How the margin of safety is calculated
The margin of safety is the distance between what a business is worth and what it costs. It is calculated at the end, in a single step, and never folded into the assumptions: a realistic base with the discipline applied once is more useful than a pessimistic base where caution was applied five times with nobody keeping count.
The four steps
First, estimate the value of the business over a visible horizon, which for us is five years. Second, set the minimum return required from the investment. Third, solve for the maximum price payable today to obtain that return, discounting the future value at the required rate. Fourth, the margin of safety is how much cheaper the market trades than that maximum price.
The required-return scale we always show has three rungs: 4%, which barely covers the loss of purchasing power; 10%, roughly the historical average return of equities; and 15%, which corresponds to a clearly good investment. Demanding unrealistic rates — 18% and up — does not make the analysis more prudent: it makes everything look expensive and ends up neutralising the tool.
The intuition: the gap between what you get and what you demand
The margin of safety is the difference between the return the current price offers and the return you required. If at market price the business offers 23% a year and you only demanded 15%, those eight extra points are the cushion. They exist to absorb an estimation error, not to predict the share price.
That is why the verdict is expressed as expected total return rather than as a recommendation: what is measured is what today's price offers, dividends included, discounting each payment from its own year.
The mistake of stacking caution
The most common flaw in valuation is applying a conservative judgement to every assumption: slightly lower growth, a slightly worse margin, a slightly lower multiple, slightly less cash. Each decision looks sensible in isolation and the combined result is unreal, because the cautions multiply against each other. We call this concatenated conservatism, and it produces valuations so pessimistic that they stop informing anything.
The base has to be realistic and unbiased. The discipline lives in the required return, which is applied once and is visible.
The mirror image, which is just as dangerous
The opposite flaw also exists and shows up above all in cyclical businesses: valuing with the commodity price at the top of its normalised range and, on top of that, with an exit multiple above the historical range. Those are two optimisms multiplying each other; they inflate the base and hide a verdict that was in fact negative.
The rule for a cyclical is to use the midpoint of the mid-cycle range and a multiple inside the historical range, not above it. In one of our analyses of an integrated oil company, correcting those two things together moved the verdict from preserving value to overvalued — and neither assumption was outrageous on its own.
Why risk does not go into the discount rate
The discount we use is the risk-free rate with a floor, not a rate adjusted for the volatility of the share. The reason is that volatility is not risk: risk is the permanent loss of capital, and a share that moves a great deal is not necessarily more likely to destroy value.
Risk is charged in two places where it can be seen: in the margin of safety demanded, and in the judgement about business quality. The floor on the rate exists so as not to discount at a depressed rate, which is the arithmetic with which bubbles get justified.
Where this shows up in the analyses
Every analysis shows the full calculation; these are good for seeing the edge cases:
Cyclical: the midpoint of the mid-cycle range and a multiple inside the historical range, rather than stacking both optimisms.
A sum of the parts where each piece has its own metric and its own multiple before being added.
A predictable business: the case where the estimate is most stable and the margin depends almost entirely on price.