Bank statements for a mortgage application: what underwriters look for

The lender has asked for 24 months of bank statements. That is roughly two thousand rows across every account you hold, and an underwriter is going to read them looking for reasons the file does not work. Nothing about that is adversarial. The underwriter has to document a conclusion someone else will audit, and every line they cannot explain becomes a condition on the file and a delay for you.

The useful move is to read your own run first, the way they will read it. Requirements vary by lender, by programme and by country, so treat the conditions your lender actually stated as the rules and everything below as the concern behind them.

What the underwriter is trying to establish

The specifics below are common rather than universal, and where a lender's stated conditions differ, the conditions win. What is stable across programmes is the set of propositions the file has to support.

• That income is consistent and is what you said it is. Deposits that look like the income on the application, at roughly the same amount and cadence, from a source that matches the employer or business named.
• That money arriving in the account can be accounted for. A credit that is not payroll raises the question of where it came from, because funds borrowed to inflate a balance or cover a down payment change the debt position being underwritten. Which credits attract that question, and at what size, is set by the programme.
• That obligations are disclosed. A regular outgoing payment matching no account on the credit report suggests a debt the affordability calculation is missing.
• That the account is managed. Returned items, sustained overdrafts and balances that hit zero before every payday describe a cash position, and they are read as one.

All four are answered from the transaction rows rather than from any summary document, so the run gets read line by line. It is worth being clear about the limit of what statements carry: as with any use of a statement as evidence, they show what moved through the account, not why.

Deposits are not income

The first thing worth doing to your own run is separating deposits into payroll and everything else. The two are treated very differently and the arithmetic shows why.

Take an applicant paid 3,200.00 twice a month. Over 24 months that is 6,400.00 a month, or 153,600.00 of payroll. In month nine a single credit of 14,500.00 arrives from a house sale. Total deposits across the run are 168,100.00, and dividing by 24 gives an average of 7,004.17 a month.

That average of 7,004.17 is not a figure anyone should qualify on, and the reason is visible in the arithmetic: 14,500.00 of the 168,100.00 is a one-off asset sale rather than recurring income. Qualifying income is meant to be income that will still be arriving after completion, so on most programmes a credit like this comes out of the income calculation and moves into the sourcing question instead, where the completion statement from the sale answers it in one document. How your own lender treats a one-off credit is a matter for the programme rules, but you want to know both figures, 6,400.00 and 7,004.17, and which of the two you can defend.

The applicant who has the completion statement to hand loses no time. The applicant who does not spends a fortnight finding it while the file sits on hold.

Returned items and the shape of the month

Returned payments, insufficient-funds fees and reversed entries carry weight out of proportion to their size. They are a direct statement about whether the account can absorb a new monthly obligation.

Recency generally counts for more than the raw total. A returned direct debit twenty months ago and three in the last quarter describe different accounts, and the second is the one that generates questions. Find them in your own run by filtering the description column for the bank's fee and return wording, then check the balance in the days around each one.

An average balance hides the same information. An account that runs from 6,400.00 on payday to 84.00 the day before the next one has no buffer, and averaged across the month it can look comfortable. Read the minimum balance in each of the 24 months and look at the 24 figures as a series.

In plain text: payroll credits are checked for a consistent amount and cadence and support the stated income; large non-payroll credits need a document that traces them, and are commonly left outside the qualifying income figure; returned items are weighed on recency and frequency and are worth a one-line factual explanation each; a recurring debit with no matching credit account is better disclosed by you than found by the underwriter; and the monthly minimum balance shows a buffer the average hides.

The obligation hiding in the debit column

Credit reports do not carry everything. A private loan repaid by standing order, a family arrangement, a support payment ordered by a court, a lease paid direct to a dealership, an instalment plan run through a merchant: all of these can appear as a fixed monthly debit, and whether any given one reaches a credit file depends on the counterparty, the country and which bureau the lender pulls. Some will be there. The safe working assumption is that the statement is where the lender sees the payment, and that anything on the statement but not on the report is something you should name yourself.

