Kaaj raises $3.8M in seed funding to power the future of small business lending 🎉Read more
Bank Statement Analysis

Bank Statement Analysis for Lenders

Utsav Shah·August 18, 2026·8 min read
Kaaj — Bank Statement Analysis Automation for Lenders
Table of contents

About the author

Utsav Shah

AI and decision-systems operator with experience building large-scale systems at Uber and Cruise.

Bank statement analysis is how lenders decide whether a small business can repay, when tax returns are stale, credit files are thin, and the P&L was built for the application. You read the deposits, the NSFs, the transfers that look like revenue, and the MCA payments hiding in the debits. Then you decide.

Doing that by hand across 3, 6, or 12 months of PDFs is slow and inconsistent. Automated bank statement analysis is the same work, structured: classify every line, separate operating revenue from loan proceeds, flag stacking and tampering, and hand the underwriter a sourced summary instead of 80 pages.

This page covers what bank statement analysis actually is in lending, the steps a credit team should run, the signals that change a decision, and how to tell a useful tool from an OCR export.

What is bank statement analysis?

Bank statement analysis is the review of a borrower''s bank transactions to judge cash flow, repayment capacity, and risk. For SMB lenders it is often the primary credit file, not a supplement.

A useful analysis answers five questions:

  • Is the deposit stream real operating revenue, or transfers, refunds, and loan proceeds?
  • Can the business cover existing debt, including MCA and other stacking, after a new payment?
  • Are there NSFs, overdrafts, or end-of-month balance collapses?
  • Do the statements match the application, invoices, and tax returns?
  • Does the PDF show document-integrity signals that require source verification?

If your process only totals credits and debits, you are not doing bank statement analysis. You are adding.

Why it matters for SMB credit

FICO and last year''s tax return miss the last 90 days. Bank statements do not.

Lenders use them to:

  • Verify stated revenue against actual deposits
  • See cash-flow seasonality and average daily balance
  • Find undisclosed debt (especially MCA payments)
  • Catch non-business spend and owner drains
  • Build alternative risk features: debit-credit ratio, bounce rate, end-of-month dips

That is why equipment finance, MCA, brokers, and community banks all end up in the same place: the statement is the underwrite.

How to analyze a bank statement

1. Collect the right window

Three months is a snapshot. Six months shows a season. Twelve months is what you want when the business is cyclical. Take every account the borrower actually uses, not just the one on the application.

2. Separate revenue from money movement

Credits are not revenue. Split:

  • Operating deposits (customers, card settlements, ACH from buyers)
  • Transfers between own accounts
  • Loan proceeds, MCA advances, merchant cash
  • Refunds, chargebacks, one-off asset sales

A restaurant Zelle from a regular is often revenue. A trucking company Zelle from an owner is often not. Industry context is the difference between a good classification and a bad one.

3. Map the true debt burden

Debits that look like vendor or ACH are often existing positions. Tag MCA, daily/weekly ACH to funders, equipment notes, and owner draws separately. This is where stacking lives.

4. Read balances, not just flows

NSFs, overdraft days, negative-balance days, and average daily balance tell you whether the revenue story survives a slow week. A high-deposit month with a $400 ADB is a different credit than the same deposits with a $40,000 ADB.

5. Cross-check the rest of the package

Match legal name, EIN, and addresses to SOS, the application, and the voided check. Tie large deposits to invoices. Tie tax-return revenue to the deposit trail. Mismatches may reflect stale records, naming variations, packaging errors, or fraud risk. Treat them as exceptions that require source-level review, not as conclusions.

6. Write the finding, not a transcript

The output a credit committee can use is a short, sourced summary: verified monthly revenue, ADB, NSF count, stacking, exceptions, and a note on what you still do not believe.

What a good analysis should output

A lender-ready bank statement analysis is a structured object, not a spreadsheet dump.

FieldWhy it matters
Verified monthly operating revenueAfter stripping transfers and loan proceeds
Average daily balanceLiquidity, not just inflow
NSF / overdraft / negative daysStress and bank relationship risk
Existing MCA / daily ACHStacking and true DSCR
Owner draws and related-partySustainability of stated cash flow
Large unexplained depositsPossible non-operating cash that requires source review
Document-integrity flagsWhether the file needs source verification or a replacement copy
Source linksEvery number back to a page

If a tool cannot show you how it classified a deposit, your analyst will redo the work.

