Precisely Imprecise: How to Avoid False Precision in DCF Models
Standard modeling and valuation practices are backwards
It’s hard to find someone today who will tell you that they find discounted cash flow analysis useful and make it a part of their investment process. The popular opinion paints the DCF model as a defunct and deceptive old tool that is made irrelevant if your qualitative and strategic understanding is up to par.
The criticism is well-deserved but misplaced. The flaws of the DCF aren’t in the tool itself but in the way it is applied - a case of “blame the archer, not the arrows.” As Michael Mauboussin reminds us in his paper, Everything is a DCF Model, you are using a DCF whether you know it or not. After all, it’s how the market values cash-generating assets.
Conventional wisdom today says to combine historical financial information with your view about how the future will play out and project a company’s performance line by line. Then use your projections to “calculate” the company’s “value” and compare it to today’s market price. Better yet, forecast EPS three years out and slap a “market multiple” on those earnings.
The problem with these methods is obvious. The future is unknowable and so governed by randomness that any estimates you make will be either wrong or lucky. How can you know what SG&A expense will be seven years from now? This false rigor leads to confident investments based on a shaky foundation of layered assumptions and excel formulas.
The main objection to all financial models, including those that use a DCF, is that they give you a false sense of precision - and I agree.
Instead, I would propose that it’s more valuable to shift your attention to the far more calculable question of “what’s priced in?” Using the same principles of the DCF, you can understand a company’s current market value, roughly measure what is priced in, and then make a qualitative judgment about whether you agree or disagree.
“What is Peloton worth?” is an abstract question. But “will Peloton have more than five million subscribers by 2026?” is very tangible. And it’s a question for an investor, not a spreadsheet.
By being precisely imprecise - understanding which parts of the model are important and which are not - you can simplify your model and ground your reasoning in the real, not the abstract.
The Reverse-DCF
"Stock prices are the clearest and most reliable signal of the market's expectations about a company's future financial performance,” Mauboussin writes in Expectations Investing. The framework laid out in this book, centered on the reverse-DCF, allows you to decode and approximately calculate what level of expectations are implied by a company’s share price.
For this to be useful, you have to believe a few fundamental things that seem intuitive, but not when you look at the state of the market today:
Stock prices change as a result of revisions in investor expectations. The market is forward-looking, and prices are set based on unknowable future cashflows. When those expectations change, prices move to reflect the adjustment.
“Value,” however you define it, comes from a discrepancy between what is priced in and what plays out in reality. Mauboussin often talks about a chapter in a book called Crist on Value, in which Steven Crist, a famous horse racing handicapper, explains that successful betting is not about picking which horse is going to win, but rather which horse is going to perform the best relative to the odds set before the race. It’s a simple and elegant reminder that expectations are everything.
In a reverse-DCF model, you manipulate the variables until the value equals the current share price. Provided that the numbers reflect the broader market’s assumptions rather than your own, it will give you a rough idea of the expectations that are priced in.
This is valuable because a rough idea is all you should need. After doing a few of them, you’ll quickly realize that the price-implied expectations are really only sensitive to a few variables. This is the key to avoiding the pitfalls of modeling. Find what matters for a company and focus on that - the rest is noise that should be ball-parked.
Where to be precise and imprecise
To avoid false precision, you need to be thoughtful about what to focus on and what to ignore. You want the model to be detailed enough to be useful, but not so detailed that it bogs you down.
To know where to focus, look at the various levers that can be pulled in a DCF:
Things that affect the amount of free cash flow
Revenue
Operating costs
Investments
Taxes
Competitive advantage period (number of years before terminal value kicks in)
Terminal growth rate
Things that affect the discount rate applied to that free cash flow
Risk-free rate
Equity risk premium (best source is Aswath Damodaran’s site)
Beta
Then think about where your opinion is likely to differ materially from the rest of the market. Spoiler alert, it’s almost always revenue. Operating costs and investments, expressed as a free cash flow margin, are a distant second. The emphasis on revenue is especially true for high-growth companies.
Your opinion doesn’t come into play much in determining the discount rate. Beta is an imperfect measure that involves some subjectivity, but the risk-free rate and risk premium are more or less set.
You may have a variant view on how much a given company will need to invest in capex to compete in the near future. Or you may think they will be more operationally efficient resulting in lower SG&A or R&D. These expenses are ultimately reflected in a free cash flow margin.
Remember that as an investor, you are trying to anticipate revisions in market expectations. The difference in your opinion and the market’s needs to be wide enough that, if you are correct, the stock moves as a result of the market arriving at your view. As Mauboussin writes, “not all expectations revisions are equal.”
Changes in revenue expectations are by far the most important because they happen the most frequently, are typically the largest in magnitude, and have the biggest impact on shareholder value. Margins, especially for mature businesses are less dynamic, largely set by the nature of their industry, and have a smaller impact on the model. Changes to a company’s beta, tax rate, or the equity risk premium matter even less.
The slow way to figure this out is to do a handful of these reverse-DCFs and notice what happens to the value when you manipulate different variables. A 5% increase in the revenue growth rate could take an opportunity from boring to interesting, but a 5% decrease in operating expenses won’t change much for most businesses.
Different companies are sensitive to different variables - a hyper-growth SaaS business will be triggered by revenue, a slow-growing stalwart with high net income will be affected more by tweaking margins.
The point of being precisely imprecise is to be thoughtful about which metrics are most important, get a feel for the market’s expectations for them, then close the model and reason independently why your expectations are higher or lower.
Note: this framework is most applicable to a certain breed of investing. I’m looking through the lens of owning high-quality companies at a good price and holding them for a long time. If you’re searching for things like net-nets, hidden assets, or event-driven trades, this isn’t a useful framework. In those cases, your model does need to be accurate. Being precisely imprecise is for those looking for deltas between what’s priced in and what is realistic so wide you can drive a truck through them.