Founders prepare for diligence as though it were an audit: gather the documents, hand them over, wait to be told whether they pass. It is not that. An auditor asks whether the numbers are right. An investor is asking a different question, and asking it through the numbers.

The question is whether the business is what you have said it is, and whether you are the person who would know.

Revenue, and what counts as revenue

The first thing looked at, always. Not the total, but what is inside it.

Signed or delivered. Recurring or one-off dressed as recurring. Recognised over the contract or taken up front. Whether a large contract signed in the last month of the period is in the number, and why. Whether anything was invoiced early because the quarter needed it.

The number itself is rarely the issue. The issue is whether your definition of revenue is the same one you used last year and the same one in the deck.

The test. Write your revenue definition in one paragraph, then check it against last year's accounts and the current deck.

Where the revenue is concentrated

Who the top ten customers are and what share they represent. How long each has been there, whether any are on notice, whether any are related to a shareholder.

Concentration is not disqualifying and everybody knows it. What matters is whether you volunteered it. A founder who opens with "one customer is 28 per cent of revenue, here is the contract term and here is what we are doing about it" is a different proposition from one where diligence discovers it in week three.

The test. Work out your top ten by revenue share this week, and decide which of them you raise before you are asked.

Gross margin, and what you put in it

Margin is where definitions do the most damage, because nearly every business defines cost of sales slightly differently and nearly none write it down.

Whether support sits above or below the line. Whether the implementation team is cost of sales or overhead. Whether hosting is in there. Move two of those and the margin changes materially without a single transaction changing, which is why the definition gets tested before the number does.

The test. Write down what sits in cost of sales and what does not. If it is not written down, you do not have a definition.

Churn, retention, and the arithmetic underneath

Whether churn is measured on customers or on revenue. Whether an account that halved counts as retained. What happens to the number when the definition is applied to the raw data rather than to the summary.

This is the one most often calculated once, quoted for two years, and never recomputed. It is also the one most likely to be recomputed by somebody else during diligence.

The test. Recompute it from the raw data yourself, before somebody else does it for you.

Working capital, and what the business actually needs

How long customers take to pay against what the contracts say. How long you take to pay suppliers, and whether that has been quietly lengthening. What the seasonal low point looks like rather than the year-end position.

Year-end balance sheets are the tidiest day of the year in most businesses. Diligence looks at the other days.

The test. Look at debtor days and creditor days month by month across two years, not at the year-end position.

Related parties, and everything informal

Loans to or from the founder. A property owned by a shareholder and rented to the business. A relative on payroll. A supplier connected to a director.

None of these are fatal and most are entirely ordinary in an owner-led business. All of them are found. The only version that damages you is the one that emerges after you were asked a question it should have answered.

The test. List every one of them now and put the list in the data room before anybody asks for it.

The gap between the model and the accounts

The model says one thing. The management accounts say another. Both may be defensible and the difference may be nothing more than timing.

It still has to be explained, by you, in one paragraph, without hedging. An unexplained gap between what you forecast and what you booked is read as either weak forecasting or weak record-keeping, and the investor does not have to decide which it is to be concerned.

The test. Put last year's forecast next to what you booked, and write the explanation in one paragraph.

What is really being assessed

Every item above is a proxy. The thing being assessed is whether the person running the business knows what is in it.

Which is why the worst answer in diligence is not a bad number. It is "let me check and come back to you", repeated. And the best is a founder who says the number is wrong, here is why, here is the corrected version, and here is what we changed so it does not happen again. That answer buys more credibility than the clean number would have.

Diligence is survivable with imperfect numbers. It is not survivable with a management team that appears to be learning about its own business in real time. The difference is whether somebody owned the answers before the questions arrived.