The DCF Trap: Why Estimates Fail

We’re going to talk about one of the most powerful tools in finance, and how it can also be the most dangerous trap for the unwary investor.

When you hear about a company being "valued" using a model, you might get the impression that there is a magical formula—a precise equation that reveals the exact true value of a company. You might think that if the numbers line up, you have found a sure thing.

It isn't.

This lesson looks at why a tool that is elegant in theory disappoints so often in practice. The short version: a DCF is not a measurement, it is an argument expressed in numbers. Its output is exactly as reliable as the assumptions you fed it, and the decimal places disguise that completely.

The Illusion of Precision

The DCF model looks incredibly professional. Dates, formulas, a grid of figures. It produces a specific answer: "this company is worth 313p per share."

But that precision is an illusion. It is a lie dressed up in spreadsheet formatting.

A DCF model does not tell you the value of a business. It tells you the value of your guesses. The model is a mirror: it reflects your assumptions back at you. If you enter the wrong growth rate or the wrong discount rate, the output will be spectacularly wrong, but the math will still look perfect.

Think of it like a GPS. The GPS gives you turn-by-turn directions. But if you put the wrong starting point in the GPS, the destination will be wrong. The directions are perfect, but the result is garbage.

The "Black Box" of Inputs

To understand why the trap is so dangerous, we need to look at the "Black Box" of inputs—these are the assumptions that go into the calculation. You have to feed the machine numbers before it can work.

Here are the three biggest guesses you have to make:

1. Future cash flows. Projections for five or ten years out. Errors compound: a growth rate that is slightly too high in year one is applied again in year two, and again, and then feeds the terminal value.
2. The discount rate. As set out in lesson 3: the return you require, built from the gilt yield plus an equity risk premium plus company-specific risk. Typically 8–12%, and moving within that range alone can shift a valuation by 40%.
3. The terminal growth rate. The most consequential of the three, because terminal value usually dominates the total. A shift from 2% to 3% — both entirely defensible — can move the answer 20% or more.

The Trap: Investors often tweak these inputs until the model gives them the answer they want to hear. "If I nudge the growth rate from 5% to 6%, the valuation rises 20%. That seems fine — 6% isn't unreasonable."

The model offers no resistance while you do this. It will accept any input you give it and return a confident-looking number, and nothing in the output distinguishes a carefully researched assumption from a convenient one.

Sensitivity Analysis: The Pendulum

Because a DCF is so sensitive to its inputs, the professional practice is never to quote one number. Instead you recalculate across a range and present a grid.

Value per share, varying growth and discount rate:

Read what that grid is telling you. The same company, with four plausible growth assumptions and four plausible discount rates, is worth anywhere between 226p and 444p — a spread of almost 2:1, entirely from inputs that are all individually defensible.

This is the honest output of a DCF. Not "this share is worth 313p", but "on assumptions I can defend, this business is worth somewhere between roughly 230p and 440p."

That range is genuinely useful. At 180p the share is below every cell in the grid, which is interesting. At 600p it is above every cell, which is also interesting. At 320p it sits in the middle, and the correct conclusion is that this model cannot tell you anything.

The rule: if the market price is close to your DCF figure, the model has told you nothing — that gap sits well inside your own margin of error.

What you are looking for is a margin of safety, and its direction matters enormously:

Your estimate of value must be comfortably ABOVE the price you pay.

Value 400p, price 250p → a margin of safety of roughly 38%. ✓ Value 400p, price 390p → effectively no margin. ✗ Value 400p, price 500p → paying above your own estimate. ✗

The point is that you buy at a discount to what you think it is worth, and that discount is what absorbs your being wrong. Get this the wrong way round and you would be buying only when the price exceeds your valuation, which is precisely backwards.

The Danger of Mistaking Estimates for Predictions

The most critical trap in DCF analysis is thinking you know the future.

When you build a model, you are making a forecast. You are saying, "I believe this company will do X, Y, and Z."

