Bank statement analysis: what lenders measure and how software does it

For lenders, underwriters and credit analysts: the numbers a bank statement analysis produces, how to calculate them, and what to look for in bank statement analysis software.

Illustration of a bank statements folder feeding charts and reports on a monitor

Key takeaways

  • Bank statement analysis turns an applicant's statements into the numbers a credit decision uses, among them true revenue or income, average daily balance, overdrafts, existing debt payments and large deposits.
  • True revenue is total deposits minus money that isn't income, such as transfers between the applicant's own accounts, loan proceeds and refunds. It's usually lower than total deposits.
  • The numbers only count if the statement does, meaning balances reconcile, no pages or months are missing, and the account holder matches the applicant.
  • Bank statement analysis software extracts every transaction, categorizes it and flags exceptions, so an underwriter reviews a summary and the flagged items instead of every line.
On this page
  1. What lenders look for in a bank statement
  2. How bank statement analysis works
  3. Worked example: true revenue and average daily balance
  4. Manual review vs bank statement analysis software
  5. How to choose bank statement analysis software
  6. Where underwriter time should go
  7. Frequently asked questions

Bank statement analysis is how lenders turn an applicant's bank statements into the numbers behind a credit decision, among them true revenue or income, average daily balance, overdrafts, existing debt payments and deposits that need a source. Statements show what actually moved through the account, month by month, so lenders use them to check the application and size the loan. Bank statement analysis software, also called a bank statement analyzer, extracts every transaction and runs the math, so underwriters review the exceptions instead of every line.

Account
FieldExtracted valueConfidence
Account holderHarbor Street Bakery LLC
Address118 Harbor St, Portland, ME
Bank nameFirst Midwest Bank
Account number•••• 4821
Account typeBusiness checking
Statement period2026-07-01 → 2026-07-31
Analysis starts with extraction: account details, balances and every transaction row, scans included.

What lenders look for in a bank statement#

Each number answers one underwriting question, how much comes in, how much stays, and how much is already spoken for.

  • True revenue

    Deposits from actual sales or pay, after removing transfers between the applicant's own accounts, loan proceeds and refunds.
  • Average daily balance

    The average of each day's closing balance: the cushion available for a new payment.
  • Overdrafts and returned items

    NSF and overdraft fees and days below zero, the clearest sign that an account runs short.
  • Existing debt

    Recurring loan, lease and merchant cash advance debits that already claim part of each month's revenue.
  • Large deposits

    One-off or round-number credits that need a source. For a Fannie Mae mortgage, a single deposit over 50% of monthly qualifying income counts as large.
  • Ownership

    The account holder, address and statement periods match the applicant and the rest of the file.

Small business lenders weigh revenue and existing debt most heavily; mortgage lenders focus on assets and on sourcing large deposits. See bank statement verification for business loans and for mortgages.

How bank statement analysis works#

  • PDF statements
  • Scans and phone photos
  • Multi-month files
API
  1. 01Extract transactions
  2. 02Categorize
  3. 03Calculate metrics
  4. 04Review exceptions
LOS or decision engine
How bank statements become a credit summary

Take an HVAC contractor's June statement, the same file this guide uses for its worked example below. Extraction reads the header, the balances and every transaction row, joining the pages so a table that runs past the bottom of one doesn't lose a line at the break. Categorization then labels each entry. On this file the labels that matter most set aside the transfer from savings, the draw on the line of credit and the tax refund. All three land in the account, and none of them is income.

From there the calculation stage totals what's left into true revenue, averages the daily balances and adds up the recurring debt payments, while the review stage checks that the balance reconciles and flags anything that doesn't, such as a missing month or a name that doesn't match the application. AI bank statement analysis covers how models handle each stage.

Worked example: true revenue and average daily balance#

An HVAC contractor's June statement shows $96,850 in deposits. The underwriter takes out everything that isn't revenue.

ItemAmount
Total deposits$96,850
Transfer from the company's savings account−$8,000
Draw on a line of credit−$15,000
Federal tax refund−$3,420
True revenue$70,430

The same statement gives the average daily balance. Its 30 end-of-day balances add up to $651,000, and $651,000 ÷ 30 days = $21,700. The account never went below zero, and recurring debits show $4,120 a month in existing loan payments, about 6% of true revenue.

Run the same math on three months of your own statements below.

True revenue from three months of statements

Take out deposits that aren't sales, then see how much of what's left existing loan payments already take.

