How to Value a Stock
To value a stock, start with what the business earns and how reliably, then apply either a multiple against comparable companies or a discounted cash flow. Both produce a range rather than a figure, and the assumptions behind them matter more than the arithmetic.
Valuing a company means estimating what its future earnings are worth today. Every method does this, and the differences between them are mostly about how honest they are regarding the size of the guess involved.
Before you start
A written view of what the business does and how it earns, before any number is calculated. If you cannot describe where the money comes from in three sentences, no model will fix that.
At least two comparable companies, because every multiple is a relative measure. A multiple in isolation is a number with no unit.
An honest statement of what you would have to believe for the number to hold. This is the most useful output of the whole exercise.
The steps
1. Describe the business before valuing it
What is sold, to whom, at what margin, and what would stop that. Three sentences. Everything downstream depends on this and no arithmetic substitutes for it.
2. Establish what it currently earns
Use a figure that represents a normal year rather than the highest or the most recent. Averaging across a cycle is cruder and considerably more honest than picking a peak.
3. Compare multiples against real peers
Price against earnings, sales or book value, set beside two or three genuinely similar companies. This tells you how the market prices this business relative to that group and nothing more.
4. Build a discounted cash flow if the earnings are predictable
Project cash flows, discount them back, add a terminal value. It is the most rigorous method available and it is unusable on a business whose earnings cannot be forecast.
5. Test how sensitive the answer is
Change the growth rate by a point and the discount rate by a point, and see how far the answer moves. On most models it moves enormously, which is the single most informative thing the exercise produces.
6. Express the result as a range
A range between two defensible assumption sets. A single figure implies a precision the inputs do not contain, and it is the form in which valuations most often mislead the person who built them.
7. Write down what has to be true
“This holds if revenue grows 8% a year for five years and margins stay flat.” That sentence can be checked against reality in eighteen months. A number cannot.
How to tell it worked
The business was described in 3 sentences before any figure was calculated.
The result is a range, with both ends traceable to stated assumptions.
Sensitivity was tested by moving 2 inputs by 1 percentage point each.
And the belief statement is written down, so it can be checked in 12 months.
What every method shares
They are all forecasts. The arithmetic is arithmetic; the inputs are opinions about the future, and a model’s apparent rigour attaches to the wrong half of the exercise.
None of them says when. A company can be worth twice its price and stay there for years. On this site’s shared series 95% of bars sat below a prior peak and the longest recovery took 73 bars — being right early is indistinguishable from being wrong while it is happening.
The two honest uses
Deciding whether a price is defensible. Not what the company is worth, but whether the current price requires assumptions a reasonable person would make. That is answerable.
Finding out what the market believes. Reverse the model: solve for the growth rate that justifies today’s price. If the answer is implausible, you have learned something specific rather than produced a target.
Both are questions about the price rather than about the value, which is the framing that makes valuation useful instead of decorative.
Which method suits which business
Stable, predictable earnings suit a discounted cash flow. Utilities, established consumer brands, anything whose revenue next year is a small variation on this year’s. The forecast is the method’s weak point, and here the forecast is defensible.
Cyclical businesses suit an averaged multiple across a full cycle. Valuing a miner at peak earnings or a housebuilder at the bottom produces answers that describe the moment rather than the company.
Asset-heavy businesses suit a book-value comparison. What the company owns is a meaningful anchor when a large share of its value sits in property or equipment rather than in future growth.
And businesses without profits need a different question entirely. A multiple of revenue, or a sentence about what margin would eventually have to be achieved — which is a forecast presented plainly instead of hidden inside a model.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 2 mention valuing a stock in the
title, at a median of 130,441 views — both instruction-shaped — and 1 covers intrinsic value at 1,019,621
views. Fundamental analysis appears in 49 at 7,377. The counts come from site/corpus_count.py.
3 videos in 24,971, and one of them has over a million views. The largest audience per video of any subject measured on this site sits on a topic with three instances of coverage, which is as clear a gap between demand and supply as the corpus contains.
The answer to the question on that chart is that a model saying double usually means an assumption is doing all the work. Move the growth rate down a point and see whether the conclusion survives — and if the gap is real, it can persist for years, which makes position size a more important decision than the valuation was.
When it fails
The failure is the model built backwards from a conclusion, and nobody does it deliberately. You already like the company. The first set of assumptions produces a value below the price, so the growth rate gets revisited — reasonably, because the original was conservative. Then the discount rate, also reasonably. Each adjustment is defensible in isolation and the finished model produces exactly the answer you started with, wearing the authority of a spreadsheet.
The second failure is a single-point answer. It conceals its own error bars.
A third is using a peak earnings figure. A normal year is the honest input.
A fourth is comparing multiples across industries. They are relative measures.
A fifth is a discounted cash flow on unpredictable earnings. The method needs forecastability.
And a sixth is treating a valuation as a timing signal. It contains no date.
Related
Valuation covers the methods and where each applies. Intrinsic value is the concept underneath all of them. And discounted cash flow is the most rigorous method and its two weak points.
I stopped trying to arrive at a number and started trying to arrive at a sentence: what would have to be true for this to be worth what it is trading at. Sometimes that sentence is plausible and sometimes it requires the company to double its margin while growing, and the second case is far more useful than any figure I could have produced.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.