Long-Runway DCF (the finance-desk "valuation analyst")

The diagnosed hole: the v2 IV step anchored to a ~5-year, management-guidance window with a 4.5% terminal — which undervalues genuine 10–20yr compounders (it demoted IGI iv-high 511→370, CAMS to negative MoS). This skill corrects that without becoming a license to overpay — the margin-of-safety + IRR-beats-Nifty gates still bind. The goal is to stop falsely demoting a wonderful compounder, not to justify a heroic price.

STEP -1 — The false-precision gate (binding; run BEFORE any staging)

The binding CIO can veto the extended-window model itself. Before modelling, ask: is the value obvious without a 15-year projection? If the buy case only survives at the far end of a 10–15yr forecast, the margin of safety is a modelling artifact, not a fact (avoid-false-precision, invest-only-when-value-is-obvious, beware-of-believing-your-own-projections; desk: [[extended-window-must-pass-the-false-precision-test]]). A long runway is a reason not to truncate a value that is already compelling on conservative numbers — it is not a tool to manufacture upside that isn't there on a 5–7yr view. If the case fails this gate, STOP: this is a WATCH, and no amount of multi-stage machinery rescues it. Calibrate this as a veto on heroic far-window cases only — not a rejection of disciplined multi-stage DCF for a genuinely obvious compounder.

STEP 0 — Pull the canon (binding-safe)

Damodaran's DCF/terminal-value method-atomics are in the canon layer:

set -a && source /Users/pw/invest/.env && set +a && /Users/pw/invest/.venv/bin/python \
  /Users/pw/invest/data/scripts/32_consult_brain.py \
  --company "<name>" --model general --step intrinsic-value --corpus canon \
  --json-out extracted/grilling/<TICKER>_dcf.json

If that returns 0 / thin `principles` (the valuation method-atomics aren't model-scoped, so a bare company name can miss them), re-run with method terms appended to --company, e.g. --company "<name> DCF terminal value reinvestment fade growth ROIC" — that reliably surfaces the two/three-stage DCF, g=b×ROIC, terminal-value, and reverse-DCF atomics. Cite only returned slugs.

Also run the binding consult (Munger/Buffett CIO) for the moat-durability judgment that justifies the runway. The canon is the method; the CIO is the judge of whether the runway is real.

Pull the moat-erosion atoms to CAP the explicit window — runway length is an output of the moat verdict, not a free parameter ([[runway-length-equals-moat-durability]]). Add a moat consult:

... 32_consult_brain.py --company "<name>" --model general --step moat --corpus blended --role moat ...

moats-are-hard-to-maintain and fast-moats-can-be-lost-fast are the binding governors: a Wide+Widening moat earns a long explicit window; a contestable/eroding one caps it short. The fade (METHOD §3) then competes the excess returns away regardless.

WHEN this skill applies (the gate)

ONLY for a proven compounder: durable moat (binding CIO-confirmed), high & sustained ROIC (≫ cost of capital), a real reinvestment runway, aligned long-horizon ownership. A 5-yr window is correct for an ordinary business — do NOT extend the window for a contestable or commodity name. Extending the runway is earned by moat durability, not assumed.

Bank / financial hand-off (STOP-gate). If the subject is a bank, NBFC, SFB, or insurer, this skill does not apply — g = b × ROIC with FCFF is invalid when debt is raw material, not financing (debt-as-raw-material-diagnostic). Route to `bank-valuation`, which values equity directly via the excess-return / justified-P/B model (excess-return-model-for-equity-valuation). Do not free-hand a firm-DCF on a financial.

METHOD — multi-stage DCF

1. Fix the cash-flow object FIRST: it is OWNER EARNINGS — FCF to equity net of true maintenance capex,

not reported EPS or unadjusted FCF (intrinsic-value-is-discounted-future-cash, focus-on-owners-earnings). If you instead model FCFF, you MUST discount at WACC and never at Ke; equity cash flows discount at Ke (match-cash-flows-and-discount-rates). A mismatched numerator/denominator silently corrupts the IV.

1. Stage the model to the RUNWAY, not to guidance. Explicit high-growth period = the justified

runway (often 10–15yr for a proven compounder), then a fade/transition stage, then stable. Anchoring iv to 5yr of guided growth is the error that demoted IGI/CAMS.

1. Growth = reinvestment rate × ROIC (g = b × ROIC; canon: fundamental-growth-rate-formula). This is

the load-bearing identity: the value of a compounder comes from reinvesting a high fraction at a high ROIC, not from a high headline g. A franchise reinvesting 50% at 30% ROIC compounds intrinsic value ~15%/yr — and that is what a 5-yr window throws away. Model the reinvestment explicitly, and treat the assumed reinvestment rate as a capital-allocation judgment, not a free dial ([[reinvestment-rate-is-a-capital-allocation-judgment]]): b is only credible if management actually has the runway and discipline to deploy it above the cost of capital.