Month 1
Transfers between the business's own accounts, loan proceeds, refunds
Month 2
Month 3
Debt
True revenue, month 1
$72,200
True revenue, month 2
$82,150
True revenue, month 3
$63,650
Average monthly true revenue
$72,667
Loan payments as a share of true revenue
8.7%
How it's worked out
  • True revenue for a month = total deposits − deposits that aren't revenue.
  • The share is monthly loan and MCA payments ÷ average monthly true revenue: the part of each month's sales already committed before a new advance.

Manual review vs bank statement analysis software#

Spreadsheet review

  • Transactions are keyed or pasted into a spreadsheet, line by line
  • Each deposit is categorized by the analyst's judgment
  • Rows at page breaks get dropped or keyed twice
  • A multi-month file takes hours, and two analysts can reach different numbers

Bank statement analysis software

  • Every transaction is extracted from PDFs and scans, with tables joined across pages
  • The same categories are applied every time, and analysts correct the exceptions
  • Running balances are checked, and uncertain values go to a person first
  • Analysts review a monthly summary and the flagged items

Docsumo is an intelligent document processing (IDP) platform. Its bank statement extraction reads PDFs and scans, joins transaction tables that run across pages, checks running balances and sends fields it's unsure about to a reviewer. Cross-document validation and cash flow analytics are on the Enterprise plan. It doesn't make the credit decision. Results go to your loan origination system or decision engine through API and webhooks.

  • 99%field-level accuracy across 250+ document types
  • 94%+accuracy at Grid Finance
  • <1 minper complex statement at Grid Finance

How to choose bank statement analysis software#

Bank statement analysis tools, often called bank statement analyzers, are only as good as their handling of your worst statements. Run these checks on a sample of your own files. For vendors, see the best bank statement extraction software.

  • Reads your mix of statementsRegional banks, credit unions, scans, phone photos and multi-account statements, not just clean PDFs.
  • Accuracy per fieldAsk for field-level accuracy on your test set. A document-level score hides wrong amounts.
  • Tables across pagesTransactions that run across pages come back as one table, with nothing dropped at the page break.
  • Checks, not just extractionBalances reconcile, no pages or months are missing, and names match the application.
  • Review of uncertain fieldsValues below a confidence threshold go to a person, with the source line shown.
  • Output where you workAn API or webhooks into your loan origination system or decision engine, plus exports.
  • SecurityA SOC 2 Type 2 report, clear data retention, and a plain answer on whether your documents train shared models.
  • Total costThe per-page price plus the review time the tool leaves your team, and who retrains it when banks change layouts.

Where we land. We want the automation predictable enough that a person only sees the transactions that need a decision, each one shown with the reason it was flagged and the statement line behind it. That's a higher bar than an accuracy score. Underwriters have to trust everything they never look at.

Where underwriter time should go#

Bank statement analysis is arithmetic on messy inputs. Total the deposits and remove what isn't income, then average the balances and count the warning signs. The hard part is getting every transaction off every page correctly and knowing which lines to question. Software that extracts every transaction and flags what needs a second look lets underwriters spend their time on the exceptions instead. More on the lending workflow in IDP for lending.

Book a demo with a few of your own bank statements, or start a free trial.

Frequently asked questions#

What is bank statement analysis?

Bank statement analysis is the review of an applicant's bank statements to measure income or revenue, balances, overdrafts, existing debt and unusual deposits. Lenders use it to decide whether a borrower can carry a new payment, and on what terms.

How do you analyze a bank statement for a loan?

First confirm the statements are complete, belong to the applicant and reconcile. Then total the deposits by month, remove the ones that aren't income, calculate the average daily balance, count overdrafts and returned items, list recurring loan payments and source any large deposits.

What does a bank statement analysis report include?

Usually a month-by-month summary covering total deposits and true revenue, average and lowest balances, days below zero, NSF and overdraft counts, recurring debits, and a list of flagged transactions, each linked to the statement line behind it.

How many months of bank statements do lenders analyze?

It depends on the loan program and the lender's credit policy. For a Fannie Mae purchase loan, statements must cover the most recent full two months of account activity (one month for a refinance). Business lenders often ask for more months so they can see seasonality.

What is a bank statement analyzer?

Software that reads bank statements, extracts every transaction, categorizes it and calculates the metrics lenders use, with anything that doesn't add up flagged for a person. For a comparison of tools, see the best bank statement extraction software.

Are there free bank statement analysis tools?

Some statement converters have a free plan or free credits for a few pages a month, and several analysis platforms offer a free trial; Docsumo's covers 14 days and up to 1,000 pages. Test a free tool on your hardest statements, and check whether it flags exceptions or only exports rows to a spreadsheet.

See Docsumo read your own documents

Bring a few real samples. We'll show the fields extracted, the checks that ran and what a reviewer would see.