Manual vs. automated bank statement analysis

Manual review still wins on ugly edge cases. It loses on volume, consistency, and exam trail.

ManualAutomated
Time30-60 minutes per file, longer on scansMinutes per package
ConsistencyVaries by analystSame rules on every deal
FraudEasy to miss on page 80Layout, metadata, and pattern checks on every page
AuditNotes in a LOS commentLineage from field to source page
ScaleHeadcountSame team, more applications

The working model is hybrid: automation classifies and flags, humans decide the exceptions. For the failure modes and controls to test, see AI bank statement analysis for lenders.

Bank statement analysis tools: what to buy

Most bank statement analysis software is one of three things. Buy the layer you actually need.

Extraction / OCR (examples include Ocrolus and MoneyThumb): tools in this category primarily convert bank-statement PDFs into structured transaction data. Product scope varies; this category can fit when underwriting logic already lives somewhere else.

Open banking (Finicity and similar): borrower connects the account. Clean when it works. Incomplete when the borrower will not connect, or the account they connect is not the one they use.

Underwriting intelligence (Kaaj): statements inside the full borrower package. Revenue versus transfers and MCA proceeds, NSF and ADB, stacking, document-integrity signals, and a source-linked memo prepared for human review and connection to the lender's existing workflow.

SMB lenders usually need the third. Extraction alone does not tell you if the $48,000 deposit is a customer or yesterday''s MCA.

For a side-by-side, see best AI bank statement analysis tools for lenders.

Common mistakes

  • Treating all credits as revenue
  • Ignoring small daily ACH that is actually an MCA
  • Reading one month and calling it a trend
  • Skipping the PDF forensics (cropped pages, font mix, balance math that does not foot)
  • Never matching statements to invoices and tax returns
  • Leaving the finding in a spreadsheet instead of a memo the committee can review

How Kaaj does this

Kaaj is the underwriting OS for SMB lending. Bank statement analysis is one agent in a package workflow, not a standalone OCR product.

  • Reads scanned, photographed, and digital statements
  • Classifies revenue, transfers, MCA proceeds, owner draws, and loan payments with industry context
  • Tracks NSFs, overdrafts, negative days, ADB, and stacking
  • Cross-checks SOS, invoices, and the rest of the file
  • Writes a sourced credit-memo section your underwriter can edit
  • Syncs to the LOS or CRM you already use

Kaaj is designed to work alongside an existing LOS or CRM. Implementation scope and timing depend on the lender's workflow, data sources, and integrations.

Book a demo and run one of your own packages through it.

FAQs

What is bank statement analysis?

It is the review of bank transactions to judge a borrower''s cash flow, debt burden, and risk. In small business lending it is often the main credit file.

How do you do bank statement analysis?

Collect 3-12 months of statements, classify credits and debits, strip transfers and loan proceeds from revenue, flag NSFs and stacking, then cross-check the file. Write a sourced summary. Do not stop at a credit-minus-debit total.

How does bank statement analysis improve lending decisions?

It shows the last 90 days of actual cash, not last year''s tax return. That is how you catch stacking, verify stated revenue, and price the deal to the real payment capacity.

Which software is used to analyze PDF bank statements?

OCR tools extract the lines. Open-banking tools pull live accounts. Underwriting platforms like Kaaj classify the lines in context and attach them to a credit memo. Most SMB shops need the last category, not a spreadsheet export.

Ready to see Kaaj in action?

Book a demo and walk through a live deal with our team — from intake to credit memo.

Book a demo

Related articles

Bank Statement AnalysisHow underwriting systems will evolve as lending scalesMay 8, 2026 · 5 min readBank Statement AnalysisHow AI Can Expand Access to CapitalMay 8, 2026 · 6 min readBank Statement AnalysisWhat Bank Statement Analysis Is (and When Lenders Should Automate It)August 18, 2026 · 8 min read