1. FADE both growth AND ROIC toward the economy / cost of capital. Never extrapolate peak growth or

peak ROIC to perpetuity — excess returns get competed away. The fade is what keeps the extended window honest.

1. Terminal-value discipline. TV = CF_{n+1} / (r − g) (canon: stable-growth-terminal-value-formula);

stable g ≤ risk-free rate ≤ economy growth (terminal-growth-rate-riskfree-rate-cap); terminal ROIC fades toward cost of capital. The real TV guardrail is NOT a percentage cap. A high TV share is expected for a true compounder and is not a reliability test (terminal-value-percent-dcf-value-not-a-reliability-test; desk: [[high-terminal-value-share-is-not-a-red-flag-for-a-compounder]]). What actually disciplines the terminal value is (a) the reverse-implied-growth check (reverse-implied-growth-check, see SANITY CHECK below) and (b) stable-period reinvestment = g / ROC consistency (reinvestment-rate-terminal-value-consistency): the terminal g you assume must be funded by a terminal reinvestment rate the ROC can support. Get those two right and the TV share takes care of itself.

1. Discount rate: WACC for the firm (or Ke for equity-DCF). India: rf ~6.9% + ERP ~5.5% × β.

India cuts both ways — model it, don't assume it helps. Extending the explicit window for an India compounder is currently UNGROUNDED in the brain (a known gap — do not pretend canon backs it), AND India's country-risk premium RAISES the discount rate (country-risk-premium-adjustment, implied-equity-risk-premium-extraction). So "longer window + higher discount rate" has an ambiguous net effect on IV — model it both ways and let the more conservative outcome bind. The honest posture is "model both directions and demand a larger margin of safety," never a blanket India bonus or a blanket ban.

THE SANITY CHECK (already in the machinery — keep it)

Reverse-DCF (canon: reverse-implied-growth-check): back out the growth the current price implies; compare to demonstrated + plausibly-sustainable growth. If the price implies ≫ demonstrated (e.g. CAMS implied 36.8% vs demonstrated 16%), it's heroic → no margin of safety, regardless of how good the business is. The g-files (extracted/valuation/v2/g*.json) already do this; treat it as the overpaying-guardrail. This — together with the g/ROC reinvestment-consistency check — is the guardrail that replaces the old "TV >80% = trap" rule (which canon rejects as a reliability test).

THE BALANCE (the Munger caveat — do NOT skip)

Extending the explicit window raises IV — so it MUST be paired with:

• the conviction-scaled margin of safety (buy_below = iv_base × (1 − required_MoS)), and
• the IRR-beats-~10%-Nifty opportunity-cost gate.

A longer runway widens the IV range; it does not lower the MoS you demand. A wonderful compounder at a heroic price is still a WATCH, not a BUY. The fix is don't falsely demote — not "pay anything."

OUTPUT (feeds the v2 machinery)

A conservative IV range (iv_low / iv_base / iv_high) from the multi-stage DCF, with the explicit-window length, g-vs-ROIC reinvestment assumptions, fade path, and terminal g stated. sell = iv_high; buy_below = iv_base × (1 − required_MoS). If the reverse-DCF says heroic, say so honestly.

Hard rules

1. Pass the STEP -1 false-precision gate first: if the buy case only survives at the far end of a 10–15yr

projection, it's an artifact — WATCH, not BUY. Extend the window ONLY for a CIO-confirmed durable compounder whose moat verdict earns it ([[runway-length-equals-moat-durability]]); default to 5yr otherwise.

1. Always fade growth AND ROIC; never perpetuity-extrapolate peak metrics.
2. Stable terminal g ≤ risk-free (terminal-growth-rate-riskfree-rate-cap). **Do NOT use a fixed "TV >80% =

trap" rule — canon rejects TV-share as a reliability test (`terminal-value-percent-dcf-value-not-a-reliability-test`). Discipline the terminal value with the reverse-implied-growth check + g/ROC reinvestment consistency** (reinvestment-rate-terminal-value-consistency) instead; a high TV share is normal for a real compounder.

1. Always run the reverse-DCF heroic-check; the MoS + IRR gates still bind. For India, model the longer window

AND the higher country-risk discount rate both ways and demand a larger MoS — the net effect is not assumed positive (the extended-India-window itself is an acknowledged ungrounded gap).

1. Cite only canon slugs the consult returns; the binding CIO judges the moat that earns the runway. If the

subject is a bank/financial, STOP and use bank-valuation. Desk atomics (vault/desk/valuation/atomic/…) are referenced via [[slug]] as connective tissue — they are not authored or edited here, and perspectives may inform window length but never override the binding CIO's veto.