There is a number that should concern every CFO at a vertical SaaS company: $16 billion. That is the projected US ISV payment-processing revenue for 2025, according to McKinsey's January 2026 analysis, up from $6.5 billion in 2020 after five consecutive years of 20% annual growth.
The infrastructure to move money has been built. The integrations are live. The revenue is flowing.
What has not been built is the ability to understand what is happening inside it.
This is the embedded payments paradox: payments is simultaneously the largest revenue driver and the least visible line item on the P&L. The revenue is real and growing, but the operations behind it run on exported CSVs, manual reconciliation, institutional knowledge trapped in one or two people's heads, and fee complexity that makes independent verification functionally impossible without specialized tooling.
The infrastructure to move money has been built. The infrastructure to understand it has not.
The problem compounds at every layer.
The paradox is not a single failure. It is three compounding problems, each difficult independently and nearly impossible in combination.
Layer 1: The data is hard to access
Each payment processor reports in proprietary formats with different field names, fee structures, settlement timing, and API schemas. One processor reports Visa. Another reports VISA Credit. Settlement timing, decline codes, and currency handling all differ. Seventy percent of digitally advanced merchants now use multiple payment providers, according to 451 Research and S&P Global, but each provider's reporting is fundamentally single-tenant. No processor has an incentive to provide cross-processor analytics.
Finance teams at ISVs routinely spend more than 120 hours per month on payment reconciliation alone, according to SolvExia. Manual processes carry error rates around 4%, according to Ramp. Fragmented data turns a basic CFO question about payment costs into a week-long reconciliation project.
Layer 2: The data is hard to normalize
Even after extraction, making the data comparable across processors is a project, not a query. Fee categorization differs from one processor to another. Building a unified data model requires understanding both the schemas and the business logic behind every fee structure.
That logic keeps changing. Visa and Mastercard update interchange rates twice yearly and regularly introduce new programs. Every restructuring requires mapping updates that an internal system must absorb continuously. Payments adds industry-specific complexity to an enterprise data problem that is already difficult.
Layer 3: The data is hard to interpret
This is the layer most teams underestimate. Even with clean, normalized data, analysis requires domain expertise at the intersection of processor economics, data engineering, and portfolio strategy.
The US interchange system involves approximately 1,000 rate permutations. Visa alone publishes a 24-page interchange schedule with hundreds of individual rate entries. Add assessment fees, authorization fees, PCI compliance charges, gateway fees, batch processing fees, and cross-border surcharges, and a typical statement can contain 15 to 30 or more distinct fee lines.
Is a 92% authorization rate good? Is an effective rate of 2.4% competitive? Without a reference set of comparable ISV portfolios processing similar volumes in the same vertical, nobody knows.
Opacity gets more expensive every quarter.
The paradox would be merely inconvenient if payments were a stable, predictable cost center. They are not.
Total US credit and debit card swipe fees reached a record $187.2 billion in 2024, according to Nilson Report, up 70% since 2020. Average interchange rates rose to 2.35% from 2.26% the prior year. The fee environment is actively getting worse for platforms that are not watching.
Since 2010, card networks have introduced more than 40 new fee categories. Visa's Transaction Integrity Fee charges $0.10 per non-qualified transaction on top of the higher interchange rate. Mastercard expanded its Undefined Authorization Fee with assessment on all authorizations, including declines. Commercial and debit programs continue to be restructured.
The math at scale is punishing. A 0.1% overcharge on $10 million in monthly processing volume equals $120,000 annually. At $1 billion in processing volume, leakage of 0.2% to 0.5% represents $2 million to $5 million per year.
Under a standard blended plan, a debit transaction costing under 30 basis points and a premium rewards transaction costing over 200 can show the same customer-facing price. The platform cannot see which merchants are profitable, which are eroding margin, or where optimization opportunities exist.
