How to Write a Credit Paper

A credit paper has three readers, and only the first is reading it this week. The approver — a credit officer with delegated authority, or a committee — reads it to decide. The banker who inherits the relationship reads it next year to learn why the facility looks the way it does. And if the loan goes wrong, a reviewer reads it to find out what the bank knew, what it assumed, and whether the decision was reasonable on the evidence at the time.

None of the three was in the meetings. So the test of a paper is simple to state and hard to pass: a reader who knows only what is on the page should reach the same decision as the writer, and be able to see which facts it rests on. Everything below is in service of that — the order of the sections, how the numbers are written, and what is left out.

Banks name the sections differently and their templates add fields of their own, but underneath every template are the same nine questions. A small renewal may answer three of them in one paragraph; a large new borrower may give each its own pages. What an approver will not accept is a question that is not answered anywhere.

The first paragraph carries the whole paper: the borrower, the facility, the amount, the recommendation, and the one or two reasons for it. An approver reads the rest to test that paragraph, and reads it more carefully for knowing where it is going. A paper that builds to its recommendation asks the approver to form a view along the way — and an approver who has formed a different view by page three reads the recommendation as the writer's opinion rather than as the conclusion of the evidence.

The reasons in that paragraph should be the ones the decision actually turns on, not a summary of every section. If the facility is approved because the request matches what the business's cycle needs, say that. If it is declined because profit is not turning into cash, say that. The approver can then go straight to the section that proves it.

The financial section is where most papers lose their reader. A table of three years of accounts tells the approver nothing they could not read in the appendix; what they need from the body is what the numbers mean. Write each finding as a sentence with its number, its period and its direction — revenue up 36% over two years, debtor days moved from 42 to 44 — and then say why, because a number without a cause is a number the approver has to interpret alone.

State each number once, and state it the same way everywhere. A margin quoted as 12.4% in the summary and 12% in the analysis, or a facility described as US$4m in one place and US$4.0m plus an overdraft in another, makes a careful reader check everything else. Put the tables in the appendices, make sure the body agrees with them, and name the number the decision turns on so the approver knows which one to check hardest. The cash flow analysis guide sets out what the repayment section reads from the statements.

A risk section that lists fifteen risks has not decided which ones matter. Name the three or four that could actually cost the bank money on this file. For each, say what reduces it — a contract, a covenant, a guarantee, the security — and then say what is left after the mitigant has done its work. The residual is the part approvers read most closely, because it is the risk the bank is choosing to take, and a paper with no residual risk anywhere has not finished thinking.

The same discipline applies to what the writer does not know. A forecast not yet tested, a customer contract not yet sighted, a valuation more than a year old: say so, and say whether the recommendation depends on it. An approver who finds a gap the paper did not mention stops trusting the parts that were mentioned.

Every condition and covenant should answer a risk the paper has named. A borrowing base answers a risk that receivables will not convert; a clean-down answers a risk that seasonal money becomes permanent; a covenant answers a risk that the bank will learn of trouble too late. A condition that answers nothing in the paper is a template, and conditions are not free: they cost the borrower time every month and they tell a good client that the file was read with a checklist.

So the structure should be as light as the file allows. A clean file gets a short list and a quick answer; a file with a named weakness gets the one or two conditions that address it. The drill Issue 04 turns on exactly this, and the sample paper below is written for its answer.

There is no right length, only a right proportion. The paper for a clean increase to an established borrower answers all nine questions in a few pages because each answer is short; the paper for a new borrower in an unfamiliar industry, or for an existing one whose numbers have turned, is longer because each answer needs more evidence. Length that does not add evidence dilutes the evidence that is there.

What follows is a whole paper for a real decision on a composite borrower: Highview Industries, the file in Issue 04, asking to increase its working capital revolver from US$2.5m to US$4.0m. Every figure comes from that drill and its file appendix. If you would rather make the call yourself before reading someone else's, do the drill first — the paper gives the answer away in its first paragraph, as a paper should.

A paper that declines, or that approves something different from what was asked, has the same nine sections. What changes is where the weight sits. The first paragraph says what the bank will and will not do, and why; the financial section carries the finding that decided it and shows it plainly; and the conditions section, if there is one, sets out the structure the bank would support instead. Issue 01 is the same request as Highview's on the surface, and its paper would open with the opposite sentence: the working capital the business asked for would fund a cycle that has lengthened, not a business that has grown.