We were hiding 20% of every company's value inside a single number. Our two valuation models disagreed by 85%. And we were giving credit for data we didn't have. Here's how we fixed all of it.

The Problem: Terminal Value Ate Our Model

If you've ever run a DCF and noticed that 75-80% of the intrinsic value comes from the terminal value, that lump sum representing "everything after Year 5", you're not alone. It's the most common criticism of discounted cash flow analysis, and frankly, it used to apply to Tenets too.

Our original DCF engine projected free cash flow for 5 years at a constant growth rate, then slapped on a perpetuity-based terminal value at Year 5. Simple. Clean. And problematic.

A large terminal value is not directly the issue. Damodaran himself addresses this directly in his NYU valuation lectures:

"If you are valuing equity in a going concern with a long life, you should not be surprised to see the terminal value account for a high percentage of value. Contrary to what some may tell you, this is not a flaw in your valuation but a reflection of reality."

The real issue is that a 5-year projection forces too much of the analysis into a single opaque assumption. When terminal value is 76% of your enterprise value (as it was in our old model for a typical growth company), you're essentially saying: "I carefully analyzed 5 years of cash flows that represent 24% of the value, then trusted one formula for the other 76%."

That's too far away of being an analysis. It's a rounding error with extra steps.

What Changed: The Two-Phase Model

We've upgraded the Tenets DCF engine to Damodaran's canonical two-phase model. It now uses the same structure he uses in his NYU classroom spreadsheets (the famous fcffsimpleginzu.xlsx) and teaches in Investment Valuation.

Here's the difference:

Old model (5-year, single-phase)

• Years 1–5: FCFF grows at a constant estimated rate

• Year 5: Terminal value (perpetuity at 2.5% GDP growth)

• Total projection horizon: 5 years

New model (10-year, two-phase)

• Phase 1 (Years 1–5): FCFF grows at the estimated high-growth rate

• Phase 2 (Years 6–10): Growth fades linearly from the Phase 1 rate down to 2.5%

• Year 10: Terminal value (perpetuity at 2.5% GDP growth)

• Total projection horizon: 10 years

Why "fade" instead of a cliff?

No real company goes from growing at 10% one year to suddenly growing at 2.5% the next. That's what our old model implicitly assumed a hard cliff at Year 5 where growth instantly drops to the economy-wide rate.

The two-phase model is more honest. It says: "We think this company will grow at 10% for about 5 years, then competitive pressures, market saturation, and the law of large numbers will gradually erode that advantage over the following 5 years, until growth converges to the long-term GDP rate."

For a company with 10% estimated growth, the Phase 2 fade looks like this:

The linear interpolation is Damodaran's standard approach: simple, transparent, and mechanically correct.

The Terminal Value Gets Smaller (And That's the Point)

Here's the punchline: by projecting cash flows for 10 years instead of 5, we move a significant chunk of value out of the terminal black box and into the explicit projection: where you can actually see it, question it, and stress-test it.

Here's what this looks like for Adobe (ADBE), a real-world growth company with 8.6% estimated growth and a 10.4% WACC:

That 22.4% in Phase 2 was always there. It was just hidden inside the terminal value. Now it's visible. You can see exactly how much value comes from the fade period versus the terminal perpetuity.

Terminal value drops from 76% to 51%. Nearly half the value now comes from individually projected cash flows, not a single perpetuity assumption. The engine's interpretation: "Balanced between near-term cash flows and terminal value."

The practitioner rule of thumb is instructive here: if terminal value exceeds 80% of enterprise value, your model's reliability is heavily dependent on a single assumption. Our old model was at 76%; uncomfortably close to that line. The new model at 51% is in healthy territory.

Impact Across Company Types

The magnitude of the change depends on how fast the company is growing:

All figures use illustrative parameters, not live Tenets output.

The pattern is exactly what the theory predicts:

• High-growth companies see the biggest increase. They have the most above-terminal growth to capture. The old 5-year model was systematically undervaluing them by cramming 5 years of above-average growth into the terminal value.

• Stable companies barely change. When growth is already close to 2.5%, the fade phase adds almost nothing, there's very little gap to fade through.

