There is a hierarchy people absorb early and never question: a DCF is real analysis, a multiple is a shortcut for people who could not be bothered. It is wrong often enough to cost money. For a stable, low capital intensity business with predictable earnings, a well chosen multiple is a more accurate estimator than a ten year cash flow forecast, because the forecast adds error faster than it adds information. For a capital intensive business partway through an investment cycle, the multiple is close to useless and the DCF is the only thing that can see what is happening.
The choice is not about rigour. It is about which method's assumptions match the company in front of you. Three questions settle it, and they take about five minutes.
The three questions that pick the method
How far out do you need to see before the business is recognisable? If the company in three years looks like the company today with more of it, the near term earnings number contains most of the information and a multiple works. If the business is mid transition, launching a product line, or finishing a large build, then the current earnings number describes something that is about to stop existing, and only an explicit forecast captures that.
How stable are the earnings? Pull the last five to seven years of earnings and look at the spread. If the worst year is within a reasonable band of the best year, the earnings figure means something and a multiple applied to it means something. If the range spans a loss and a record, then the multiple you compute depends entirely on which year you happened to pick, and a normalised cash flow forecast is doing real work that a P/E cannot.
How capital intensive is it? This is the question most retail investors skip and it is the most decisive. Compare capital expenditure to depreciation over several years. Where the two run close together, earnings and cash flow track each other reasonably and a P/E is a defensible proxy for cash generation. Where capex runs at twice depreciation for years, earnings systematically overstate cash available to shareholders, and every multiple computed on earnings inherits that error.
Two yes answers on stability and low capital intensity, plus a recognisable business, and I use a multiple and stop. Any no, and I build the forecast. The mistake is not choosing wrong. It is not choosing at all and defaulting to whichever method is fashionable in your feed that month.
That column is where I start when a screen surfaces a name. The Company Valuation Engine leaderboard puts a trailing P/E next to a composite score and a one word verdict, and the useful moment is not when they agree, it is when they do not. A name showing a modest multiple and an over fair value verdict is telling you the multiple is resting on earnings the rest of the model does not believe are sustainable. A name showing a high multiple with a constructive read is usually a business whose cash flows are far better than its accounting earnings. Both cases are worth twenty minutes. Agreement, like on the two rows above, is worth none.
What a multiple is actually claiming
It helps to know what number a P/E has to clear to be justified on cash flow grounds, because it makes expensive multiples concrete rather than a matter of opinion.
A mature business generating a stable stream of cash, discounted at 9 percent and growing forever at 2.5 percent, is worth about 15.8 times that cash flow. Move the discount rate to 10 percent and it is 13.7 times. Move the growth to 3 percent at 9 percent discount and it is 17.2 times.
So the honest reading of a 33 times multiple is that the price is roughly double what a no growth mature business justifies, and the difference has to be made up by growth over an extended period. That may well be right. The point is that the multiple is not an alternative to having a growth view, it is a growth view expressed in one number, and most people quoting multiples have not noticed they are making the claim.
The comparison also tells you when a low multiple is not cheap. A cyclical business at 8 times peak earnings is more expensive than it looks, because the earnings in the denominator are the ones that will not repeat. The multiple did not lie. It answered exactly the question you asked, which was the wrong question.
The four situations where a multiple misleads
These are the cases where I have been burned, in rough order of how often.
First, peak cycle earnings. Semis, shipping, energy, housing, anything where the earnings series has a shape rather than a slope. The multiple looks lowest exactly when the risk is highest, which is the opposite of useful. If the industry has a cycle, apply the multiple to mid cycle earnings you estimate yourself, not to the trailing figure.
Second, capital intensity mismatch. A company spending well above depreciation to grow reports earnings that overstate distributable cash. Two companies at the same P/E, one running capex at depreciation and one at double it, are not similarly priced. The second is materially more expensive and nothing in the P/E shows it.
Third, balance sheet differences. P/E ignores the capital structure entirely. A debt free company and a heavily leveraged one at the same P/E carry different risk and different value. This is the standard argument for enterprise value based multiples instead, and it is a good one, but the simpler discipline is just to look at net debt before you compare any two P/E figures.
Fourth, accounting that diverges from cash. Large non cash charges, aggressive capitalisation of costs, or heavy share based compensation all put distance between the earnings number and cash. Where that gap is wide, valuing on earnings is valuing on an accounting convention, and the divergence tends to be largest in precisely the companies people most want to own.
The five minute version, and the decision it produces
Here is the whole routine, in order, for a name that has come off a screen.
- Pull seven years of earnings and note the high, the low, and whether there is a cycle. Two minutes.
- Pull capex and depreciation for the same period and compute the ratio. One minute.
- Look at net debt. Thirty seconds.
- Decide the method from the three questions and write down which one you chose and why, in one line.
If the answer is multiple, apply it to normalised earnings rather than trailing earnings, and compare against the 15 to 17 times benchmark a mature business justifies rather than against the sector average. Comparing to the sector average tells you whether a name is cheap relative to other things that may all be expensive, which is a different and much weaker statement.
If the answer is DCF, accept that you are now committing an afternoon rather than five minutes, and be honest about whether the position size justifies it. For a position of a few hundred dollars in a diversified holding, a normalised multiple and a hard rule about not paying more than a stated level is a rational stopping point. Building a ten year model to support a small position is not diligence, it is a way of feeling diligent, and the error band on the resulting fair value is wide enough that the extra hours did not change the decision.
The one thing worth doing regardless of method is writing the decision down before you look at anyone else's number. Method choice is where anchoring does the most damage, because whichever method makes a name you already like look cheaper is the method you will find reasons to prefer.