Skip to main content

How We Compare Processor Costs

TL;DR
  • You can't rank an interchange-plus processor against a flat-rate one without assuming what interchange costs. Helcim and Stax against Square and PayPal, say. Every comparison makes that assumption. Almost none of them print the number.
  • Ours is 0.89% + $0.158 card-present and 1.36% + $0.152 online. Derived below from Visa's published US schedule and Federal Reserve payment data.
  • We add an assumed 0.14% + $0.02 for card network assessments. Nobody publishes those. It's flagged unverified everywhere it's used.
  • Three dials in this model aren't verified, and we name all three. The 0.14% + $0.02 assessment. The split of rewards credit into standard and premium. The shift in debit share between in-person and online. Sensitivity for each is below. None of them flips a verdict here.
  • We have both networks' schedules. Mastercard's live host returns HTTP 403, which reads as unreadable and isn't. The April 2026 edition sits in the Internet Archive, and we pulled it. Visa and Mastercard land on the same number at every bound that matters. 2.30% + $0.10 premium consumer credit card-present, 2.60% + $0.10 online, 3.15% + $0.10 non-qualified.
  • We got this wrong before. This site used to assume 1.80% + $0.10 card-present. That's no blend at all. It's roughly one card type, a consumer rewards credit card. Fixing it changed several conclusions.
  • If you disagree with a weight, change it and rerun. That's why the arithmetic is here.
  • Every pricing figure here also carries a confidence tier. So you can tell whether we read it on the vendor's page, got it from a named third party, or never found it. The tier system is below.

Every processor comparison assumes a number for interchange. This page publishes ours, and the arithmetic behind it.

On this page

Why this page exists

An interchange-plus processor quotes a margin over a cost it doesn't control. A flat-rate processor quotes one number with that cost already inside. Say which is cheaper and you've decided what the underlying interchange is.

Every "cheapest payment processor" article you've read made that decision silently. So did we, for a while. Worse, we made it differently on different pages. The result: comparisons that contradicted each other while every individual rate was correct.

This page fixes the assumption in one place. Every cost comparison here runs off it. Think a weight is wrong for your business? The arithmetic is below. Redo it.


How we tier every figure

Every price on this site carries one of four labels. The label tells you how much weight the number can hold. It's printed next to the number, not buried in a methodology note nobody reads.

TierWhat it meansWhat we publish alongside it
VENDORWe read it on the vendor's own published pageThe exact URL and the date we accessed it
REPORTEDA credible third party published itThe source name, the URL, its publication or update date, and what kind of source it is: reseller, review site, industry publication, merchant survey, regulatory filing or earnings call
DERIVEDOur own arithmetic on figures somebody else publishedThe inputs, the operation, and a statement that the company never published the result itself
NOT FOUNDWe looked and failedWhere we looked, and what to ask the vendor for instead

Why bother

A blank cell isn't neutrality. It's a dead end for a merchant building a budget. And it quietly rewards the vendors that publish least. A pricing page that says nothing shouldn't come out of a comparison looking cleaner than one that publishes a rate you can argue with.

The opposite failure is worse, and it's the common one. Print a reseller's number in the same typeface as the vendor's own, and the reader can't tell them apart. The fix isn't to publish less. It's to publish more and label it.

We can't find another vendor comparison that separates what it verified from what it repeated. That separation is the point of this system.

The source quality order we use

Best first:

  1. The vendor's own published page
  2. A regulatory filing or an earnings call
  3. An established industry publication
  4. A review site with a stated methodology
  5. A reseller quoting a rate it sells
  6. An undated blog

We never cite a source that sells the thing without saying so. Several figures here come from review sites that disclose referral commissions from the vendor they're reviewing. That disclosure gets repeated next to the number every time. Not once, in a footer.

Where sources disagree, we publish the range and name both ends. "Reported at 2.6% to 2.9% by X and Y as of June 2026" beats a bare number with no provenance. More useful, more honest.