The search is for the same amount leaving on roughly the same day each month, to the same counterparty. That is a pattern search you can run once the run is in a spreadsheet, using the same grouping technique as finding recurring charges in a statement: normalise the description, group by counterparty and amount, and count how many months each group appears in. Sort the groups by that count and read from the top. Any cutoff you pick is your own working filter rather than a lender rule, so the aim is a short list of commitments you can explain, not a threshold.

Two adjustments make that search work. Group on amount with a small tolerance, since insurance and utility payments drift. And normalise descriptions before grouping, because the same counterparty routinely appears with a trailing reference number that differs every month.

Reading 24 months without reading 24 documents

1. 1

   Inventory the periods

   List the 24 statement periods for every account the lender asked about, by date, so a missing month is obvious before you convert anything.
2. 2

   Convert each statement

   Convert every statement document, keeping one converted file per source statement so each row stays traceable to the page it came from.
3. 3

   Chain the balances

   Confirm each file's closing balance equals the next file's opening balance and that the periods run back to back. A break means a missing statement or a misread figure, and both need resolving before you go further.
4. 4

   Split payroll from everything else

   Tag every credit as payroll or not, then list the non-payroll credits with the document that sources each one.
5. 5

   Group the recurring debits

   Normalise descriptions, group by counterparty and amount, and sort the groups by how many of the 24 months they appear in. Work down the list until the groups stop looking like commitments.
6. 6

   Read the monthly minimums

   Take the lowest balance in each of the 24 months and look at the run of them together, alongside any returned items.

Chaining the balances is the step people skip and the one that protects the rest. If the run is meant to be continuous, each statement should open where the previous one closed, and the same method that assembles a year of statements into one spreadsheet extends across 24 files without change. A break with a date gap means a statement is missing. A break with contiguous dates points at an extraction error in one of the two files.

Why the numbers have to be right first

Everything above is analysis of a spreadsheet, and analysis is only as good as the extraction underneath it. A misread amount does not just make one row wrong; it moves the running balance from that row to the end of the file, and a monthly minimum computed from a corrupted balance column is a number you should not be making decisions on.

The check that catches it is arithmetic. Recalculate opening balance plus credits minus debits row by row and compare each result against the balance the statement printed. A wrong digit breaks that row's equation and every equation after it, which is what makes it the strongest automatic check available on a converted statement, and why it points at the row that caused the problem instead of telling you the file is out by something somewhere. A missing row behaves the same way. Lose one 3,200.00 payroll credit in month nine and the printed balance beneath it stops following from the row above, so the omission announces itself rather than quietly shrinking that month's deposit total.

That property is the reason to recompute every row rather than only check where the file closes. Compare closing figures alone and a dropped 3,200.00 credit sitting alongside a duplicated one cancels exactly: the closing balance ties, the month's deposits are understated by nothing at all, and two rows are still wrong.

What the row-by-row comparison leaves alone is everything outside the amount column. It has nothing to say about dates, which matters when a credit sits near a month boundary and drags the sourcing question into the wrong month, or when a swapped day and month misaligns the run of monthly minimums. It has nothing to say about descriptions, and the payroll split in the step above rests entirely on that field. A row whose amount is zero contributes nothing to any equation and is invisible to it. The check is also unavailable on statements that print no running balance beside the transactions, where the closing figure is all there is. For an underwriting file the amount column is where the decision is built, so the chain covers the part that matters most; the dates and the tagging still need a sample read against the source pages.

Where a run mixes business and personal activity in one account, do the split between business and personal transactions before you look at the income figures, because separation done by hand at the underwriter's desk tends to be slow and conservative. Hand-keying two years of statements gives you no arithmetic check at all: a transposed figure enters the file silently, and the person who finds it is deciding how much of the rest of your numbers to rely on.