The $70 Phone Call: How Fintechs Are Killing the Income Verification Tax

Here's a number worth sitting with: a community bank can pay more to verify a borrower's income than it costs to buy a tank of gas, a nice dinner, or a month of streaming services. Sometimes all three.

The Work Number, Equifax's employment verification database, charges lenders $55 to $70 per standard verification order. For a two-borrower mortgage file requiring verification at underwriting and again before closing, that's $140 to $280 per loan, before accounting for the cases where verification fails and someone has to pick up the phone. The Community Home Lenders of America (CHLA) formally asked regulators to scrutinize the pricing. An antitrust class action filed in 2024 documented a 272% price increase since 2012, when the same verification cost $17.85. Equifax has a near-monopoly on automated employment data for GSE lending. They know it.

Meanwhile, Plaid returns income verification results in seconds. For somewhere between $1 and $3 per pull.

The Work Number charges $55–$70 per verification. Plaid does the same job in seconds for $1–$3. That gap is not regulatory necessity. It's infrastructure debt.

That gap is not a feature of borrower complexity or regulatory necessity. It's the cost of doing business on aging infrastructure, and it's one reason why fintechs serving personal loan and auto markets are lapping community banks on speed and economics, even when the underlying borrower population is identical.

What Plaid Actually Does

Plaid Income is not one product. It's three verification paths that route automatically based on what data is available.

The first path is payroll connectivity. The borrower authenticates directly into their payroll account (ADP, Gusto, Workday, and roughly 250,000 other employers covering an estimated 80% of the US workforce). Plaid pulls structured data from the same source as the pay stub: gross income, pay frequency, employer name, job title, employment start date. The result is instantaneous once the user logs in.

The second path is bank income analysis. The borrower connects a checking account. Plaid's machine learning pipeline runs against up to 24 months of transaction history, extracts recurring income streams, classifies them into 13 categories (salary, gig economy, pension, long-term disability, and more), and returns a structured income report. For straightforward W-2 earners with direct deposit, this takes about 11 seconds. That's the actual measured response time from a published case study with Purpose Financial, a high-volume consumer lender.

The third path is document upload: pay stubs, W-2s, 1099s. Plaid's OCR and fraud detection layer checks more than 30 signals including photoshopped paystubs and name alterations, assigns a fraud risk score, and extracts the relevant figures.

In April 2026, Plaid announced it rebuilt its income classification layer using a transformer-based large language model trained on financial transaction data, claiming a 48% improvement in accuracy and 86% precision on earned income classification. The self-employed income category, historically the hardest to classify correctly because of irregular payment patterns, is now a named category for the first time.

The product also operates as a consumer reporting agency under Plaid Check, meaning income reports are FCRA-compliant and can be used in credit decisioning with adverse action notice requirements.

What a Community Bank Actually Does

The traditional income verification chain for a mortgage involves several sequential steps, most of which cannot be parallelized and several of which involve waiting on parties the lender cannot control.

The borrower submits two recent pay stubs, two years of W-2s, and one to two years of federal tax returns. The lender orders tax transcripts from the IRS using Form 4506-C through the Income Verification Express Service (IVES). The IRS targets 10 business days for processing. That's two calendar weeks, minimum, assuming nothing goes wrong.

Here's what goes wrong: the IRS rejects 30–40% of all 4506-C requests. When a form is rejected, the IRS returns it without an error code. The lender has to diagnose the problem, correct the form, and resubmit from scratch, restarting the 10-business-day clock. A single rejection cycle can consume the entire back half of a 30-day purchase closing window.

Meanwhile, the lender also needs to verify employment. If the employer is in The Work Number's database, an automated report comes back instantly. But The Work Number's coverage is inconsistent for small businesses, nonprofits, newer companies, and gig platforms. When an employer isn't in the database, a human initiates outbound verification: independently validate the employer's phone number and physical location, call HR, verbally confirm employment status and job title, and obtain income details in a separate written format. Fannie Mae requires a verbal VOE completed within 10 business days of the loan note date regardless, which means lenders cannot do this step early. It becomes a closing-week bottleneck on every single file.