Two rules that follow from this

A REPORTED figure is never dressed up as the vendor's own. Toast is the sharpest example here. Toast publishes 2.49% and 3.09% on its own shop pages. The "+ $0.15 per transaction" that every comparison table bolts on comes from NerdWallet and two SEO content sites. So we write it as two things. Toast publishes 2.49%. The fixed component is reported at $0.15.

A DERIVED figure is labelled as our arithmetic, not as a disclosure. Riskified's revenue over its GMV works out to 0.237%. But Riskified never published a take rate. It says only that revenue is "directionally correlated" with GMV. Adyen is the reverse case. It publishes its own take rate of 17.0 basis points in its annual report, so that one is VENDOR.

What changed when we applied this properly

Several cells that used to read "Not published" turned out to be fetch failures, not absences. The difference matters:

  • Clover renders its pricing client-side, so a plain HTTP request returned a page shell. Clover publishes a full vertical-by-vertical rate card.
  • Toast returns HTTP 403 to automated fetches. Served to a browser, its plan cards still render client-side. Two retrieval problems stacked into one wrong verdict.
  • Square publishes rates in plain HTML and plan prices in JavaScript. That's why an earlier pass found one and missed the other.
  • Worldpay puts an indicative UK rate card behind a cross-host redirect. A naive fetch misses it.
  • Lightspeed's implausible "1.5%" was an extraction artefact from a summarising fetch. The real published rate is 2.6% + 10c.

The lesson is now part of the method. A single failed fetch is never evidence that a vendor publishes nothing. Before a cell may say NOT FOUND, we try a browser user agent, a JavaScript-rendering pass, regional variants, help-centre and support articles, PDF rate cards, sitemaps and archived copies. Checkout.com and Global Payments survived all of that. The Details panel on each says exactly what was tried.


The short version

Card presentOnline
Assumed interchange0.89% + $0.1581.36% + $0.152
Assumed network assessments (unverified)0.14% + $0.020.14% + $0.02
Total pass-through cost1.03% + $0.1781.50% + $0.172

Those are blends. One card isn't a blend. The published spread runs from 0.05% + $0.21 on regulated debit to 3.15% + $0.10 on a non-qualified transaction, and the full card-type table is on the interchange page. Got a lopsided mix? A B2B counter on commercial cards, a bar on mostly debit? Use your band, not ours.

What that means in practice:

ChannelTicketInterchange onlyFull pass-throughEffective rate
Card present$15$0.29$0.332.22%
Card present$50$0.60$0.691.39%
Card present$150$1.49$1.721.15%
Online$25$0.49$0.552.19%
Online$75$1.17$1.301.73%
Online$200$2.87$3.171.59%

The effective-rate column runs off the full pass-through. That's the figure every comparison here uses. The interchange-only column sits beside it, so you can see how much of the total is the unverified network assessment. Two cents plus 14 basis points, everywhere.

Notice how much the effective rate moves with ticket size. That's why a bare "X is the cheapest processor" tells you nothing. Ranking at a $15 ticket and ranking at a $150 ticket are different questions with different answers.


Where the rates come from

Every interchange figure comes from Visa USA Interchange Reimbursement Fees, 18 April 2026. Visa publishes it openly. Verified 2 August 2026.

Card-present rates used

Card typeVisa fee programRate
Consumer debit, regulatedRegulated Visa Check Card0.05% + $0.22
Consumer debit, exemptCPS/Retail, Debit0.80% + $0.15
Consumer prepaid, exemptCPS/Retail, Prepaid1.15% + $0.15
Consumer credit, no rewardsProduct 2, All Other Products1.51% + $0.10
Consumer credit, standard rewardsProduct 2, Traditional Rewards and Visa Signature1.65% + $0.10
Consumer credit, premium rewardsProduct 2, Signature Preferred2.10% + $0.10
Commercial and businessCommercial Card Present 2.50%, Business Product 2 around 2.10%2.26% + $0.10

