Bank Data API Pricing: What Lenders Pay

Bank data for lending is priced in three different units, and the unit matters more than the rate. Connection based providers meter each borrower connection and each API call. Document based providers meter pages. Manual review meters analyst hours. A lender comparing quotes on headline rate alone will usually pick the wrong one, because the number that decides your invoice is your approval rate, not the price per call.

This guide walks through how each pricing unit behaves inside a real lending funnel, where the hidden costs sit, and the four questions to settle with any vendor before you sign.

The three units bank data gets sold in

Almost every option in this market reduces to one of three billing units. They are not interchangeable, and a spreadsheet that compares them as though they were will mislead you.

Open banking providers sit in the first two rows, often in both at once. Plaid bills across five separate models depending on which products you enable: Auth and Identity charge once per connected Item, Transactions charges a monthly subscription for as long as the access token exists, and Balance charges per successful call. A lender running all three is on three meters at the same time.

Why your approval rate drives the bill more than the rate card

Here is the part most vendor comparisons skip. In a lending funnel, most applications do not fund. Small business and MCA desks commonly decline the large majority of what comes in. Every one of those declines still consumed data.

Under a per connection model, a declined applicant costs the same as a funded one at the moment of connection. If the product bills as a monthly subscription, the declined applicant keeps costing you every calendar month until somebody deletes the access token. Plaid documents this directly: subscription products bill as long as a valid access token exists regardless of API activity, and fees are not pro rated for Items removed mid month. Connect on the first, decline on the second, and you have paid for the month.

Now put numbers on it. Take 1,000 applications a month at a 10 percent approval rate. Under a per connection model you pay for 1,000 connections to fund 100 loans, so your true cost per funded loan is ten times the per connection rate. If subscription products stay live on the declines for an average of two months, you are paying for roughly 2,000 borrower months to fund 100 loans. Under a per page model you pay only for the pages you actually read, and a thin decline that gets rejected on the first statement costs a fraction of a full file. The arithmetic here is illustrative and uses no vendor's published rate, because the point is the shape of the curve rather than any particular price.

The shape flips when approval rates are high. A consumer lender approving half its applicants on single account borrowers gets excellent value from a connection, and the verified live feed is genuinely harder to fake than a submitted PDF. That is a real advantage, not a marketing claim.

Multi account borrowers multiply connection costs, not page costs

A commercial borrower rarely has one account. An operating account, a payroll account and a savings account at two banks is three or four separate connections, each one a billable object under a per connection model. The same borrower submitting twelve months of statements is simply more pages, which scales linearly and predictably.

This is why the same vendor can look cheap to a consumer lender and expensive to a commercial one. It is not a difference in the price, it is a difference in how many billable units one borrower generates.

What a bank connection does not cover

Budgeting only for bank data understates the job on a commercial file. An open banking feed returns transactions. It does not return business or personal tax returns, K-1s, year end or interim financial statements, or a debt schedule, and a commercial credit decision leans on all of them.

So a lender who buys a connection still needs a way to read the rest of the file. That is either analyst time or a document analysis layer, and it belongs in the comparison from the start rather than appearing as a surprise six months in. A borrower who hands over a bookkeeping export rather than formatted accounts is a common case, and turning that export into a clean set of financial statements is its own step before any spreading happens.

The costs that do not appear on the rate card

• Connection success rate. If a meaningful share of borrowers will not or cannot link an account, you are paying for a channel that only works on part of your volume, and running a document workflow in parallel for the rest.
• Integration engineering. A connection flow is a build, plus ongoing maintenance as institution coverage changes. Page based analysis with an upload screen and an API is a lighter lift.
• Token housekeeping. Under subscription billing, deleting tokens on decision is a direct saving. If nobody owns that job, it does not happen.
• Minimums and commitments. Annual minimums and twelve month terms mean your first year cost is set by the contract, not by your usage.
• Re-pulls. Every refreshed balance or updated report is another billable event under per call pricing.

Frequently asked questions

How much does bank data cost per loan application?

There is no single rate, because providers meter different things. The useful figure is cost per funded loan, not cost per application: divide your total monthly data spend by loans funded, not by applications received. At a 10 percent approval rate, a per connection model costs roughly ten times its headline rate per funded loan, while a per page model charges only for pages actually read.

Is open banking cheaper than bank statement analysis?

It depends on approval rate and borrower mix rather than on the rate. Open banking tends to win for high volume consumer lending with single account borrowers and high approval rates, where a verified live feed is both cheaper per funded loan and stronger evidence. Document analysis tends to win for commercial and MCA underwriting, where borrowers hold several accounts, most applications are declined, and the file includes tax returns and financials a feed never touches.

Do you pay for a bank connection if the borrower is declined?

Usually yes. One time fees are charged when the product is added to the connection, which happens before any credit decision. Subscription products keep billing monthly until the access token is removed. Since declines are the majority of most funnels, this is often the single largest avoidable line in a bank data budget.

Why do bank data providers not publish pricing?

Most sell several products across multiple billing models to buyers ranging from solo developers to national banks, and they price on projected volume and product mix, so one list price would not describe most deals. The practical consequence is that you cannot shortlist on price without entering a sales process, which is why published alternatives are worth checking early. MoneyThumb publishes a full volume ladder and LenderAnalyzer publishes flat plans, so both can be priced without a call.

What should I ask a bank data vendor before signing?

Four questions settle the real invoice. Which billing model applies to each product I plan to enable? Is there a monthly or annual minimum, and how does it compare to realistic first year volume rather than a projection? What happens to billing when an application is declined, and how do I stop it? And what connection or extraction success rate should I expect on borrowers like mine? Get those in writing and quotes become comparable.

How to run the comparison

Build one model for your own funnel before you collect quotes. Take your monthly application volume, your approval rate, the average number of accounts per borrower and the average page count per file. Then price each shortlisted vendor against that model and divide by loans funded rather than applications received. The ranking changes often enough that the exercise is worth an afternoon.

If your borrowers reliably connect their accounts and you approve a healthy share of them, a connection is likely the better buy and the verification quality is a genuine bonus. If you underwrite businesses from documents, decline most of what arrives, or need the tax returns and financials analyzed alongside the statements, page based bank statement analysis software tends to cost less per funded loan and works on every file rather than on the share that connects. For a fuller view of how the two provider types compare on coverage, see our guide to the best bank statement APIs for lenders, and for the wider category budget see loan underwriting software pricing.