• Distressed value (growth below terminal) actually decreases slightly. This is correct and subtle: when a company grows at 2% (below the 2.5% terminal rate), the fade phase actually increases growth from 2% toward 2.5%. But the extra 5 years of discounting at a high WACC (12%) more than offsets this. Net result: a small decrease.

This last case is a good sanity check. If the upgrade made everything go up, we'd worry about systematic bias. The fact that it can go both directions means the model is doing math, not marketing.

The Scenario Spread Widens

The bear/base/bull scenario sensitivity also changes:

The spread widens by 27%. Which is the model effectively being more honest about uncertainty. When you give growth assumptions 10 years to compound instead of 5, the difference between optimistic and pessimistic scenarios has more room to diverge.

In practice, this means the model gives you a wider confidence interval, which is appropriate for forward-looking valuation. Damodaran's own terminal value research shows this relationship clearly: the higher the excess growth rate relative to cost of equity, the more the high-growth period assumptions matter to the final value: even though terminal value is still a large percentage of the total.

What We Didn't Change

A few design decisions worth noting:

WACC stays constant across all 10 years. Damodaran sometimes adjusts WACC in Phase 2 (as the company matures, beta declines, capital structure stabilizes). We kept it constant for now. This is what his basic spreadsheet does, and what most practitioners use. Dynamic WACC is on our roadmap but adds complexity without proportional accuracy gains for most companies.

Terminal value still uses perpetuity growth with exit-multiple fallback. The terminal value formula is the same, only when it's applied changes (Year 10 instead of Year 5). If WACC falls below the terminal growth rate (rare but possible), we still fall back to a sector-appropriate EV/EBITDA exit multiple.

Bear/base/bull deltas are the same magnitude. Bear = base growth − 2 percentage points, WACC + 1pp. Bull = the opposite. The deltas now apply to the Phase 1 growth rate, with Phase 2 always fading to terminal. This means the scenarios are testing "what if the high-growth period is stronger/weaker than estimated?" — which is the right question to ask.

Backward compatibility. The API still returns cashflow_pct (the combined Phase 1 + Phase 2 percentage) alongside the new phase1_pct and phase2_pct fields, so nothing breaks for existing integrations.

How the Growth Estimate Works Now

The growth rate that feeds Phase 1 comes from the median of up to three independent signals:

1. Historical FCFF CAGR: How fast has the company's free cash flow actually grown? (Backward-looking, but anchors expectations.)

2. Analyst consensus: earnings growth and revenue growth estimates from analyst coverage, where available. (Forward-looking market signal.)

3. Fundamental growth: Reinvestment Rate × Return on Invested Capital. This is the Damodaran-canonical approach: a company can only grow as fast as its reinvestment rate allows, given its returns on that investment. We now compute the reinvestment rate as a median across all available annual periods (not just the latest year), which smooths out one-time capex spikes. And we cap fundamental growth at ROIC, because if a company is reinvesting more than it earns, that's not sustainable growth.

The median-of-three approach means no single signal dominates. If analysts are wildly bullish but historical growth is modest and fundamentals suggest moderate reinvestment, the estimate gravitates to the middle. This is where you want to be when projecting a decade of cash flows.

Five More Things We Fixed

The two-phase DCF was the headline upgrade, but we shipped five more changes alongside it. Each one closes a gap we found during our engine audit.

1. Owner Earnings DCF: From $720 to $508

Our biggest embarrassment was that two DCF models, both analyzing the same company, disagreed by 85%.

For Adobe, the Buffett Owner Earnings DCF was showing $720. The Damodaran FCFF DCF was showing $389. Same company, same data, wildly different answers. That's more than "model diversity" it's an internal contradiction.

The root cause: the Owner Earnings DCF was using a fixed 10% discount rate and a single-phase growth projection (no fade). The Damodaran DCF was using a dynamic WACC and now had the two-phase fade. Different discount rates. Different growth structures. Of course they disagreed.