Online rates used

Card typeVisa fee programRate
Consumer debit, regulatedRegulated Visa Check Card0.05% + $0.22
Consumer debit, exemptCPS/e-Commerce Basic, Debit1.65% + $0.15
Consumer prepaid, exemptCPS/e-Commerce Basic, Prepaid1.75% + $0.20
Consumer credit, no rewardsProduct 1, All Other Products1.89% + $0.10
Consumer credit, standard rewardsProduct 1, Traditional Rewards and Visa Signature2.05% + $0.10
Consumer credit, premium rewardsProduct 1, Signature Preferred2.50% + $0.10
Commercial and businessCommercial Card Not Present 2.70%, Business Product 1 around 2.85%2.79% + $0.10

Correction, 2 August 2026. The premium rewards row used to read 2.47% online and to name both Signature Preferred and Infinite. Neither was right. Visa publishes Signature Preferred Product 1 at 2.50% + $0.10. And 2.47% sits below both tiers, so it wasn't a blend of them either. The row now uses Signature Preferred alone, at the published rate.

Visa's top consumer tier, Infinite Spend Qualified, is higher still. 2.30% + $0.10 card-present, 2.60% + $0.10 online. Mastercard's World Elite lands on the same two numbers. We keep the premium bucket on Signature Preferred anyway. Infinite volume is small enough that folding it in moves the blend by less than a basis point. And we'd rather publish a tier you can look up than a weighted average you can't check. The correction moved the online blend from 1.359% to 1.362%. That still rounds to the 1.36% published everywhere on this site.

The regulated debit rate is a federal cap, not a network choice. Regulation II sets it at "$0.21 plus 0.05 percent multiplied by the value of the transaction, plus a $0.01 fraud-prevention adjustment". It binds every network identically. That matters below.

How we read Mastercard, and what it says

The blend above is built on Visa's tables. We checked it against Mastercard's, and the two agree.

Mastercard's schedule looks blocked and isn't, and the mistake is easy to repeat. The live host returns HTTP 403 to automated requests on both mastercard.com and mastercard.us. That's a retrieval failure, not an absence. The April 2026 edition sits in the Internet Archive and comes back cleanly:

http://web.archive.org/web/20260616151957id_/https://www.mastercard.com/content/dam/mccom/us/business/documents/merchant-rates-2026-2027-april2026.pdf

That's "Mastercard 2026-2027 U.S. Region Interchange Programs and Rates", effective 17 April 2026, snapshotted 16 June 2026, 286KB. Retrieved and extracted 2 August 2026.

What it shows next to Visa:

BoundVisa, 18 April 2026Mastercard, 17 April 2026
Regulated debit0.05% + $0.210.05% + $0.21
Premium consumer credit, card present2.30% + $0.10 (Infinite Spend Qualified)2.30% + $0.10 (World Elite Merit III Base)
Premium consumer credit, online2.60% + $0.10 (Infinite)2.60% + $0.10 (World Elite Full UCAF)
Non-qualified3.15% + $0.103.15% + $0.10
Highest rate published anywhere in the document3.15% + $0.103.30% + $0.10 (small business Standard, Level 5)

Two independently published schedules landing on the same three numbers is about as good as corroboration gets. Nothing to average, nothing to hedge between. The debit floor matches because Regulation II sets it federally, not because either network chose it.

Note the last row. Neither document contains a 3.5% rate anywhere. We checked every rate at or above 3.00% in both. The "up to 3.5%" that circulates in fee articles is a processor's keyed-entry price. Clover 3.5% + 10c, Square 3.5% + 15c. Somebody copied it into an interchange table.

One input we didn't re-fetch. The Federal Reserve Payments Study transaction counts behind the 63/37 debit-credit split carry over from the earlier pass. They weren't re-verified against the source document this session. Everything else on this page was read fresh on 2 August 2026.


