Two investors can hold the exact same view about a business and walk away with opposite results. The difference is rarely the analysis. It's the price they paid relative to how wrong they could afford to be. A margin of safety is the gap between what you estimate something is worth and the lower price you actually pay for it — a buffer that absorbs the errors in your own estimate. The bigger the gap, the more you can be wrong and still come out fine.
The term comes from Benjamin Graham, but the idea isn't financial. It's engineering.
Start from the bridge, not the stock
An engineer who calculates that a bridge will carry 10 tons does not rate it for 10 tons. They rate it for 3 or 4, because the inputs — material strength, load estimates, corrosion over time — are all approximations, and approximations are sometimes wrong in the same direction at the same time. The gap between what it can take and what you let it carry is the margin of safety. It exists to cover the error you can't see yet.
Investing has the same problem. Every estimate of what a business is worth rests on assumptions — growth, margins, how long an advantage lasts — and each one can be off. The margin of safety is how you price in your own fallibility instead of pretending it away.
How it works: value, price, and the gap
Three terms, defined as we go:
- Intrinsic value — your estimate of what the asset is actually worth, based on the cash it can produce over time. It is an estimate, not a fact.
- Market price — what you can buy it for today. Set by other people, often emotional ones.
- Margin of safety — how far the price sits below your value estimate, usually written as a percentage.
The arithmetic is deliberately simple:
Margin of safety = (Intrinsic value − Price) ÷ Intrinsic valueEstimate a business is worth $100 a share and buy at $70, and your margin of safety is 30%. The price can fall, or your $100 estimate can turn out to be $80, and you still haven't overpaid. The buffer did its job.
Why the size of the gap matters
A margin of safety is not a yes/no switch. It scales with how uncertain your estimate is. The shakier the inputs, the wider the buffer you should demand.
| Margin of safety | What it buys you | Reasonable when |
|---|---|---|
| ~10% | Almost nothing — one bad quarter erases it | Estimate is unusually reliable (rare) |
| ~25–35% | Room for ordinary estimate error | Stable, predictable business |
| ~50%+ | Survives being meaningfully wrong | Cyclical, opaque, or hard to forecast |
Notice what the table implies: a wider margin doesn't mean a better investment. It means a less certain one. The buffer is priced to match your ignorance, not your enthusiasm.
A worked example
Two analysts both value a company at $100 per share and are both correct about the business.
- Analyst A buys at $95. Margin of safety: 5%.
- Analyst B waits and buys at $65. Margin of safety: 35%.
A recession hits and the true value turns out to have been $80 — both of them overestimated by the same 20%. Analyst A paid $95 for something worth $80 and is down ~16%. Analyst B paid $65 for the same $80 and is up ~23%. Same company, same analytical error, opposite outcomes. The only variable was the buffer.
This is the quiet point Graham was making: you do not get paid for being right. You get paid for the gap between price and value at the moment you commit.
Where it shows up beyond stocks
The same logic governs any decision under uncertain estimates. Keeping a cash reserve sized for more than your expected expenses is a margin of safety against a wrong forecast. Planning a project with schedule slack is a margin of safety against optimistic time estimates. In each case the buffer exists for the same reason: your central estimate is your best guess, and best guesses are wrong often enough to plan around.
Try this
Pick one position you hold or are considering. Write down two numbers: your honest estimate of what it's worth, and the price you'd pay. Compute the gap as a percentage. Then ask the harder question — how confident is the value number? If it's a rough guess, a 5% gap isn't a margin of safety; it's a rounding error.
Capture both numbers in JustJot.ai as a dated note so the buffer you demanded is on record, not reconstructed from memory after the fact. Pair it with the [investment thesis](./what-is-an-investment-thesis.md) — the thesis says why you think the value is there; the margin of safety says how wrong that can be and you're still fine. Log the eventual outcome in your [decision journal](./the-investing-decision-journal.md), and over enough positions you can check the one thing that actually predicts results: not whether your estimates were right, but whether your buffers were wide enough when they weren't.