The model does not predict the share price. It calculates the consequence of your assumptions. If the assumptions are wrong — and some will be — the output is wrong in exactly the same proportion, while looking every bit as authoritative.

The Mental Shift:

• Wrong: "My model says 400p and it trades at 390p, so it's cheap." — a 2.5% gap against inputs that could move the answer 40%.
• Better: "My model gives a range of roughly 230p to 440p on defensible assumptions. It trades at 180p, below my entire range. That is worth investigating properly."

The Margin of Safety: The Antidote

Since you cannot be precise, you must be conservative. This is the philosophy of Benjamin Graham and Warren Buffett, and it is the best way to use DCF.

Since you cannot be precise, be conservative. This is the discipline associated with Benjamin Graham and Warren Buffett, and it is the only sensible way to use a DCF.

Never buy purely because a model says "undervalued" — the model says that because of numbers you typed into it. Instead:

1. Build a base case on assumptions you'd defend.
2. Build a downside case where growth disappoints and margins compress.
3. Ask whether the price still looks reasonable against the downside case.

If it does, you have a genuine margin of safety: the business can do materially worse than you expect and you may still be all right.

One thing a margin of safety cannot do: protect you against a business that is genuinely worthless. If a company is heading for administration, no discount to a mistaken valuation helps — the true value is nil, and any price above nil is too much. The margin protects against estimation error, not against being fundamentally wrong about the business. That is what the balance sheet work is for.

How wide should it be?

It depends on how confident you are in the forecast:

The last row is the important one. If a business cannot be forecast, the answer is not a bigger discount. It is a different valuation method, or a different company.

An analogy. You are deciding whether to cross a frozen lake. Your measurement says the ice is 15cm thick. You know 25cm is the safe threshold, and you know your measurement could easily be out by 5cm either way.

You don't cross. Not because the measurement said unsafe — it didn't — but because the margin between what you measured and what you need is smaller than your own measurement error.

That is a margin of safety. It isn't about the estimate being accurate. It is about leaving enough room that being wrong doesn't matter.

The Most Useful Thing You Can Do With a DCF

Given everything above, here is the application that survives all of the criticism — because it does not require your forecast to be right.

Run the model backwards. Instead of estimating growth to produce a value, take the current share price as given and solve for the growth rate that would justify it.

The output is not a valuation. It is a statement of what the market is currently assuming, and it converts an unanswerable question into an answerable one:

• "Is this share expensive?" — a matter of taste, and unresolvable.
• "Does this company grow free cash flow at 11% a year for the next decade?" — a question you can research, test against the company's history and its market, and form a genuine view on.

Do it with the DCF calculator: adjust the growth input until the output matches today's share price. Whatever rate you land on is the market's implied assumption.

Sometimes the answer settles the matter immediately. If a mature company in a low-growth market needs 15% annual growth for a decade to justify its price, you do not need a valuation model to know that is demanding. Equally, if a business only needs 2% to justify its price and has grown at 8% for a decade, that is worth a closer look.

This is covered further in How much growth is priced in?.

When to Use DCF

So when is this tool worth the effort?

• Use It For: Understanding your own "margin of safety." To prove to yourself that a company is cheap. To understand how sensitive a company is to growth assumptions.
• Don't Use It For: Timing the market. Don't use it to try and find the exact bottom. Don't use it to beat a computer algorithm.

Summary

• It's a Guess, Not a Fact: DCF relies on assumptions, not precise data. Input a bad assumption, and you get a bad result.
• It Is Extremely Sensitive: A small change in growth rate can change the value of a company by billions.
• Margin of Safety: Because the model is noisy, you need a wide gap between the price and the calculated value to protect yourself.
• Treat it as a range. Build a sensitivity grid rather than quoting a single figure, and act only on gaps wide enough to survive your own error.
• Run it backwards. The most reliable use of a DCF is solving for what the market already assumes, which requires no forecast of your own.

The bottom line: a DCF is a tool for thinking, not a device for producing answers. Its real value is that it forces you to write down what you believe and then shows you the consequences. If the output ever lands exactly where you hoped, check what you changed to get there.