The card mix

Why we weight by transaction count, not by dollars

For a merchant with average ticket A, the cost of a transaction is the share-weighted sum of (rate x A + fixed fee) across card types, where the shares are shares of transactions.

Weighting by dollar value is the wrong operation. It's also a common error. It over-weights credit cards, which carry bigger tickets. And it mishandles the fixed cent component entirely, because a $0.22 debit fee is charged per transaction, not per dollar. Blend by value and you overstate interchange every time.

The transaction shares

From the Federal Reserve Payments Study, CY2024, released July 2026:

Card typeTransactions (billions)Value ($ trillions)Average ticket
General-purpose credit64.36.17$96
Non-prepaid debit99.34.34$44
General-purpose prepaid11.80.43$36

That's 63% debit and prepaid, 37% credit by transaction count. The Fed says the same thing independently. Debit was 64 percent of all card payments in 2024 by number.

We exclude private-label credit and EBT cards. Store cards and benefit cards don't carry general-purpose network interchange at an ordinary merchant.

Two more splits, both from primary sources:

  • Within debit, the Fed's 2024 Regulation II data gives 60.7% regulated and 39.3% exempt by transaction count. Regulated means the issuer holds over $10 billion in assets. That's most large banks.
  • Within credit, the CFPB's December 2025 report to Congress says it plainly: "Overall, 92 percent of general purpose card spending was accounted for by rewards cards in 2023 and 2024."

That last figure surprises people. It's why casual estimates of interchange run low. Nearly all US credit card spending is on a rewards card.

The blend, card present

BucketShare of transactionsRateContribution
Regulated consumer debit38%0.05% + $0.220.0190% + $0.0836
Exempt consumer debit and prepaid25%0.84% + $0.150.2100% + $0.0375
Consumer credit, no rewards3%1.51% + $0.100.0453% + $0.0030
Consumer credit, standard rewards24%1.65% + $0.100.3960% + $0.0240
Consumer credit, premium rewards7%2.10% + $0.100.1470% + $0.0070
Commercial and business3%2.26% + $0.100.0678% + $0.0030
Total100%0.885% + $0.158

The blend, online

BucketShare of transactionsRateContribution
Regulated consumer debit33%0.05% + $0.220.0165% + $0.0726
Exempt consumer debit and prepaid22%1.66% + $0.1550.3652% + $0.0341
Consumer credit, no rewards3%1.89% + $0.100.0567% + $0.0030
Consumer credit, standard rewards30%2.05% + $0.100.6150% + $0.0300
Consumer credit, premium rewards9%2.50% + $0.100.2250% + $0.0090
Commercial and business3%2.79% + $0.100.0837% + $0.0030
Total100%1.362% + $0.152

The online profile moves debit from 63% to 55% of transactions. Card-not-present volume skews toward credit. The Payments Study publishes no channel split, so that shift is a judgement call rather than a sourced figure. It's the third of our three unverified dials. And the softest one that actually moves an answer.


Does it check out?

The debit half can be tested against an independent measurement the model never saw.

The Fed reports actual 2024 debit interchange. Regulated issuers averaged $0.23 per transaction. Exempt issuers $0.51. All debit $0.34, on an average ticket of $46.32.

Running the Visa rates we use at that same ticket:

Our modelFed observedGap
Regulated$0.243$0.231.3 cents
Exempt$0.521$0.511.1 cents
Blended$0.352$0.341.2 cents

Within about a penny on the largest component of the blend. The small overshoot is expected. Not every issuer certifies for that $0.01 fraud-prevention adjustment.

This is the check that convinced us 0.89% was right and 1.80% was wrong. The old figure made debit a rounding error in a small merchant's mix. The Fed's own data says otherwise.


The parts we're least sure about

Three inputs are unverified. We name all three here rather than bury them. Each one carries a sensitivity test, so you can see what would have to be true for a verdict to move. No verdict on this site rests on any of them.