For self-employed borrowers, add a year-to-date profit and loss statement to the stack. Underwriting complexity scales up from there.

Total income verification cost per mortgage file: IRS transcript ($40+ per order, plus resubmission costs), Work Number ($55–280 depending on borrower count and timing), staff time for manual calls and document chasing. The average cost to originate a mortgage was approximately $11,800 in Q2 2025. Income verification is not a trivial line item.

The Step-by-Step Comparison

Here's the same borrower moving through both systems.

Salaried W-2 employee, direct deposit, major employer:

| Step | Plaid | Traditional | |---|---|---| | Income data pull | Borrower logs into payroll portal: ~90 seconds | Borrower uploads pay stubs: minutes to days | | Verification result | 11 seconds (bank income) or immediate (payroll) | Work Number: instant if employer is listed | | Tax transcript | Not required for bank/payroll path | 4506-C: 10 business days minimum, 30–40% rejection rate | | Verbal VOE | Not required | Required within 10 days of note date | | Cost per file | $1–3 | $55–280+ |

Gig worker or self-employed borrower:

This is where the gap gets uncomfortable. Plaid's bank income ML now specifically handles irregular income streams and launched a dedicated self-employed classification category in April 2026. A borrower who drives for Uber, delivers for DoorDash, or runs a freelance design practice can connect their bank account and Plaid identifies, clusters, and categorizes the income streams automatically.

The traditional process for that same borrower requires two years of tax returns, a year-to-date P&L, possible additional scrutiny from the underwriter, and all of the IRS transcript delays described above. Self-employed borrowers are one of the most underserved segments in mortgage lending. The documentation burden alone disqualifies people who would otherwise qualify on the merits.

The GSE Wrinkle

The obvious question: if Plaid is faster and cheaper, why isn't every lender using it for mortgages?

The honest answer is that GSE lending has specific requirements that bank transaction data alone cannot fully satisfy. Fannie Mae and Freddie Mac have their own approved vendor lists for income and employment validation. Plaid is a certified Fannie Mae Day 1 Certainty provider, which is meaningful progress. Fannie Mae updated its Desktop Underwriter in March 2024 to allow a single 12-month bank asset report to simultaneously validate income, employment, and assets, a structural change that directly benefits open banking providers.

But the full mortgage stack still requires tax transcripts for most self-employed borrowers, the verbal VOE requirement doesn't disappear, and The Work Number remains the default for large employers because it produces a report that maps directly to Fannie Mae Form 1005. Plaid supplements the traditional chain more than it replaces it, at least in the mortgage context.

For consumer installment loans, auto lending, and rental applications, the substitution is more complete. Purpose Financial, which has originated over 134 million loans, reported a 71% lift in customers able to instantly verify income after switching to Plaid, with a 99.8% approval rate for applicants who verified via the platform versus 78% for those going through the manual process.

What This Means for Community Banks

The verification stack is a cost center that community banks have largely inherited rather than chosen. The Work Number's pricing power is a known problem: CHLA formally asked regulators to investigate in 2024, and antitrust litigation is now in the courts. The IRS transcript process has a rejection rate that would be unacceptable in almost any other operational context. Both constraints are structural, not borrower-driven.

Fintechs serving non-GSE loan categories have walked away from that infrastructure entirely. They've built on open banking APIs where the data is faster, cheaper, and more current than anything in The Work Number or an IRS file. Community banks originating consumer loans, personal lines of credit, or auto products have the same option.

The mortgage market is more constrained by GSE requirements, but even there, Plaid's Fannie DU certification means a bank could offer a genuinely faster pre-qualification experience by pulling bank income data upfront, before the full application stack starts, reducing borrower churn during the wait.

The verification delay is not an immutable fact of lending. It's a product of infrastructure built on top of institutions (a credit bureau, the IRS) that did not design their systems around lender speed. The data already exists in the borrower's bank account. How long lenders pay a premium to access it the slow way is increasingly a choice.