Why Growing Free Cash Flow Directly Can Mislead Your DCF

Question:
I used to build FCF DCFs by taking CFO minus total capex, then running a two-stage model. For the high-growth period, I’d take the average of (last year FCF − this year FCF), use that as a growth proxy for 10 years, and taper it down toward GDP growth in stage one.
Are you saying this approach is wrong?

Answer:
It’s not complete nonsense, but yes—your process is fundamentally mis-specified for what you think you’re modeling, and it will fail precisely in the edge cases you’re worried about.

You did something very common:

You took a residual output (FCF),
Treated it as if it were a primary business driver,
And then extrapolated it.

That’s the core flaw.

1. What you were actually doing

Your old pipeline looked roughly like this:

Approximate FCFF as:
FCFF ≈ CFO − total capex ✅ (acceptable as a rough starting point)
Inspect historical FCF.
Calculate some average change or growth in FCF.
Build a two-stage DCF:
- Stage 1 (10 years): grow FCF at that historical rate, fading toward GDP.
- Stage 2: terminal value at a GDP-like growth rate.

In other words, you were growing the FCF line directly, instead of modeling the underlying economics that create that FCF.

That’s where the model breaks.

2. Why projecting FCF itself is structurally risky

Conceptually:

FCF = NOPAT − Reinvestment

where reinvestment includes capex (net of D&A), working capital investment, and any other increments to invested capital.

So FCF is:

Not a clean “operating” metric,
But the end result of:
- Growth rate,
- Capital intensity,
- Where you are in the cycle,
- One-offs in working capital, tax timing, etc.

Typical failure modes

a) High-quality growth business

High ROIC, heavy reinvestment → low or negative FCF.
Historical FCF looks weak or volatile.
Your method says: “low/choppy FCF → low value.”
Reality: the business is compounding NOPAT at high returns; intrinsic value is strong. You undervalue.

b) Melting ice cube

Management stops reinvesting; capex sits below depreciation.
Near-term FCF looks fantastic.
Your method says: “FCF trending nicely → let’s project it forward.”
Reality: they’re liquidating the franchise; earnings power will roll over. You overvalue.

c) Working capital whiplash

One large working capital inflow (running down inventory, stretching payables, etc.) spikes FCF.
That noise gets embedded in your “growth rate.”
You treat timing effects as if they were structural economics.

Implicitly, your old method assumes:

“The historical pattern of FCF is a reliable proxy for the future structural cash generation of the business.”

That assumption simply does not hold in many real-world cases.

3. Why the ROIC / reinvestment framing is a step-change

The ROIC/reinvestment framework fixes this by shifting focus to the engine instead of the exhaust.

Instead of:

“FCF grew around X%, so I’ll grow FCF at X%.”

You move to:

Growth in revenue / NOPAT
– What is a realistic growth path given industry structure, competitive position, and scale?
ROIC level
– What returns on incremental capital are sustainable given history and competition?
Reinvestment rate implied by growth
– Reinvestment / NOPAT ≈ g / ROIC.
Derive FCFF = NOPAT − Reinvestment each year.

Now you are explicitly modeling:

Growth,
Capital intensity,
Returns on capital,

and letting FCF be the output, not the input.

That’s why this framework feels like you’re “seeing through” the numbers: you’ve stopped asking a residual to do the job of a primary driver.

4. How to think about your old process

Parts that were workable

Using CFO − capex as a first-pass proxy for FCF: acceptable in many cases.
Two-stage DCF structure: standard.
Tapering long-term growth toward GDP/inflation: sensible.

Parts that were structurally off

Using historical FCF growth or change as the central forecasting hook.
Not decomposing whether FCF movements came from:
- Real economic growth vs. under/over-investment,
- Changes in ROIC,
- Shifts in capital intensity,
- One-off working capital effects.

The math in your sheet “worked,” but it was solving the wrong economic problem.

Think of it this way:

Old method: “Extrapolate whatever cash fell out after all decisions were made.”
New method: “Model the decisions (growth, reinvestment, returns) and then see what cash falls out.”

The second is how professional investors and corporate finance teams actually think about DCF, even if they don’t always write it in those exact terms.

5. Practical way forward for a two-stage DCF

Going forward, a cleaner structure looks like:

Forecast revenue and margins → derive NOPAT.
Tie capex and working capital to that growth using realistic capital intensity assumptions.
From that, compute FCFF year by year.
Sanity-check the economics:
- Does g ≈ ROIC × reinvestment rate roughly hold over time?
- If it doesn’t, adjust growth, ROIC, or reinvestment until the story is coherent.
Terminal stage:
- Use modest, GDP-ish long-term growth.
- Terminal reinvestment ≈ (g / ROIC) × terminal NOPAT.

So yes, your earlier approach was directionally trying to do a DCF, but it leaned on the wrong variable. The upgrade you’re making now—shifting from “grow FCF” to “model growth and returns, let FCF be implied”—is exactly the step that separates a back-of-the-envelope model from a proper valuation framework.