Margin of safety: building for the load you did not plan for

The bridge

A civil engineer sizing a bridge for 30-tonne trucks does not build it to carry 30 tonnes. The steel is specified to hold several times that. Not because anyone expects a 90-tonne truck, but because the engineer does not fully trust the load survey, the concrete batch, the weld inspection, or thirty years of weather. The extra capacity is not optimism about traffic. It is a budget for being wrong.

Valuation has the same problem and the same fix. You are forecasting a decade of cash with four or five numbers you cannot verify, and then paying a real price today.

Your valuation is wrong. The question is by how much

Look at what goes into any growth-based model. One input is a fact: trailing earnings, straight off the filing. Everything else is a judgement — how fast the business grows, what multiple a buyer will pay in ten years, what return you require. Three estimates and one fact, and the three estimates compound against each other.

The sensitivity is not subtle. Move an assumed growth rate from 9% to 15% — a range any reasonable analyst could argue over the same company — and the computed value can move by two thirds. You did not get better at forecasting between those two numbers. You just picked a different one.

The arithmetic of a half-price rule

The convention is to take your estimate of value and halve it. If the models say a share is worth 70, the price you are willing to pay is 35. Spelled out, that is an unusual thing to say out loud: you have just declared that you will not buy something you believe is worth twice the asking price unless it gets to half.

The 50% is a convention, not a derivation — nobody proved that half is the correct haircut. What makes it defensible is the size of the error it absorbs. At a 50% discount, you can be wrong about growth by six percentage points, wrong about the exit multiple, and wrong about the timing, and the price you paid is still not above value. At a 10% discount you have bought a forecast, not a business.

What the discount is protecting

It is tempting to read the half-price rule as a return strategy: buy at half of value, value is realised, double your money. That is a possible outcome, and it is not the purpose.

The discount is not there so you make more when you are right. It is there so you survive when you are wrong.

The asymmetry is the whole argument. A position that loses 50% needs a 100% gain to get back to where it started, and the loss arrives in months while the recovery takes years. Permanent loss of capital is the only investing mistake that does not compound in your favour — everything else is a delay. The discount buys room between "my estimate was too optimistic" and "I lost the money".