeback fee line and all-region fraud lines, with a judgment call left for chargebacks or inquiries that look "disproportionate". For Discover, get your current monitoring limit from your acquirer in writing. Its 2011 regulations mention excessive returns or disputes as a security trigger, without setting a number; that's an old trigger, not proof of today's sole standard.
Chargeback Composition
NOT PUBLISHED Reason-share targets or diagnostic cutoffs.
- Fraud-coded disputes: review purchase and delivery evidence before assuming stolen-card fraud
- Unrecognized charges: check the billing descriptor (the name on your customer's statement) and purchase reminders
- Service and recurring billing: check complaints, renewal notices and cancellation requests
Note: A fraud reason code alone doesn't tell you who made the purchase. Friendly fraud can appear in fraud-coded disputes. Use your order evidence to classify cases; don't use an industry estimate as your expected share.
Refund Benchmarks
Refund Rate
NOT PUBLISHED All-merchant refund bands.
- Check access to eligible refunds; read disputes alongside refunds
- Investigate changes by product, delivery, season and support issue
Refund-to-Chargeback Ratio
NOT PUBLISHED Healthy ratio.
- Use the same period; keep both counts beside the ratio
- Check refund delays, duplicate credits and policy abuse
- Are earlier refunds resolving problems and reducing disputes? Measure it
Alert Performance Benchmarks
Ethoca/CDRN/Verifi Metrics
NOT PUBLISHED Universal match, refund or prevention targets.
- Track covered disputes against your issuer and card mix
- Check prior refunds and duplicates before issuing credit
- Measure prevented disputes before and after deployment
Real-Time Alert Response
| Product | Published response guidance |
|---|---|
| Ethoca through Kount | VENDOR Respond within 24 hours; prevention or credit eligibility isn't guaranteed after that. Kount guidance, updated 5 August 2026 |
| Verifi CDRN | VENDOR Participating sellers refund within 72 hours. Verifi RESOLVE |
| Verifi RDR | VENDOR Automatic decisions in real time, using seller rules |
Use your provider's agreed deadline. A reporting window isn't a reason to wait.
By Business Model
Subscriptions
| Metric | Published context |
|---|---|
| Initial auth rate | Not published |
| Recurring auth rate | Not published |
| Involuntary churn | Not published |
| Dunning recovery | VENDOR 12.7% median attempted recovery among 119 active U.S. B2B SaaS customers using Baremetrics Recover, May 2026. Baremetrics measurement |
| Subscription fraud | Not published |
Baremetrics excludes failures it didn't attempt and accounts with no attempts. It isn't recovery over all failures. Separate initial payments, renewals and retries; track payment-driven churn by billing cycle.
Digital Goods
NOT PUBLISHED Auth or fraud bands.
- Split approvals by issuer, country and authentication flow
- Dispute win rate: track it separately. Access and delivery records are your proof
Physical Goods
NOT PUBLISHED Auth, fraud or not-received-share bands.
- Investigate not-received claims by carrier and fulfillment route
- Dispute win rate: track it separately by reason code and carrier. For not-received claims, check your delivery records
Track your own win rate by reason code.
B2B
| Metric | Published context |
|---|---|
| Auth rate | Not published |
| Fraud rate | Not published |
| ACH return rate | Not published |
| Invoice payment terms | REPORTED 46 days average agreed terms from invoicing. Atradius merchant survey: 240 U.S. companies, fielded late Q2 to mid Q3 2025. 2025 report, printed pp. 10 and 13 |
Agreed terms aren't time taken to pay. Separate ACH return reasons from card disputes.
Keyed Transaction Benchmarks (CP)
NOT PUBLISHED Normal keyed-entry share.
