Guide

Terminal value in DCF valuation explained

Harbor Analytics pitched a buy on a vertical SaaS name at $142 per share. The ten-year DCF looked rigorous: revenue grew 18% in year one, margins expanded each year, and unlevered free cash flow turned positive in year four. But 72% of enterprise value sat in terminal value — driven by a 3.5% perpetual growth rate on a business already at 34% EBITDA margins in a category growing 9%. When the associate dropped perpetual g to 2.5% and cross-checked with a 14× exit multiple (peer median), intrinsic value fell from $142 to $89. The stock traded at $118.

Terminal value (TV) captures every cash flow after your explicit forecast horizon. In most ten-year models it is 60–80% of enterprise value. That concentration makes TV the lever that wins or loses DCF debates — not whether year-three capex was modeled to the nearest million. This guide covers Gordon growth perpetuity, exit-multiple cross-checks, H-model fade periods, TV sensitivity and attribution, the Harbor Software refactor, a technique decision table, pitfalls, and a production checklist for analysts and investors.

What terminal value represents

A DCF splits time into two buckets. Years 1 through n use explicit forecasts for revenue, margins, working capital, and capex. Year n+1 onward collapses into a single lump sum: terminal value. You then discount TV back: PV(TV) = TV / (1 + WACC)ⁿ, using the same WACC as the explicit period.

TV answers: “What is this business worth as a going concern after the forecast window, assuming it does not liquidate?” The answer must be internally consistent. If explicit years show margin compression and slowing growth, TV cannot assume perpetual re-acceleration without a documented reason (new product cycle, regulatory moat, etc.).

TV as a percentage of enterprise value

Report TV% = PV(TV) / EV in every model. Bands to watch:

Below 50% — explicit period carries most value; common for high-growth early-stage models with long horizons or heavy near-term FCF
60–80% — typical for standard ten-year mature-company DCFs; sensitivity tables are mandatory
Above 85% — red flag: you are mostly valuing an assumption about infinity, not the next decade

When TV% exceeds 75%, run a second model with a five-year explicit period extended to fifteen years. If EV barely moves, your TV assumptions are doing the work; tighten g or justify with exit-multiple evidence.

Gordon growth (perpetuity) method

The Gordon growth model values a perpetual stream growing at constant rate g:

TV = FCF_n × (1 + g) / (WACC − g)

FCF_n is unlevered free cash flow in the final explicit year. The (1 + g) term projects one more year of growth before perpetuity begins.

Constraints on g

g < WACC — required for a finite TV; if violated the formula blows up or goes negative
g ≤ long-run GDP + inflation — roughly 2–3% nominal for developed markets unless the company is tiny in a huge addressable market
g ≤ industry growth — a 40% share leader cannot outgrow its category forever
Margin stability — perpetual g implies stable reinvestment rate; margin expansion in TV double-counts improvement already in explicit years

A worked micro-example: final-year FCF = $100M, WACC = 9%, g = 2.5%. TV = 100 × 1.025 / (0.09 − 0.025) = $1.58B. Change g to 3.0% and TV rises to $1.72B (+9%). Change WACC to 8.5% at g = 2.5% and TV hits $1.70B. Small input moves, large output swings — that is the point of sensitivity analysis.

Exit multiple method

Instead of a perpetuity, apply a trading or transaction multiple to a terminal-year metric:

TV = Metric_n × Multiple

Common choices: EV/EBITDA, EV/EBIT, EV/FCF. Pick the metric least distorted by one-time items in year n. Source multiples from trading comps or precedent transactions, then haircut for maturity (terminal-year growth should be lower than today’s growth, so multiples often deserve a discount).

Reconciling Gordon and exit multiple

Implied perpetuity growth from an exit multiple: g_implied = (Multiple × WACC − 1) / (Multiple + 1) (approximate shortcut; exact algebra ties Gordon FCF to EBITDA). If Gordon g = 3.5% but exit multiple implies 2.0%, your model is internally inconsistent. Analysts should converge the two within 50 bps or document why (e.g., comps reflect acquisition premium not in perpetuity).

H-model and fade periods

High-growth companies rarely jump from 15% explicit growth to 2.5% perpetuity in one step. The H-model (two-stage fade) linearly declines growth from g_high to g_terminal over 2H years, then perpetuity at g_terminal.

