Limitations
DCF is a powerful tool but it breaks badly on certain business types. Here are the categories where this model will produce unreliable or meaningless output — and why.
Auto/industrial companies with captive finance subsidiaries (Ford, GM, GE Capital)
Financial subsidiary debt inflates invested capital massively, creating an inverse effect (massive deflation) on return-on-invested-capital. Operating and financial earnings are commingled in reported financials so manual statement splitting would be required.
Banks and insurance companies
A dividend discount model or excess return model works way better as there is no distinction between debt and equity, since the debt a bank takes on is what it then lends out, so debt is the product itself.
Real estate companies & investment trusts (REITs)
Value comes from asset appreciation and rental yields, not operating cash flows. Standard DCF misses the property value component entirely. FFO (funds from operations) is the correct metric, not FCFF.
Early stage / pre-revenue startups
No historical series to anchor assumptions on, negative ROIC for entire history, and value is almost entirely based on future products that DCF can't capture.
Commodity companies (oil, mining, steel)
Earnings are driven by commodity prices, so commodity price forecasts are more reliable for finding a fair EV.
Turnaround companies in distress
Negative invested capital or negative equity distorts every ratio, and the DCF can't model bankruptcy optionality (equity has value even when enterprise value is negative). Many US auto companies are an example.
Holding companies and conglomerates (Berkshire, SoftBank)
Value is sum of stakes in unrelated businesses each requiring different assumptions. Consolidated financials blend incompatible business models; can't use a single set of long-term assumptions from the NYU tables as currently done.
Companies with massive off-balance-sheet obligations
Operating leases, pension obligations, and contingent liabilities that aren't in total_debt_series understate the actual $ of employed capital and overstate ROIC.
Pharmaceutical/biotech companies with binary pipeline
Most of the value is in whether whatever new drug the company invested massively into gets approved.
Network effect platform businesses
Reinvestment generates nonlinear returns.
Companies undergoing major business model transitions
Historical series give information about a business that no longer exists. Forward multiples (looking at what a typical business with the model the firm is transitioning into is worth) based on price-to-sales ratios, for example, would value more fairly.
This list reflects the author's own experience running this model across hundreds of companies. The specificity here is intentional — knowing exactly where a tool fails is as important as knowing where it works.