Network assessments: 0.14% + $0.02, unverified. Card networks charge assessments and per-authorization fees on top of interchange. Neither Visa nor Mastercard publishes them on any public page we could read. They matter here for one reason. An interchange-plus merchant pays them visibly, on top. A flat-rate merchant pays them invisibly, inside the headline rate. Leave them out and you flatter interchange-plus. Helcim's pricing page confirms the pass-through happens, "we pass exactly what banks and payment networks charge you", without attaching a number. Stax never mentions network fees. We carry 0.14%, the high end of plausible, so the answer we reach isn't flattered. Move this dial between 0.10% and 0.25% and no conclusion on this site changes.

The standard-versus-premium rewards split, unverified. We assume 24% of rewards spending sits on Signature Preferred or Infinite tier cards. That's anchored on the CFPB's finding that 16% of general-purpose accounts carry an annual fee, adjusted up because those cardholders spend more. Move it between 10% and 40% and the card-present blend shifts 4 basis points. Nothing turns on it.

The online debit share, unverified. Explained above. We move debit from 63% of transactions in person to 55% online, and the Payments Study publishes no channel split to check it against. This is the softest input that actually matters. Test the debit share across a 50% to 75% range and the card-present blend swings between 1.065% and 0.721%. Helcim still beats Square Free by $160 to $220 a month in the reference cell, at either end.

And one thing we left out on purpose. Visa runs a Small Merchant Fee Program with genuinely lower credit rates. It covers merchants under $280,000 a year in Visa consumer credit sales, on pages 12 and 13 of the schedule. A qualifying merchant pays 1.29% + $0.10 card-present on a non-rewards credit card, against 1.51% + $0.10. We don't bake it into the blend. It applies only to Visa and only to consumer credit, so the realised benefit is roughly 4 basis points card-present. It's real and it's worth knowing about. Leaving it out keeps our published number the conservative one.


What this changed

Correcting the assumption moved conclusions, not just decimals. On a $50 card-present ticket, the old assumption overstated pass-through cost by 79 basis points. On $20,000 a month that's $159 wrongly charged to the interchange-plus side of every comparison.

Worked example, card present, $50 ticket, $20,000 a month, which is 400 transactions:

  • Pass-through: 1.03% x $20,000 = $206.00, plus $0.178 x 400 = $71.20, so $277.20
  • Helcim entry band markup: 0.40% x $20,000 = $80.00, plus $0.08 x 400 = $32.00, so $112.00
  • Helcim total: $389.20, an effective rate of 1.95%
  • Square Free: 2.6% x $20,000 + $0.15 x 400 = $580.00, or 2.90%

Helcim saves $190.80 a month. The old assumption put the same saving near $10. That's close enough to nothing that it would have changed nobody's mind. That was the error.

The other conclusions that moved:

  • Helcim wins more scenarios on the corrected blend, and by much wider margins.
  • Helcim isn't cheapest "at any volume." At a $15 card-present ticket it loses to PayPal Zettle at every volume up to $500,000 a month. Zettle's $0.09 fixed fee is the lowest published here, and small tickets reward a small fixed fee. Zettle leads by $13.00 a month at $5,000, $52.00 at $20,000 and $43.33 at $100,000. It flips at $500,000 a month. Helcim enters its $500K-1M band at 0.20% + 6c and its crossover ticket drops to $13.96.
  • Stax's crossover numbers were mislabelled. What we published was the point where Stax passes flat-rate pricing. We presented it as the point where Stax becomes cheapest overall. Helcim is usually still cheaper there.

See Processor Comparison for the full scenario tables and the verdict for each combination of channel, ticket size and volume.


What we won't rank

A model that ranks everything is guessing about something. These processors stay out of the cost verdicts on this site, with the reason stated each time.