Alternatively, extend the explicit forecast with a fade block: years 11–15 at declining growth rates before applying Gordon on year-15 FCF. This lowers TV% and forces you to show how competitive advantage erodes. For Harbor Software, adding a five-year fade from 8% to 2.5% cut TV% from 72% to 61% and intrinsic value from $142 to $104 before the exit-multiple cross-check.

Sensitivity and attribution

Publish a two-way table: rows = WACC (±100 bps around base), columns = terminal g (±50 bps). Each cell is equity value per share. Investors should see a range, not a single point.

TV attribution bridge

Decompose changes in intrinsic value into explicit-period FCF revisions vs TV revisions. When a model update moves price target by $20, ask: was it year-six revenue or perpetual g? Teams that cannot answer that question are optimizing the wrong cell.

Tornado chart — rank inputs by impact on EV at ±10% swings; WACC and g usually top the list
Scenario sets — bear (high WACC, low g), base, bull; weight by probability if presenting to a committee
Margin of safety — require purchase price below bear-case or below base minus one standard deviation of historical modeling error

Harbor Software refactor

Harbor’s initial SaaS DCF used a ten-year explicit forecast with terminal Gordon g = 3.5% on $118M year-ten FCF, WACC 9.0%. TV = $2.05B; PV(TV) at year ten = $865M vs $335M from explicit FCFs — TV% = 72%. The stock looked 17% undervalued at $118 vs $142 target.

Review flagged three issues: (1) perpetual g above nominal GDP while the company was already a category leader; (2) year-ten EBITDA margin of 38% with no competitive mean-reversion in TV; (3) no exit-multiple cross-check. Fixes: reduce g to 2.5%, add a five-year fade block in the spreadsheet appendix, and apply 13× EV/FCF (peer median 14×, haircut for maturity). Revised EV implied $108 per share; at $118 the name moved from “buy” to “hold.” Six months later the stock traded at $94 as growth decelerated — the TV sensitivity had been the entire thesis.

Technique decision table

Approach	Best for	Weak when
Gordon growth perpetuity	Mature, stable FCF; quick sensitivity; academic consistency	High-growth fade; g near WACC; margins still shifting
Exit multiple TV	Cross-check against market; M&A-style valuation	Comps bubble or trough; no true peers; metric distortion in year n
H-model / fade block	Mid-growth companies; lowering TV% concentration	Extra assumptions; still ends in Gordon or multiple
Extend explicit forecast (15–20 yr)	Visible long pipelines (infrastructure, pharma)	Forecast error compounds; false precision
Skip DCF; use comps only	Money-losing firms; hyper-growth with no FCF	Cyclical trough/peak; missing intrinsic floor

Common pitfalls

g ≥ WACC — produces infinite or nonsensical TV; spreadsheet may not error, but economics break.
Double-counting growth — aggressive explicit years plus high perpetual g.
Ignoring TV% — debating year-four capex while 75% of value sits in perpetuity.
Inconsistent multiples — Gordon implies 18× FCF while comps trade at 11× with no narrative.
Using levered metrics in unlevered DCF — apply EV/EBITDA or unlevered FCF multiples, not P/E, when discounting at WACC.
One-way sensitivity only — WACC alone hides interaction with g; use two-way tables.
False precision — reporting $101.47 fair value when bear-to-bull range is $78–$128.

Production checklist

Report TV% of enterprise value on every DCF output.
Enforce g < WACC and g ≤ nominal GDP unless documented exception.
Cross-check Gordon TV with exit multiple on year-n metrics.
Reconcile implied g from multiple vs stated perpetuity.
Use fade block or H-model when explicit growth > 2× terminal g.
Publish two-way WACC × g sensitivity table.
Run TV attribution on material estimate changes.
Extend horizon if TV% > 85% and value barely shifts.
Haircut comps multiples for terminal maturity.
Present intrinsic value as a range with margin-of-safety threshold.
Triangulate with trading comps and precedent transactions.
Document competitive mean-reversion assumptions in TV.

Key takeaways

Terminal value is most of a standard DCF — treat it as the main risk, not an afterthought.
Gordon and exit multiples must agree or you must explain the gap.
Harbor cut target $142 → $108 by fixing 3.5% perpetuity on a mature SaaS name.
Fade periods lower TV% and force explicit competitive logic.
Sensitivity tables are not optional when TV% exceeds 60%.