Four Free Cash Flow Definitions That Give Four Answers

Free cash flow is not one number. It is at least four, they come off the same cash flow statement, and on a typical large company they can sit fifteen or twenty percent apart. If you have ever built a model, got a fair value, then compared it to a free cash flow yield you read somewhere and found the two irreconcilable, this is usually why. Nobody was wrong. You were using two definitions and treating them as one.

The fix takes about twenty minutes per company and it is the kind of housekeeping that pays for itself the first time it stops you buying something on a yield that was never really there.

One cash flow statement, four numbers

Take an illustrative filing set. These are round numbers chosen to make the arithmetic legible, not figures from any real company. Cash from operations is 1,000. Capital expenditure is 300, of which you judge 200 to be maintenance and 100 to be growth. Cash interest paid is 60 and the tax rate is 25 percent. Net new borrowing over the year is 100. Share-based compensation of 120 was added back as a non-cash charge inside that 1,000 of operating cash flow.

Now run the four definitions.

The gap between the highest and lowest is 120, which is roughly 18 percent of the lowest. Put that through a valuation and an 18 percent difference in the cash flow line is an 18 percent difference in the answer before you have argued about a single growth assumption. It is larger than the discount you were probably hoping to buy at.

Each number is doing something specific. Free cash flow to the firm is the cash available to everybody who funded the business, so interest is added back after tax because it is a payment to one class of funder rather than a cost of operating. Free cash flow to equity is what is left for shareholders after the lenders have been served, which is why new borrowing counts as a source. Owner earnings tries to answer a different question entirely: what could be pulled out of the business each year without shrinking it, which is why only maintenance spending is deducted and why stock comp comes back out. Reported free cash flow is a convention, usually operating cash flow minus all capex, and its main virtue is that it is consistent within one company's own reporting history.

What each method is entitled to use

The rule is short and there are no exceptions worth learning as a private investor.

• Discounting at a weighted average cost of capital requires free cash flow to the firm. The result is an enterprise value. You then subtract net debt to reach equity value.
• Discounting at a cost of equity requires free cash flow to equity. The result is already an equity value. You do not subtract debt again.
• An owner yield check, meaning the cash a full owner could take out divided by the price of the whole business, uses owner earnings.
• A free cash flow yield comparison across several companies uses reported free cash flow, but only if you compute it yourself the same way for every name rather than lifting each company's own headline figure.

That last point catches people. Companies define their own headline free cash flow and the definitions differ. One may deduct capitalised software, another may not. One may include finance lease payments, another may not. Comparing two published figures is comparing two accounting policies.

The double count that shows up in every second retail model

Two errors account for most of the damage, and both are silent because the model still produces a plausible looking number.

The first is discounting free cash flow to equity at a weighted average cost of capital. The cash flow has already had the lenders paid out of it, and the discount rate has then been lowered to reflect cheap debt, so the benefit of leverage gets counted twice. On a business with meaningful debt this inflates the answer by a chunk that grows with leverage, which is exactly backwards.

The second is arriving at an equity value from free cash flow to equity and then subtracting net debt anyway, out of habit. Say a model produces 800 of equity value and the company carries 200 of net debt. Subtracting again gives 600, a 25 percent haircut applied for no reason. On a 5,000 dollar position that is the difference between a name that looks cheap and one that never makes your shortlist.

There is a third error that is less about arithmetic and more about self-deception. Using reported free cash flow, where stock comp has been added back and never deducted, then valuing the result on today's share count. The cash was never paid out, so operating cash flow is genuinely higher. But the compensation was real and it was settled in your ownership stake. Either deduct it from the cash flow, as owner earnings does, or model the share count rising. Doing neither counts the benefit and skips the bill.

Where the screen ends and your definition starts

A composite screen is good at the part of the job that is tedious, which is ranking a large universe consistently so you are not reading filings at random. The Company Valuation Engine scores thousands of listed names into undervalued, fairly valued and overvalued buckets, and at capture it had 4,420 companies on the board with 2,207 reading below fair value. That is a starting shortlist and a reasonable one.

What it is not is a substitute for your definition choice. When your own model disagrees with a screen verdict, the first thing to check is not growth rates or discount rates. It is whether the two of you are even discounting the same cash flow. In my experience most of those apparent disagreements dissolve at that step, and the ones that survive are the ones worth spending an evening on.

A definition line at the top of the model

The habit that fixes this permanently costs one line. At the top of every valuation sheet, write the definition in words, not a formula reference: cash from operations plus after-tax interest minus total capex, discounted at WACC, enterprise value output. Then, when you add a comparable next month, that line tells you what to compute for the new name instead of leaving you to reconstruct what past you meant.

This week, pick the largest holding you own and compute all four numbers for it from the last annual report. Twenty minutes. If they land within a few percent of each other, the company is lightly levered, spends close to maintenance, and pays little in stock, and you can use whichever you like. If they land 20 percent apart, you have just found out that the yield you thought you owned depends entirely on which definition someone else picked for you.