The fix: We upgraded the OE DCF to use a dynamic cost of equity via CAPM (risk-free rate + beta × equity risk premium) and the same two-phase growth schedule as the Damodaran model. Critically, we kept the discount rate as cost of equity — not WACC — because Owner Earnings is an equity-level cash flow. Discounting equity cash flows at WACC would be a textbook error; discounting them at cost of equity is what the CFA curriculum prescribes.

Result: Adobe's OE DCF dropped from $720 → $508. The gap between the two DCF models compressed from 85% to 30.5%. The composite model spread dropped from 161% to 123%. Still wide (the Graham Number at $104 anchors the low end — that's Graham's framework working as designed for a high-P/B tech company), but dramatically more coherent.

2. Margin Convergence Flag

Here's something that should make you nervous about any DCF: Adobe's operating margin is 37.9%. The Technology sector median is 22.0%. That's 1.72× the sector median.

Our terminal value assumes Adobe's cash flows grow at 2.5% forever. But it implicitly assumes those margins stay at 37.9% forever too. For most companies, competitive pressure, market saturation, and new entrants erode excess margins over time. If Adobe's margins converge to the sector median over the next decade, the terminal value is overstated.

We now flag this explicitly. Every Damodaran DCF result includes a margin convergence signal:

• > 1.5× sector median: "Margins significantly above sector - reversion risk in terminal phase"

• > 1.2× sector median: "Margins above sector - monitor for sustainability"

• < 0.7× sector median: "Margins below sector - potential for improvement"

This isn't baked into the cash flow projection yet (that's a future upgrade). But seeing "1.72× sector median, reversion risk" next to the intrinsic value is exactly the kind of context that prevents you from blindly trusting the number.

3. Altman Z'' (For Companies That Don't Make Widgets)

The original Altman Z-Score was calibrated on 66 manufacturing firms in 1968. One of its five terms, Revenue / Total Assets, is heavily industry-dependent. A SaaS company's asset turnover looks nothing like a steel manufacturer's, and the original model wasn't designed for companies where the primary asset is intellectual property.

We switched to Altman's Z'' (Z double-prime), his 1993 revision specifically designed for non-manufacturing and service firms. It drops the industry-sensitive Revenue/Assets term and uses book equity instead of market capitalization (which avoids circular pricing dependency). The thresholds change too: Safe is now > 2.6 (vs > 2.99), Gray is 1.1–2.6, Distress is < 1.1.

Adobe's Z'' score: 7.74 - firmly in Safe territory.

4. Honest About What We Don't Know (B3 Removal)

Our Buffett Score used to award 3 out of 5 points for "business simplicity" based on a proxy: if we know the company's sector, give partial credit. The logic was that some credit was better than zero.

That logic was wrong. Awarding phantom points for data we don't have introduces systematic upward bias. Every single company got +3 free points. Graham would be appalled, the essence of value investing is honesty about uncertainty, not manufactured confidence.

We now score this criterion as 0/5 with a clear note: "Cannot assess, segment data unavailable." Adobe's Buffett Score dropped from 92 to 89. The effective maximum is now 95/100, making the scoring thresholds slightly more stringent. A company that scores 89/95 should carry more conviction than one that scored 92/100 with 3 phantom points.

5. Real Dividend History (No More Proxy)

Graham's defensive investor criterion #4 requires "uninterrupted dividends for at least the past 20 years." Our engine was checking... whether the current dividend yield was greater than zero. That's like verifying someone's employment history by checking if they're wearing a suit today.

We now pull actual dividend payment records and count consecutive years of payments. Adobe correctly fails: "No dividend history found", a company that has never paid a dividend shouldn't pass a dividend criterion on a technicality.

For companies like Johnson & Johnson, the check would show 60+ consecutive years. For a company that just started paying dividends 3 years ago, you'd see "3 consecutive years, less than 5yr required." Honest, specific, and auditable.

Why This Matters for You

If you start using Tenets to analyze stocks, here's what changes practically:

Growth stock valuations shifted. Adobe's Damodaran DCF now shows $389.32 at the base case (bear: $307.75, bull: $502.64). At $255.62, the Reverse DCF shows the market is pricing in just 1.0% Phase 1 growth: against our fundamental estimate of 8.6%. That's a low bar to beat.

