← All publications

From the literature

Reading a business plan with Knight: eight gently awkward questions

A century after Knight, his framework remains an excellent method for reading projects. Not to predict the future, but to know which parts of an application are a matter of calculation and which are still a matter of judgement.

This final publication in the series brings the previous seven together into a single reading framework. Each question stems from a chapter of the book, and each can be put to a real application in an hour's work.

1. Which assumptions are genuinely measurable?

Price volatility backed by a rich history does not have the same status as an assumption about the adoption of a new technology. Putting both in the same "risks" column is convenient; treating them the same way is not. The first sort consists in physically separating the two families in the document: what rests on an observable frequency, and what rests on a conviction.

2. Which risks can be transferred?

Insurance, hedging, contracts, diversification or pooling sometimes make it possible to convert a hazard into a cost. When a risk can be bought or sold, two questions arise: why does the project keep it, and at what price? Retaining an insurable risk is a legitimate decision, provided it is deliberate and the premium saved is visible in the profit and loss account.

3. Which unknowns depend on judgement?

The size of a market that does not yet exist, a team's ability to deliver a world first, a competitor's reaction, the regulatory trajectory: these elements must be made explicit as judgements, not disguised as natural constants. A good application names its bets. A fragile one dilutes them in averages.

4. What does everyone already know?

Expected sector growth is not necessarily a source of advantage. If the market has priced it into assets and skills, it explains the activity, not necessarily a superior profit. The question to ask fits into one sentence: what does this project know, or know how to do, that its competitors do not?

5. Who bears the consequences of an error?

This is the most Knightian way of locating the entrepreneurial function. The decision-maker who sets the assumptions, commits the resources and bears the residual is not in the same position as someone who gives an opinion without exposure. Following the trail of exposure across the investor line-up is often more instructive than reading biographies: it reveals who really believes in the project.

6. What is the plan if the estimate is wrong?

The purpose of a stress test is not only to measure a fall in net present value. It serves to check whether the organisation survives when residual uncertainty materialises in a way the central scenario had not foreseen. A useful test does not ask "what is the impact of a 20% fall in revenue?" but "at what level does this project become unviable, and by what path does it get there?".

7. What information can be produced before deciding further?

Knight insists on the possibility of reducing uncertainty through knowledge. A study, a pilot, a proof of concept or a commercial trial is sometimes worth more than one more decimal place in the financial model. Good practice consists in sequencing commitment: what is the smallest outlay that would remove the biggest unknown?

8. How much does reducing uncertainty cost?

More data, control and guarantees are useful until their cost exceeds the expected benefit. A solid application therefore does not promise the absence of uncertainty. It shows that it knows where uncertainty lies, how part of it can be reduced, and how the rest will be borne.

What a good application demonstrates

This is perhaps the best way to read Knight today. A business plan is not a contract with the future. It is an architecture of decisions under imperfect information.

The eight questions above do not seek to catch an application out. They seek to distinguish two qualities that are often confused: apparent solidity, which comes from precision, and real solidity, which comes from clear-sightedness about what remains unknown. A good model is not the one that makes the unknown disappear: it is the one that avoids mistaking it for a correctly formatted cell.

Contemporary application based on Frank H. Knight, Risk, Uncertainty, and Profit (1921), in particular chapters II and VII to XII. Editorial adaptation, not a quotation.