The optimization opportunity is not a one-time fix. CMSPI reports that processors absorb an average of 55% of interchange cap savings, while 40% of savings erode within eight years through network fee increases. Platforms that do not monitor continuously fall further behind.
Every intuitive fix has a structural flaw.
"We will hire someone."
The person who combines deep processor economics, data engineering capability, and portfolio strategy rarely exists in a standard hiring pool. Even after a successful hire, first-year costs compound across executive compensation, recruiting, supporting engineering, and analytics infrastructure. The organization still begins without benchmark data, automated monitoring, or protection from a critical single point of failure.
"We will build it internally."
A smart team identifies the data problem and hires or reassigns a data engineer. Six months in, the team is still normalizing CSVs and building basic reconciliation. The scope expands to profitability analysis, benchmarking, anomaly detection, and optimization recommendations. A year later, the result is often a brittle tool that breaks when a processor changes an export format and covers only a fraction of the required analytics surface.
"Our processor dashboard is good enough."
Processor dashboards provide transaction history, payout tracking, and basic analytics, but only for their own slice. No processor has an economic incentive to provide cross-processor analytics or recommendations that illuminate how its pricing compares with alternatives. For ISVs using two or more processors, a single-processor view means managing the largest revenue line with a partial picture.
"We will bring in a consultant."
Specialist consultancies can produce excellent forensic diagnostics and substantial savings, but traditional engagements scale with headcount and rarely provide continuous monitoring. The static analysis begins aging the moment it is delivered.
The room for operating blind is shrinking.
Overall global payments revenue growth is decelerating. BCG projects 4.0% annual growth through 2029, down from 8.8% since 2019. Embedded payments may be capturing share from traditional channels rather than creating entirely new demand. The growth story is increasingly one of redistribution.
Potential interchange regulation and network competition could shift billions of dollars out of the margin pool that ISVs currently monetize. At the same time, processor consolidation puts more proprietary analytics and negotiating leverage on the other side of the table.
The implication is direct: when the total margin pool shrinks, extracting maximum value from every basis point becomes existential rather than incremental. ISVs with transaction-level visibility, comparable benchmarks, and continuous monitoring will retain margin. Those operating blind will watch it erode.
Margin compression makes optimization more valuable, not less. The $16 billion revenue line that nobody can report on is also the one where reporting matters most.
Visibility is an operating system, not a report.
The paradox resolves when the requirements are made explicit. Managing payment economics at an ISV with rigor requires each of the following:
- 01Multi-processor data normalizationA unified model that ingests transaction, fee, and settlement data from every processor and makes them comparable.
- 02Transaction-level fee decompositionPass-through interchange, assessments, processor markup, vendor costs, revenue share, and net income at the individual transaction level.
- 03Interchange qualification monitoringTracking which transactions qualify for target rates, which downgrade to higher-cost tiers, and why.
- 04Merchant-level P&L analysisKnowing which merchants are profitable, marginal, losing money, or at risk of churning.
- 05Comparable benchmarksA reference set of similar ISV portfolios so competitive rate questions have empirical answers.
- 06Continuous monitoringAnomaly detection that catches fee changes, downgrade spikes, processor errors, and revenue leakage as they happen.
- 07Payments domain expertisePeople who can interpret the data, recommend specific actions, negotiate with processors, and model changes before they are made.
No spreadsheet delivers all of this. No single hire delivers all of it. No processor dashboard delivers more than a fraction. No quarterly consulting engagement delivers it continuously.
The gap between what embedded payments requires and what platforms can actually see is structural, worsening, and central to the next generation of payment-enabled software companies.
This article draws on analysis and data from McKinsey, Nilson Report, CMSPI, SolvExia, Ramp, AutoRek, 451 Research and S&P Global, Payrails, Optimus Fintech, Glassdoor, TSG and Fiserv, Salesforce, Integrate.io, Spreedly, BCG, PaymentGenes, Deloitte, and Medallion Partners. Time-sensitive market figures should be read in the context of the cited reporting periods.