The three-way IV composition tells you where the risk lives. Instead of seeing "24% cash flows, 76% terminal value" (unhelpfully vague), you now see "26% high-growth phase, 22% fade phase, 51% terminal value." If a company has 65%+ in terminal value, that's a sign the valuation is heavily dependent on long-run assumptions - tread carefully. If it's mostly in Phase 1 and Phase 2, the value is better grounded in the near-to-medium term.

Our two DCF models now agree. The Owner Earnings DCF ($508) and Damodaran DCF ($389) are within 30% of each other, down from 85%. When two independent valuation methods converge, conviction goes up. When they diverge, you know to dig deeper.

You get a margin reality check. Adobe's 37.9% operating margin is 1.72× the Technology sector median. The engine now tells you: that's a reversion risk. The terminal value assumes those margins hold forever. Would you believe that?

The Buffett Score is more honest. Adobe dropped from 92 to 89 because we stopped giving 3 phantom points for data we don't have. An 89 built on real data carries more weight than a 92 built on assumptions.

The Reverse DCF is now more precise. When we solve for "what growth rate does the market price imply?", we're solving for the Phase 1 rate in a model where Phase 2 fades to terminal. "The market is pricing in 1.0% Phase 1 growth, fading to 2.5% over 10 years" is a much more actionable statement than the old "the market implies 3% constant growth forever."

Your verdicts might shift. Since both DCF values feed into the composite verdict, some stocks near the BUY/HOLD boundary may tip over. Adobe stays at BUY, but the reasoning now cites real numbers: "score 89/100, price $255.62 below base-case DCF $389.32, IC 34×, FCF margin 41%."

THE TENET: JOIN OUR FREE WEEKLY NEWSLETTER

One stock. Three mentors. Every Sunday.

Each issue applies Graham's safety check, Buffett's business evaluation, and Damodaran's valuation model to one real company, with a clear verdict at the end.

Extra: join the newsletter to be the first to know when we launch the Tenets mobile app that runs any stock through the same Graham/ Buffett/ Damodaran workflow, in under 10 minutes, on your phone.

Sources

• Damodaran, Aswath. Terminal Value. NYU Stern lecture notes. pages.stern.nyu.edu/~adamodar/pdfiles/country/TerminalValue.pdf

• Damodaran, Aswath. Investment Valuation, 3rd ed. Ch. 12: Closure in Valuation.

• Damodaran, Aswath. fcffsimpleginzu.xlsx — FCFF two-phase DCF model spreadsheet. pages.stern.nyu.edu/~adamodar/

• Schiano, Samuel. "Terminal Value Mistakes Analysts Make — And How to Fix Them." Financial Modeling, Jan 2026. financial-modeling.com/terminal-value-mistakes

• Hangeldiyeva, Nazli. "Terminal Value — The Most Dangerous Number in a DCF." Grid Oasis, Mar 2026. gridoasis.com/guides/stock-valuation/terminal-value

• CFA Institute. "Free Cash Flow Valuation." CFA Program Curriculum, 2026. — FCFF at WACC, FCFE at cost of equity.

• Buffett, Warren. 1986 Berkshire Hathaway Annual Letter — Owner Earnings definition.

• Altman, Edward I. (1993). Z'' model revision — non-manufacturing/service firm variant. Discussed in Altman & Hotchkiss (2014), SSRN #2536340.

• Brattle Group. "Solvency Shortcuts: The Use and Misuse of Simple Tools for Predicting Financial Distress." (2022). — Z'' accuracy: 90.9% bankruptcy prediction, 97% non-bankrupt identification.

• Graham, Benjamin. The Intelligent Investor, Ch. 14 — Defensive investor criteria including "uninterrupted dividends for at least the past 20 years."

• Novy-Marx, Robert. "The Other Side of Value." (2013) — Gross profitability as a predictor of returns; margin mean-reversion.

• Mauboussin, Michael. The Base Rate Book. Credit Suisse. — Operating margin convergence toward industry medians.