The list is much shorter than it was. Five entries turned out to be retrieval failures rather than genuine absences. What remains:

ProcessorWhy it's still excluded from a cost verdict
Checkout.comQuote-only, and verified as such. Two fetches, a full sitemap scan of 1,552 URLs and a third-party sweep produced no figure of any kind
Global PaymentsNo rate published. Verified with a browser user agent to rule out a 403-shaped false absence. Its own scale disclosure is "trillions in payments volume", which is not a number you can divide by
NuveiNo rate published, and it stopped filing in November 2024, so the derived 56 bps blended rate from its final Q2 2024 filing is the last public anchor and is not comparable to anyone else's take rate
AdyenAdyen labels its own published rates indicative, and requires a minimum monthly invoice whose amount it doesn't publish
WorldpayPublishes an indicative UK card in GBP. No US rate card exists, and the page doesn't state whether interchange and scheme fees pass through at cost
MonerisCanadian acquirer, all figures in CAD
Stax above $250,000 a yearSubscription becomes "$199+" with no published ceiling
Lightspeed, online onlyCard-present is published at 2.6% + 10c. No online, keyed or card-not-present rate is published anywhere, so only the card-present side can be ranked
Clover, through a resellerClover-direct rates are published and rankable. A reseller quote is not, and Clover says so itself: contract terms and fees "can vary based on your service provider"

Now rankable, having previously been on this list:

ProcessorWhat closed the gap
StripeUS rates confirmed from a US connection: 2.9% + 30c online, Terminal 2.7% + 5c
Shopify PaymentsUS rates confirmed: 2.9% / 2.7% / 2.5% + 30c online, 2.6% / 2.5% / 2.4% + 10c in person, plans $39 / $105 / $399 monthly
Square Plus and PremiumPlan prices confirmed at $49 and $149 per location. Square renders rates in HTML and plan prices in JavaScript
Toast2.49% Traditional and 3.09% Pay-as-you-Go published on Toast's own shop pages. The + $0.15 fixed component is reported, not published, and is labelled that way everywhere it appears
Clover, direct2.3% + 10c restaurant, 2.6% + 10c retail entry plans, 3.5% + 10c keyed and online
Lightspeed, card-present2.6% + 10c, US region, USD

For the ones still excluded, the useful output isn't a number we invented. It's the reported range where one exists. Plus what to ask for and what to push toward, which sits on each processor's page.


Reviewing and reproducing this

The blend is six multiplications and two additions per profile. Every rate is in the tables above. Every weight is stated. If your business genuinely differs, change the weights and rerun. A B2B supplier taking mostly commercial cards. A bar with a $12 average ticket and almost all debit. The method survives. Only the answer moves.

Sources, all accessed 2 August 2026:

  • Visa USA Interchange Reimbursement Fees, effective 18 April 2026, usa.visa.com. Fetched directly, 877KB PDF
  • Mastercard 2026-2027 U.S. Region Interchange Programs and Rates, effective 17 April 2026, mastercard.com. Live host returns HTTP 403; retrieved from the Internet Archive snapshot of 16 June 2026, 286KB PDF
  • Federal Reserve, Regulation II average debit interchange fee data (2024), page last updated 19 December 2025, federalreserve.gov. Covered issuers $0.23 on a $48.95 average ticket, exempt issuers $0.51 on $42.27, all issuers $0.34 on $46.32
  • Federal Reserve Payments Study, CY2015-24 top-line data, released July 2026, federalreserve.gov
  • Consumer Financial Protection Bureau, The Consumer Credit Card Market Report to Congress, December 2025, consumerfinance.gov
  • Helcim pricing, helcim.com/pricing
  • Stax pricing, staxpayments.com/pricing

Review schedule: both networks republish in April and October. So the model gets re-derived within a month of each release. Next check is October 2026. It also gets re-derived the moment the Federal Reserve changes the Regulation II debit cap. That's the single largest input in the blend. The Fed's debit figures are 2024 actuals and update annually.