- Investigate a rise by terminal and employee
- Check terminal issues, training gaps and CNP transactions recorded as CP
- Review unexplained keying for employee fraud; log why entry was needed
Benchmarks by Revenue Tier ($50K-$5M)
MRC defines SMB as $50K to under $5M annual e-commerce revenue (Figure 2, p.4). It doesn't publish subtier targets. Revenue isn't a performance ranking.
| Metric | Published context for the whole SMB band |
|---|---|
| Auth rate (CNP) | Not published |
| Fraud rate | REPORTED MRC merchant survey: 3.4% of annual e-commerce revenue lost to payment fraud; 3.3% of accepted orders later found fraudulent over the preceding 12 months. Trimmed averages, Figure 27, p.36 |
| Chargeback ratio | Not published |
| Refund rate | Not published |
| Processing cost | Not published |
| Manual review rate | REPORTED MRC merchant survey: 21% of orders screened manually, average; SMB base 127, Figure 31, p.40 |
| Reconciliation method | Not published |
These share the survey period above. Match reconciliation tools to count and complexity. Costs need inputs:
The fixed per-transaction cents punish small tickets. VENDOR Square's U.S. Free plan charges 2.6% + 15c for tap, dip or swipe in person, and 3.3% + 30c on Square Online and invoices. Its Online API rate is 2.9% + 30c, and manual entry is priced separately; the table below uses the tap, dip or swipe and Square Online figures. These effective rates are DERIVED: percentage rate + (fixed fee / ticket x 100).
| Average ticket | In person (tap, dip or swipe), 2.6% + 15c | Square Online and invoices, 3.3% + 30c |
|---|---|---|
| $25 | 3.20% | 4.50% |
| $50 | 2.90% | 3.90% |
| $100 | 2.75% | 3.60% |
A small online seller with a $50 average ticket is at 3.90% on those inputs. That isn't a negotiating failure, it's arithmetic.
VENDOR Helcim's U.S. pricing lists interchange + 0.25% + 7c card present for $100K-500K monthly card volume, with no monthly fee. On this site's published pass-through assumption of 1.03% + $0.178 card present, that's 1.28% + $0.248. These effective rates are DERIVED, using the same formula:
| Average card-present ticket | Helcim $100K-500K band, effective |
|---|---|
| $25 | 2.27% |
| $50 | 1.78% |
| $100 | 1.53% |
That's a model. Your card mix can differ. Compare your statement at the same ticket and channel; add other fees.
The volume trap: At low volume, a handful of chargebacks can spike your ratio above network thresholds. A merchant doing 500 transactions/month only needs 5 chargebacks to hit 1.0%. See network program thresholds for the exact numbers.
That ratio alone won't put you in the Visa or Mastercard programs above. Visa's also needs 1,500 fraud reports and disputes in the month, and its merchant test applies when your acquirer isn't Above Standard or Excessive. Mastercard's needs 100 chargebacks. Amex is different: its US fee can apply after a location's monthly chargebacks minus reversals exceed 1% of its sales minus credits for three straight months, with no count floor. Exactly 1.0% doesn't cross that line. Your processor can act sooner than any of them.
How to Use These Benchmarks
Step 1: Establish Your Baseline
- Choose a period covering your normal billing and sales cycle
- Segment by CP/CNP, geography, method
- Document as your baseline
Step 2: Compare to Benchmarks
- Does the source match your country, channel and definition?
- Which of your own segments changed?
- Where can you test a fix?
Step 3: Track Trends
Start with a routine you can keep:
- Weekly: Auth rate, fraud rate, chargeback ratio
- Monthly: All metrics, segmented
- Quarterly: Deep dive, re-baseline
Step 4: Set Alerts
NOT PUBLISHED Universal alert cutoffs.
- Auth: drops from your segment's normal range; check decline codes
- Fraud: rises from baseline; allow for reporting lag
- Chargebacks: approaches your limit; check counts too
- Refunds: shifts from normal; check products, season and support issues
Where This Breaks
- Mixing CP and CNP. Never combine in single number.
- Seasonality. Holiday, promo weeks skew baselines. Compare like periods.
- High-risk MCCs. Travel, gaming and crypto need their own bands. Ask your processor for a comparison.
- International traffic. Separate it from domestic traffic. Get a local acquirer before judging.
- Low volume. Small sample sizes create noise. Read the counts beside the rate; see the volume trap above.
Do next
Look up
- Reference - The whole reference section
- Processor Reporting Checklist
- Dispute Monitoring
Deeper
Establishing your baseline?
Comparing to industry?
Improving metrics?