Bank statement analysis for a loan turns a stack of statements into a short set of facts: consecutive months, tagged deposits, a qualifying average, flagged events and a written summary. You line up the months, tag every credit, remove what is not income, check continuity and write the findings so an underwriter can verify each figure.
What should bank statement analysis for a loan produce?
A loan analysis should produce five things the underwriter can use without reopening the statements: a month-by-month deposit table, a total and a qualifying deposit average, a list of excluded items with reasons, a list of flagged events, and a short note on trend. Everything else is working paper.
The output, defined
- Monthly deposit table. One row per month: total credits, excluded credits and qualifying credits.
- Averages. The total average and the qualifying average over the lookback, which is the number of months the program reads.
- Exclusions. Each excluded deposit with its date, amount and reason.
- Flags. Large deposits, NSF fees, overdrafts and unusual debits, each with the statement line.
- Trend note. Whether deposits rise, fall or hold across the period.
If a file contains only the average, the underwriter must redo the work to trust it. The goal is a package where every number points back to a line.
Lining up statement months before the loan analysis
Lining up months means putting the statements for every account in consecutive order and confirming that each month has a complete statement, from the first page to the last. You cannot analyse a gap. Doing this first stops you from averaging over a missing month and getting a number that looks fine and is wrong.
Step 1: Build the month grid
Make a grid with one row per month in the lookback and one column per account. Mark each cell as received, partial or missing. A partial cell means a page is absent or the statement period does not run the full month.
Step 2: Check the balances chain
For each account, the closing balance of one month should equal the opening balance of the next. A break means a statement is missing, belongs to a different account or was altered. Flag it before going further.
Step 3: Confirm the account and holder
Check the name and the last four digits of the account number on every statement. Two accounts with similar names are easy to mix, and a deposit counted twice or credited to the wrong person will inflate the result.
Tagging each deposit during loan statement analysis
Tagging means giving every credit one label that says what it is. The label decides whether the deposit counts toward qualifying income. Untagged deposits default to being counted, which is how averages get inflated.
Step 4: Tag every credit
Use a short, fixed set of tags so two analysts reach the same answer. The set used in the example below is payroll, client payment, own transfer and one-time sale. Real files add refunds, loan proceeds and gifts.
A worked example on a fictional six-month file
All figures here are examples for a fictional borrower. The borrower is a self-employed designer with one checking account. The table lists the deposits in one month and how each is tagged.
| Date | Description | Amount | Tag | Counts as qualifying? |
|---|---|---|---|---|
| Mar 2 | PAYROLL PART-TIME TEACHING | 1,200 | Payroll | Yes |
| Mar 9 | ACH CLIENT A INVOICE 118 | 3,400 | Client payment | Yes |
| Mar 14 | XFER FROM SAV 4417 | 2,000 | Own transfer | No |
| Mar 21 | ACH CLIENT B INVOICE 77 | 2,600 | Client payment | Yes |
| Mar 28 | MOBILE DEPOSIT | 5,500 | One-time sale (sold equipment) | No, pending source |
Total credits for March are 14,700. Qualifying credits are 7,200. The difference is the transfer and the one-time sale. If the analyst had counted everything, March would look twice as strong as it is.
Step 5: Run the same tagging on every month
Repeat for all six months. Client payments should be matched by payer name so you can see which clients are recurring. A payer that appears in only one month is not a stream.
Step 6: Sample the statement lines yourself
Before you finalise, pick three deposits at random from the table and trace each back to its statement line. Confirm the date, the amount and the tag. If one of the three is wrong, assume others are and recheck the month. This takes five minutes and catches the transposed digit or the mislabelled transfer that would otherwise reach the underwriter. It also gives you a habit that scales: a fixed check you run on every file, in the same way, whoever did the tagging.
A note on multiple accounts
Borrowers with a personal and a business account need the same analysis on each, followed by a combined view. Tag transfers between the two as own transfers on both sides so they cancel out. If only one side of a transfer is in the file, the other account is missing, and the file needs it before the average can be trusted.
Loan analysis mistakes that inflate the deposit average
The loan analysis mistakes that inflate the deposit average are missed exclusions: own transfers, double-counted deposits, refunds, one-time sales and the wrong divisor. Each one adds a credit that should be out or shrinks the number of months the total is divided by. Each can be checked in a minute.
- Counting transfers between the borrower's accounts. Money moving from savings to checking is not new money. The posts on which transfers come out of average monthly deposits set out the rules.
- Counting the same deposit in two accounts. A client payment deposited in business checking and then moved to personal checking appears twice.
- Counting refunds and reversals. A returned purchase is a credit but not income.
- Leaving in a one-time sale or loan. One large item can lift a six-month average noticeably. Take large deposits out until they are sourced.
- Averaging over the wrong number of months. Dividing by the months with deposits instead of the months in the lookback.
- Ignoring NSF fees. They do not change the average, but they change the file. See how lenders count NSF fees before you finish.
Continuity: check the stream, not just the sum
A six-month average can hide a stream that stopped. Compare the first three months with the last three. If the last three are much lower, the borrower's current income is below the average, and the underwriter will want to know why. Note the comparison in the summary even if the lender uses the full average.
When does manual loan statement analysis stop scaling?
Manual analysis works for a few files a week and stops working when volume, layout variety or the number of accounts per file grows faster than the analyst's time. The sign is not a date; it is the point where the checks get skipped.
The signs
- The grid takes longer than the analysis.
- Different analysts tag the same deposit differently.
- Statements arrive as scans, and numbers are retyped by hand.
- Borrowers with three or more accounts take an afternoon each.
- A reviewer asks where a figure came from and nobody can find the line quickly.
At that point, the part worth automating is the mechanical part: reading lines, grouping repeat payers and adding totals. The judgment about what counts stays with a person. The post on which steps of bank statement analysis can be automated draws that line step by step.
TrueBankFlow does these mechanical steps for lending teams: it reads 3, 6, 12 or 24 months of statements per account, lines them up month by month, sorts transactions into editable categories and recalculates the averages when you include or exclude a deposit. It is onboarding teams in small groups, and you can request a spot.
Writing up the loan statement analysis for the underwriter
The write-up is the part of the analysis the underwriter reads, so it must stand alone. It should state the months covered, the accounts included, the qualifying average, what was excluded and why, and anything the underwriter should ask the borrower. Keep it to one page.
A structure that works
- Scope. Accounts, holder names, months covered and the lookback used.
- Result. Total average and qualifying average, stated as separate numbers.
- Exclusions. A short table: date, amount, reason.
- Flags. Large deposits, NSF fees and overdrafts, with dates.
- Trend. One or two sentences comparing earlier and later months.
- Open items. Documents or explanations still needed.
Wording that helps the reader
Say what you did, not what you feel. "Excluded a 2,000 transfer from savings on March 14; matching debit found in account ending 4417" is checkable. "Looks like a transfer" is not. Where you made a judgment call, say so. The underwriter then decides, and the file shows that the call was theirs.
Write the open items as questions the broker can pass on: "Source of the 5,500 mobile deposit on March 28." That turns a vague concern into a document request. A report that lists clear exclusions, line-cited flags and plain open items is quick to approve, and that speed is the point of doing the analysis well in the first place.