Nanonets Pricing and Cost Per Loan File

Nanonets pricing starts with $50 in free credits, then $100 a month for 100 credits, and every workflow step is billed per run: $0.02 for simple operations, $0.10 for standard AI and $0.30 for complex AI. For a lender the catch is what counts as a run. Nanonets' billing documentation charges extraction per page and formatting per table row, so twelve months of business bank statements can cost around 90 credits to process, far more than the one-invoice examples on the pricing page suggest.

The figures below come from Nanonets' own pricing page and its billing documentation (the "Consumption based Billing" and "Base price per unit" pages), plus Capterra. If you are still choosing a vendor rather than budgeting for this one, our comparison of Nanonets alternatives and competitors for lenders puts it next to Ocrolus, DocuClipper, Docsumo, Textract and Hyperscience.

Nanonets pricing at a glance

How much does Nanonets cost per page?

Extraction alone is 0.30 credits per page. Nanonets' billing docs say Data Extraction AI counts each page processed as one run, so a 5-page document is 5 runs. On the Starter plan, where $100 buys 100 credits, that is about 30 cents a page. Export to your system adds 0.10 credits per page, and any formatting or lookup steps come on top, so the real per-page cost depends on how your workflow is built.

Why bank statements cost more than invoices on Nanonets

Nanonets' own example says a typical workflow runs four to six blocks per document and comes in under $2 per invoice. An invoice is short and has a handful of line items. A business bank statement is several pages with a long transaction table, and that is exactly where the per-row billing bites. The docs are explicit: formatting data in one table column with 10 rows is 10 runs, and in two columns it is 20. A statement with 150 transactions and two formatted columns is 300 formatting runs on its own.

Lenders also cannot skip the formatting. Dates come out of statements in every layout a bank has ever printed, amounts arrive with currency symbols, parentheses and trailing minus signs, and debits and credits sit in separate columns on some statements and a single signed column on others. Normalizing that is the work the formatting blocks are for.

What does one credit file cost on Nanonets?

Here is a worked example using Nanonets' published block prices. Assume twelve months of business bank statements, four pages and 150 transactions a month. That is 48 pages and 1,800 transaction rows.

At Starter pricing that is roughly $91 for one year of statements, before tax returns, financial statements, any Python analysis blocks at 0.18 credits per run, or the $500 a month for confidence scores. Growth plan discounts of up to 40 percent would bring the statement figure closer to $55. Your numbers will differ, because transaction counts vary enormously between a sole proprietor and a restaurant group, so rebuild the table with your own averages. The shape of the result will not change: on statements, rows drive the bill.

For comparison, a flat monthly tool changes the math completely. LenderAnalyzer is bank statement analysis software priced from $99 a month with volume tiers, so the cost per file falls as volume rises instead of climbing with every transaction a borrower makes.

Is Nanonets expensive for lenders?

It depends on volume and on what you expect to get back. For a handful of files a month, the $50 in starting credits and the $100 Starter plan are cheap ways to test extraction. At steady volume, a statement-heavy workflow burns credits quickly, and Capterra reviewers (4.8 out of 5 from 81 reviews) most often name price as the main drawback. The bigger cost is usually not on the invoice at all. Nanonets returns fields and a transaction table. Turning that into an underwriting answer is still work.

What you still pay for after extraction

A clean transaction table does not tell a credit analyst whether to approve the loan. Someone still has to separate revenue from transfers between the borrower's own accounts, count NSF events and negative balance days, compute average daily balance, find recurring loan and merchant cash advance debits, flag stacked positions, and check the deposits against the tax return. On Nanonets you can build some of that in Python or LLM post-processing blocks, billed at 0.18 credits per run, and then maintain it every time a bank changes its statement layout. Or an analyst does it in a spreadsheet.

Either way it costs real money. An analyst spending 45 minutes on a file at a $45 hourly loaded cost is about $34 of labor per file on top of the extraction bill, and it recurs on every application, including the ones you decline. When the statements also feed month-end close on the finance side, matching them to the ledger is a separate job for account reconciliation software, and it should not be confused with underwriting.

Does Nanonets have hidden costs?

Not hidden, but easy to miss. Nanonets lists its add-ons openly in the "Base price per unit" documentation page. The ones that matter to a lender are AI Confidence Scores at 500 credits a month on every plan, and on Enterprise, SAML SSO at 500 credits a month, the audit and file management package at 100, the analytics package at 200, per-user access levels, and managed human review at 1 credit per page. Region-specific AWS hosting and SLAs are custom. A bank or credit union with SSO and audit requirements should price those lines before comparing Nanonets with anything else.

How Nanonets pricing compares with other options for lenders

Two of these publish nothing, so treat any comparison with them as a starting point for a quote. Our breakdown of Ocrolus pricing covers what can be verified for the closest lending-specific vendor, and document AI pricing for lenders lays out the wider market with the billing unit labeled on each row.

Which should a lender choose?

Choose Nanonets if you process many document types beyond lending (invoices, purchase orders, IDs), you want them routed into QuickBooks, Xero or a database, and your bank statement volume is modest. Choose a developer API such as Textract if you have engineers and want the lowest cost per page. Choose a lending tool if the output you need is a credit decision: LenderAnalyzer reads bank statements, tax returns and financial statements and returns cash flow, DSCR, NSF and negative days, recurring revenue and existing debt at a flat monthly price. Teams that want those numbers inside their own systems can use the bank statement analysis API. The quickest way to judge is to upload one real statement at the top of this page and compare the output with what your current workflow produces.