Why a DCF's precision is false comfort

A DCF hands you a figure like ₹166.42 per share. It looks like a measurement — as solid as a weight on a scale or a reading off a thermometer. That solid feeling is the most dangerous thing a DCF produces, and this module is about resisting it.

The number is exact because arithmetic is exact: feed a spreadsheet precise inputs and it returns a precise output, to as many decimals as you like. But the inputs were guesses — growth rates, discount rates, a terminal value about the far future. A precise sum of rough guesses is still rough. The exactness is real; the accuracy is imagined. This is false precision: the comfort of a sharp number resting on soft foundations.

Computing has an old saying: garbage in, garbage out. A DCF adds a cruel twist — garbage in, gospel out. You feed it uncertain assumptions and it returns them laundered into a clean, authoritative, decimal-pointed figure that feels far more certain than anything you put in. The spreadsheet strips the doubt off your inputs and hands you back false confidence.

Three features of a DCF make this worse than in most estimates:

• The inputs are unknowable, not merely unknown. You cannot look up next decade's growth or the right discount rate the way you look up a share price. They are judgements, and reasonable people disagree by wide margins.
• The output hides the disagreement. A single number gives no hint of how much it would move if you had chosen slightly differently. The uncertainty is invisible in the very figure that matters.
• The maths has enormous leverage. As you saw with the terminal value, dividing by small numbers means small input changes produce large output swings. A DCF amplifies the uncertainty in its inputs rather than dampening it.

This is why the wise old rule, usually credited to Keynes, matters so much here: it is better to be roughly right than precisely wrong. A DCF tempts you into precise wrongness. The defence is to keep the roughness of the inputs visible in how you report the answer.

Return to our composite. illustrative Everything is held exactly as in module 006 — ₹100 crore of cash, 10% growth for five years, 4% terminal growth — and we move only the discount rate, the yearly rate we shrink future cash by, across a band any careful analyst might choose: 11%, 12%, 13%.

Sit with that. A discount rate of 11% versus 13% is not a wild disagreement — it is the ordinary spread between two sensible people arguing about how risky the same business is. Yet it moves the value from ₹1,907 crore to ₹1,472 crore: a swing of about ₹435 crore, roughly a quarter of the whole valuation, from a single input nobody can measure. And this is with everything else held still. Let growth and terminal growth wobble within their own plausible bands at the same time, and the true range of "defensible" answers is wider still.

Per share, at 10 crore shares, that band is roughly ₹147 to ₹191 before the other inputs even move. The tidy ₹166.42 was never a point. It was the middle of a cloud.

Picture two analysts handed the same company. illustrative The first builds a lean, three-input model and reports: "worth somewhere around ₹140 to ₹200 a share, most likely near ₹165 — and the answer is very sensitive to the discount rate, so I'd want a real margin below that before buying." The second builds a fifteen-year, three-stage model with forty line items and reports: "intrinsic value ₹166.42".

Which is more trustworthy? The instinct says the second — it is more detailed, more effortful, more precise. But the first is the honest one. It says what it knows and, crucially, what it does not. The second has spent its extra effort hiding uncertainty rather than reducing it: every additional line item was another guess, and forty guesses do not average out into accuracy — their errors compound. The precise figure is not a sharper measurement; it is a rougher estimate wearing a lab coat.

The tell is the decimal point. ₹166.42 claims to know the value to within a few paise. Nobody knows any company's intrinsic value that finely — not Buffett, not the analyst, not the algorithm. A number reported that precisely is announcing, without meaning to, that its author has confused the exactness of arithmetic with knowledge of the world.

This is the heart of the module, so it is worth stating plainly.

A precise output cannot make uncertain inputs certain. The decimals are borrowed from the arithmetic, not from any real knowledge of the future. No amount of modelling detail turns a guess about year-twelve growth into a fact.

It cannot tell you which of its own numbers to trust. A single figure gives no sense of its own fragility. That is why the point value must always be replaced by a range — the subject of the very next module.

It cannot protect you from a confident story. The greatest danger is not that a DCF is imprecise, but that its precision feels like proof, and quietly ends the argument. A ₹166.42 shuts down the very doubt that should keep you cautious. The number's job is to open a discussion about assumptions, not to close it.

1. Trusting the decimals. ₹166.42 feels known to the paisa. Round hard — to ₹165, or better, to "the mid-160s" — so the figure cannot pretend to a precision it does not have.

2. Mistaking complexity for accuracy. More stages and line items add guesses, not certainty. A lean model whose few assumptions you can defend beats a sprawling one whose forty you cannot.

3. Letting the output kill the debate. The moment a precise number appears, questioning it feels like nit-picking. Keep asking "how much would this move if I'm wrong about growth or the rate?"

4. Optimising the wrong thing. Effort spent adding decimals is effort not spent widening the range and stress-testing the two or three inputs that actually drive the answer.

5. Comparing a point value to the price as if both were exact. "₹166 beats ₹150" is only meaningful if ₹150 sits outside your value band. Inside it, the DCF is silent, not bullish.