Bank Statement Analysis
Read every line on a bank statement, with context.
LendPipe extracts every transaction off a PDF bank statement — even scanned and photographed ones — then classifies revenue, flags NSF activity, and surfaces the cash flow patterns your underwriters actually need to make a decision.
$667,857
$525,320
+$142,537
New loan payment began in February, leaving only ~$10K to cover operating expenses.
Significant outflows to related entities totaling over $84K in February alone.
Core revenue is highly diversified with ~103 individual fee payments per month.
What it does
Bank statement analysis without the data entry, without the OCR babysitting.
Bank statements are where most commercial credit decisions are made — and where most analyst hours are lost. Statements arrive as scanned PDFs, phone photos of paper statements, exported CSVs, or printouts taped together over multiple months. Before LendPipe, an analyst keys it all into a spreadsheet or works through a brittle OCR pipeline that produces near-correct numbers your team has to spot-check anyway. LendPipe reads bank statements the way an experienced underwriter does. Every transaction is extracted at the line level, classified into a cash-flow category that matches the borrower's business model, and explained — not just transcribed. Transfers between owner accounts are recognized as transfers, not revenue. MCA proceeds are flagged as financing, not income. Recurring payers are identified, NSF events are counted, and the underlying cash flow trend is computed across months. What lands on the analyst's screen is a structured cash-flow workpaper, not a wall of OCR text.
Capabilities
What the bank statement analysis surface includes
Line-level transaction extraction
Every debit and credit pulled from scanned PDFs, photographed statements, and exported files — including multi-page statements where transactions span page breaks.
Revenue classification
Recurring customer payments separated from owner transfers, related-party movements, refunds, and one-time deposits. Revenue counted the way your underwriting policy counts it.
NSF, overdraft, and chargeback detection
Every NSF event, overdraft fee, and returned item is identified and counted. Patterns that signal stress — clustering, end-of-month timing, repeat NSFs — surfaced in the summary.
MCA and loan position tracking
Merchant cash advance proceeds, daily/weekly debits, and existing term-loan payments detected from descriptor patterns. Existing debt service is quantified before underwriting begins.
Recurring payer and customer concentration
Repeating payer names are deduplicated and aggregated. Customer concentration metrics — top 3 / top 5 / single-customer share — computed across the statement window.
Multi-account and multi-month consolidation
Statements from multiple accounts and across multiple months consolidated into one cash flow view. Transfers between the borrower's own accounts eliminated so revenue isn't double-counted.
How it works
From statement upload to cash flow analysis in four steps
- 01
Upload the statements you have
Drop scanned PDFs, phone photos, exported CSVs, or forwarded broker emails. LendPipe accepts whatever format the bank sends — including the multi-page, low-resolution scans no other tool will read cleanly. Multiple accounts and multiple months can be uploaded together.
- 02
Transactions extract at the line level
Every debit and credit is read off the statement and tagged with the date, amount, descriptor, and balance. Transactions that span page breaks are reconciled. The opening and closing balances of each statement are checked against the sum of activity to catch extraction errors before they affect downstream analysis.
- 03
Classification runs with industry context
Each transaction is classified against the borrower's business type — a $4,200 Zelle deposit looks like revenue at a trucking company and like a transfer at a family office. Owner sweeps, related-party movements, payroll, and financing transactions are isolated from operating revenue. NSF and overdraft events are counted and timestamped.
- 04
Cash flow workpaper produced
Monthly inflows, outflows, and net cash flow are computed. Revenue recognized, transfers excluded, NSF count, debt service identified, and customer concentration land in a structured workpaper. Behavioral findings — structural deficits, related-party patterns, diversified revenue — are summarized in the analyst's review queue.
What you get back
Concrete artifacts your team uses, not a black box
Structured cash-flow workpaper
Monthly inflows, outflows, net cash flow, recognized revenue, and excluded transfers — every figure traceable back to the source transaction. Exports to Excel for committee packets.
NSF and stress-signal log
Every NSF event, overdraft fee, and clustering pattern documented with date, amount, and statement page reference. Used directly in risk write-ups and committee discussion.
Customer concentration breakdown
Top payers ranked with dollar share and transaction count. Single-customer concentration, top-3, and top-5 ratios computed across the analysis window.
Behavioral findings summary
Plain-language flags — structural cash flow deficits, related-party transfers, diversified revenue patterns — with the supporting transactions linked. The analyst reviews, doesn't excavate.
Built for lenders
Bank statement analysis sized for commercial and small-business lenders
Community banks, credit unions, and SBA lenders see a different mix of borrower types than mid-market shops — small businesses, owner-operators, professional practices, restaurants, contractors, and equipment operators. The bank statements those borrowers produce are messy: paper printouts, hand-stapled multi-month packets, three accounts at three institutions, and statements with cents-level activity that the borrower's accountant has never reconciled. LendPipe is built for that reality. The classification logic understands MCA-heavy industries, owner-managed cash flow patterns, seasonal businesses, and the kind of related-party activity that's normal in family-run companies and a red flag in a leveraged buyout. Findings are written in the vocabulary your credit committee already uses — not generic AI summaries that need translation.
Common questions
What lenders ask before they switch
How accurate is the extraction on scanned and photographed bank statements?
Clean digital PDFs from the major banks extract with very high accuracy — typically over 98% of transactions captured correctly. Scanned statements and phone photos extract at lower confidence and are flagged for analyst spot-check on any transaction below threshold. Every extracted line is linked back to its position on the source page, so verification is a single click rather than a re-read of the original statement.
How does LendPipe handle related-party transfers and owner draws?
Transfers between the borrower's own accounts are detected when both sides of the transfer appear in the uploaded statement set, and they are eliminated from the cash-flow analysis so revenue isn't double-counted. Owner draws and related-party payments — typically the largest source of cash flow distortion in small-business statements — are surfaced as their own category in the workpaper, with dollar volume and frequency reported separately from operating cash flow.
Can we configure what counts as revenue for our institution's policy?
Yes. The classification rules — what qualifies as recognized revenue, what counts as excluded transfer, how to treat MCA proceeds, whether to include refunds, and similar policy choices — are configured at the institution level. A community bank lending to professional practices and a fintech-funded MCA shop produce very different cash-flow workpapers from the same statements; both versions are correct, both are what those lenders' policies actually say.
Does this replace our existing bank statement OCR or cash-flow tool?
Most teams use LendPipe to replace a combination of manual data entry, generic OCR tools, and ad-hoc Excel work. The output — structured cash flow with revenue classification, NSF tracking, and behavioral flags — is what teams previously assembled by hand from raw OCR text. If you have an existing bank-statement tool that produces structured output, LendPipe can ingest its CSV exports as a starting point and add the classification and behavioral layer on top.
How does it handle a borrower with multiple accounts at different banks?
All statements are uploaded together regardless of institution. LendPipe matches account numbers and routing details across the upload set, identifies transfers between the borrower's accounts even when they cross institutions, and consolidates the cash-flow view at the borrower level. The output is a single workpaper covering all accounts, with per-account detail available when committee discussion calls for it.
How long does a typical statement analysis take to run?
A standard three-month statement on a single account analyzes in under a minute. A larger borrower with three or four accounts across multiple months — common in SBA underwriting — runs in three to five minutes. The analyst's actual review time is where the time savings show up: instead of two hours assembling a cash-flow spreadsheet, the workpaper is ready when the analyst opens the deal.
Related capabilities
Built to work together, not in isolation
See it run on a real borrower file.
Walk through one of your own deals — document drop to committee-ready output, end to end.
Book a 10-minute demo