In the last lesson we traced interchange, the largest single component of a card transaction's cost. Interchange is the part everyone argues about. But it is not the part the network keeps. Interchange flows from the acquirer to the issuer; the network sets the rate and routes the money, then takes a separate, smaller cut for running the rails. That cut is the scheme and network fees, and it is the line item most merchants and most product teams never actually read.
This matters because it is the part of the cost stack that goes directly to Visa and Mastercard, and it is growing faster than the part that does not. A UK Payment Systems Regulator review published in 2024 found scheme and processing fees charged to acquirers rose more than 30 percent in real terms between 2019 and 2024, with little change in the underlying service. If you are pricing a payments product or modeling unit economics, treating these fees as a rounding error is how your margin quietly erodes.
What scheme and network fees actually pay for
Interchange compensates the issuer for funding the transaction, carrying credit risk, and handling fraud and disputes. Scheme and network fees compensate the network itself for the infrastructure: authorization switching, clearing and settlement, the BIN system, fraud and security programs, tokenization, and a long tail of optional data and analytics services.
The mechanical distinction is the cleanest way to remember it. Interchange is a transfer between two banks that the network administers but does not keep. Scheme and network fees are revenue the network books for its own account. Same transaction, two different pockets.
Outside the US and Canada these are usually called scheme fees. Inside the US you will hear them split into two ideas: assessments (the percentage-based piece) and a set of named per-transaction fees. They are the same category of cost.
The two shapes: assessments and per-transaction fees
Network fees come in two shapes, and both hit nearly every transaction.
Assessments
The assessment is the headline percentage the network charges on volume. Visa's assessment is roughly 0.14 percent on credit and around 0.13 percent on debit; Mastercard's runs near 0.1375 percent on consumer credit, with a higher tier on transactions of $1,000 or more. These rates are small relative to interchange, which is why they get ignored, and they apply uniformly. There is no fine print that lets a high-volume merchant negotiate the assessment rate down the way they might shape interchange exposure through better data or routing.
Per-transaction network fees
The second shape is a flat fee per authorization or transaction, independent of ticket size. The two most common in the US:
- Visa's Acquirer Processing Fee (APF), about $0.02 per authorization. It charges on authorization, so it lands even when the transaction is later reversed or not settled.
- Mastercard's Network Access and Brand Usage (NABU) fee, $0.0195, applied to authorizations and refunds alike.
These look trivial in isolation. On a $4 coffee they are not. Two cents of fixed network fee on a $4 ticket is half a percent before interchange, before the acquirer markup, before anything else. Fixed fees are regressive against small tickets, and any business built on low average order value should model them explicitly.
A worked example
Take a $50 in-store US consumer credit purchase on a Visa card.
- Interchange (issuer's cut, network-set): roughly 1.65 percent plus $0.10, about $0.93. This leaves the network and goes to the issuer.
- Visa assessment (network keeps): 0.14 percent of $50, about $0.07.
- Visa APF (network keeps): about $0.02.
The network's own take on this transaction is roughly $0.09, against $0.93 of interchange it merely routed. The merchant's acquirer will bundle all of it, often add a markup, and present a blended rate. On a clean interchange-plus statement the $0.09 sits in its own column. On a tiered or blended plan it disappears into a single percentage, which is exactly why it goes unread.
Now run the same $0.09 across a business processing 10 million transactions a year. That is real money flowing to the network, separate from and on top of everything interchange already moved.
Why it is the part nobody reads
Three reasons the scheme line gets ignored, and why ignoring it is a mistake.
First, it is dwarfed by interchange on any single transaction, so it reads as noise. At volume it is not noise.
Second, the category has fragmented. Beyond assessments and the headline per-transaction fees, networks have added dozens of smaller line items tied to authorization behavior, cross-border activity, data quality, settlement timing, and optional services. Visa and Mastercard publish hundreds of distinct rate categories across their fee schedules, and the non-transactional fees in particular have multiplied. A typical acquirer statement can carry a long tail of network charges most merchants have never seen itemized.
Third, much of it is genuinely non-negotiable. Assessments and core network fees are set by the network and passed straight through; the merchant's leverage is on the acquirer markup, not the network's own rate. That makes the scheme line feel like weather, something that happens to you, which is precisely how it escapes scrutiny.
The leverage that does exist is behavioral. Several network fees are penalties or surcharges for how you transact: authorizing without settling, missing data fields, reauthorizing, or routing cross-border when you did not need to. Those are avoidable. Knowing which line items are fixed weather and which are self-inflicted is the entire practical payoff of reading this part of the bill.
Takeaway
Interchange is the cost the network moves between banks. Scheme and network fees are the cost the network keeps for itself, and they come in two shapes: a small percentage assessment on volume and flat per-transaction fees that bite hardest on small tickets. They are individually tiny, collectively material at scale, mostly non-negotiable on rate, and partly avoidable on behavior. Read the line. It is where the network's own margin lives, and it is rising faster than the interchange everyone else is watching.