Discipline
Margin of safety and sensitivity testing
Your valuation is wrong. Not carelessly wrong — wrong in the ordinary way that all forecasts are wrong. The margin of safety is the discipline of pricing that certainty in, and sensitivity testing is how you find out how much you need.
Last updated: August 2026
Run the calculator and it returns a single number: $146.29. The precision is an artefact of arithmetic. What you actually know is something much vaguer — "somewhere around $130 to $170, assuming nothing surprising happens". Treating the point estimate as if it were the truth is the most reliable way to lose money with a valuation model.
What the margin of safety actually is
It is the discount to your estimated fair value that you insist on before buying.
margin of safety = (fair value − price) / fair value
Buy at $110 when you think fair value is $146 and your margin of safety is 25%. That cushion has to absorb three separate things, and it is worth being explicit about which:
- Estimation error — your growth rate or exit multiple is simply off.
- Bad luck — the business was fine, the world was not.
- Unknown unknowns — the accounting was fiction, the moat was rented, the founder left.
The cushion is not there to make you money. It is there so that being wrong is survivable.
Sizing it honestly
A fixed 30% rule is better than nothing but it ignores the obvious fact that some forecasts are far more reliable than others. Scale the discount to how confident you actually are:
| Situation | Typical discount | Why |
|---|---|---|
| Stable, boring, predictable earnings; long history; you understand it well | 15–20% | The forecast range is genuinely narrow |
| Good business, some cyclicality or competitive uncertainty | 25–35% | The middle case for most quality companies |
| Turnaround, high growth, regulatory exposure, or a sector you are new to | 40–50% | Your error bars are wide whether you admit it or not |
| You cannot describe how the company makes money in two sentences | No price is low enough | A discount to an unknowable number is still unknowable |
The last row matters most. A margin of safety applied to a valuation you do not believe in is not a safety measure — it is a way of feeling disciplined while doing something reckless.
Sensitivity testing: find what actually matters
Before worrying about the discount, find out which input is driving your answer. The method is crude and effective: change one assumption at a time, leave the others alone, and watch the fair value move.
Take the worked example from the valuation guide — $6.50 EPS, 9% growth, 25× terminal, 15% required return, fair value about $146. Vary the two big inputs:
| Terminal P/E → Growth ↓ | 18× | 25× | 32× |
|---|---|---|---|
| 5% a year | $95 | $128 | $161 |
| 9% a year | $108 | $146 | $184 |
| 13% a year | $123 | $166 | $209 |
Two things jump out. The range across plausible assumptions is roughly $95 to $209 — the "true" answer is a wide band, not a point. And moving horizontally costs far more than moving vertically: the exit multiple dominates the growth rate. That tells you where to spend your research time, and it tells you that a valuation resting on multiple expansion is resting on the weakest possible foundation.
Three questions worth more than the point estimate
- What has to be true? Instead of asking what the stock is worth, set the fair value equal to today's price and solve backwards. If the market price implies 22% annual earnings growth for seven years, you now have a concrete, checkable claim to argue with, rather than a vague feeling that something is expensive.
- What is the downside case? Run the model with pessimistic-but-not-catastrophic inputs — growth halved, terminal multiple at the low end of the historical range. If that still gets you a mediocre positive return, the position is robust. If it produces a 60% loss, size accordingly.
- How wrong was I last time? Keep the assumptions you wrote down. Revisit them a year later. Most investors discover they are systematically optimistic about growth and systematically slow to lower a terminal multiple. That personal bias is worth more than any rule of thumb.
The margin of safety you cannot compute
Position sizing is the other half of the discipline, and it is the half people skip. A 40% discount on a position that is 35% of your portfolio is not conservative. Diversification, liquidity and the ability to hold through a drawdown without selling are all part of your real margin of safety — see drawdown, volatility and concentration for the numbers that measure it.
Benjamin Graham's original formulation is still the cleanest: the margin of safety is "available for absorbing the effect of miscalculations or worse than average luck." Not for generating returns. For surviving being wrong.