You’ll sell your company once. What the buyer pays and what you keep are two different numbers.

Investment committees are less dramatic than founders imagine. Nobody bangs the table. Someone asks a question, the answer takes slightly too long, and the folder moves quietly to the bottom of the pile. The real reason is never written down, which is why founders so rarely learn it.

The signals I pay attention to are small. Gross margin by segment. Why churn moved in the second half. What the largest customer actually pays. The turn to the CFO changed the room every time — not because a founder should carry every figure in his head, but because of what it revealed: the economics of the business lived with someone else.

And if they live with someone else, the committee begins to suspect that the story and the accounts were assembled separately — by different people, for different purposes, reconciled late. From that moment the question is no longer whether the plan is ambitious. It is which parts of it the seller can personally defend.

What follows is not necessarily a refusal. It is a longer diligence, a wider discount for risk, more of the price deferred into earn-out, tighter warranties. If the company is bought, it’s bought on the buyer’s terms rather than its own — and everyone involved calls it a fair process, because it was.

This is preparable, and it has nothing to do with memorising a data room. It is the management team agreeing what each number means and who owns it, far enough ahead that the agreement has already been tested by ordinary board questions before an outsider asks a hard one.

A buyer who has already decided to proceed still has to defend the price internally. Every figure he cannot verify himself becomes a figure he discounts — not out of suspicion, out of process.

Management accounts that don’t reconcile with the statutory ones don’t read as dishonesty; they read as work someone else will have to do. A shareholder from 2015 nobody has spoken to since doesn’t read as a problem; it reads as a fortnight of a lawyer’s time before the business itself can be looked at. A revenue line that lives inside one person’s relationships doesn’t read as risk; it reads as a person you are buying instead of a company.

None of it stops the transaction. All of it moves the price.

The founder experiences this as bad luck, or as a buyer who negotiates hard. It is neither. It is the accumulated cost of every number the other side had to take on trust. And the arithmetic is unforgiving: the fixing takes about eighteen months, and by the time a buyer is in the room, you have four weeks.

That is why I take the CFO seat before a process starts rather than during it. Not to make a company look better — to make it verifiable. A buyer who can check what he is being told has no reason to discount it.

The first was conservative: revenue recognised late, every cost accounted for, covenants comfortably met. It existed to keep a credit line open. The second was the growth case — addressable market, expansion multiples, what happens if two things go right. It existed to raise money. Different audiences, different questions, and inside the company nobody thought of this as unsettling.

The trouble was that the two had drifted. Not by a rounding error — by a definition. Recurring revenue meant one thing in the first and something looser in the second. Two of the largest contracts were treated differently in each. Laid side by side, they described two companies that happened to share a name.

What neither version was, was the model the company actually ran on. That one lived in a spreadsheet on the finance director’s machine, was shown to nobody, and agreed with neither of the other two. It held the real margin by segment, the real cost of the largest account, the real cash position at month five. Those are the figures a founder recites from memory in a negotiation — and the reason they cannot be found in any file handed over is that the file they came from was never meant to leave the building.

A buyer eventually sees both. Lenders talk to acquirers, old decks circulate, diligence asks for the file. When the two versions finally meet on someone else’s desk, the question stops being which number is right. It becomes: what else has this company not reconciled? That question is never answered in a meeting. It is answered in the price.

What made it hard to fix was not math. Each version had a constituency. The finance director had defended the conservative case to the bank every quarter for three years. The founder had told the growth story to every investor he had ever met. Neither wanted to be the one who had been wrong, and until someone was willing to say that the two would now become one, nothing moved.

The fix was a single set of definitions, one owner of the numbers, and a reporting pack that could be handed to either audience without editing. That work takes months — which is precisely why it shouldn’t begin when a buyer is already reading.

One company should tell one version of its numbers. Two sets of figures is not a reporting inconvenience; it is the first thing a buyer finds and the last thing a seller can explain. So before anything else I check whether the management accounts, the statutory filings and the investor deck could be laid side by side and read by a stranger without a translator. If they need one, that translator is currently the founder — and he will not be in the room during diligence.

Founder-dependency is priced long before it is admitted. If revenue moves when the founder takes two weeks off, that is a valuation input, not a character trait. The question is never whether the dependency exists in a founder-led company — it does. It is whether there is a credible answer for the eighteen months after completion, and whether anyone else in the business has been allowed to build one.

Silence in the management team is not agreement. Far more often it means nobody wants to ask the question first. A team that has not disagreed about anything in the year before a sale usually has one person quietly carrying the risk and a set of assumptions no one has stress-tested. Buyers work out who that person is, and they price the concentration.

A cap table can be legally clean and economically unresolved. Options promised in conversation and never documented. Loans that behave like equity. A co-founder who left on a handshake. A nominee arrangement everyone remembers slightly differently. The paperwork can be immaculate while the economics of ownership are unsettled — and a buyer reads the economics, not the certificate. This is the cheapest category to fix early and the most expensive to discover late, because it is the one that can suspend a signing.

The price of the company is not the price of your share. Preference stacks, liquidation multiples and the instruments above all sit between the headline number and what reaches you, in a fixed order of payment. Most founders learn the shape of that order during the negotiation, when the terms are already written and the arithmetic runs against them. It is knowable years earlier, and it changes what a good exit even looks like: a lower headline with a clean structure often pays a founder more than a higher one with three layers of preference above him. I model that distance before a buyer is in the room, so you know what an offer is actually worth to you before it arrives.

The story a company tells was written for earlier questions. Most narratives were built for a funding round, a hiring push or a set of customers, and they were true at the time. By the time a sale is coming, the description and the reporting have quietly drifted apart. A buyer does not read that gap as marketing. He reads it as evidence that other things in the company were assembled the same way — separately, for different audiences, reconciled late. Bringing the account and the numbers back into one version is work, and it belongs to the same eighteen months.