Every DCF model rests on five numbers. Change any one of them by a reasonable amount and the "intrinsic value" output swings by 30-100%. That doesn't make DCF useless — it makes the sensitivity bands more important than the point estimate. This article is the checklist you should run before believing any DCF, including your own.

The Five Assumptions That Move Everything

g
Revenue growth rate
Year-over-year top-line growth, usually projected for 5-10 explicit years. The hardest one to estimate; small changes compound dramatically. A 12% vs 8% growth assumption can swing terminal value by > 50%.
m
Operating margin
Where margins land at maturity. Almost everyone projects margin expansion; almost no business sustains it. Stress-test by running the model at current margins, not aspirational ones.
cx
Capex intensity
Capex as % of revenue. Capital-light businesses (software) are forgiving; capital-heavy businesses (semis, autos) bury value here. Maintenance vs growth capex split matters more than total.
g∞
Terminal growth rate
Perpetual growth after the explicit forecast period. Usually tucked at 2-3%. A change from 2% to 3% can swing terminal value by 25%+. The most quietly destructive assumption.
r
Discount rate / WACC
Cost of capital. A 1pp change moves valuation 10-15%. Most retail DCFs use a single hard-coded WACC across all 10 years; in reality WACC changes with leverage, rates, and risk profile.

The Sensitivity Test That Saves You

Before quoting a DCF result, run a ±20% sensitivity grid on each of the five inputs:

  1. Hold four assumptions constant; vary one ±20%.
  2. Record the resulting intrinsic value range.
  3. Repeat for each of the five.
  4. The output is no longer "the stock is worth $200" — it is "the stock is worth $140-280 depending on which assumption you stress".

A model whose output ranges from $140 to $280 is a much weaker basis for action than the headline "$200" suggests. If price is at $190, you do not have margin of safety against the bear case — you only have it against the midpoint.

Damodaran calls DCF "a story told in numbers." The numbers are derivatives of the story; if the story is wrong, no precision in the model recovers it. The sensitivity grid forces you to admit which parts of the story you are betting most heavily on.

When DCF Is Worth Doing Anyway

DCF works best for stable cash-generating businesses where 4 of the 5 inputs are reasonably knowable (utilities, regulated infrastructure, mature consumer staples). It fails hardest where it is most often used — speculative growth stocks where every input is essentially a guess. Use multiples-based valuation cross-checks, not DCF, for those.

DCF rests on 5 numbers: g · m · cx · g∞ · r
Run ±20% sensitivity on each before believing the headline output
Best for stable cash businesses · weakest for the speculative growth where it's most used

See multi-method valuation in the app →