06.6 · Reverse DCF & expectations investing¶

Why this matters: you will never know a company's intrinsic value to within 20%. But you can know, almost exactly, what the market is assuming — because the price is public and the model is yours. Reverse DCF turns valuation from a forecasting contest into a disagreement test: here is what ₹390 requires; do I believe it? For anyone who has ever backed an implied volatility out of an option price, this is the natural way to think about equities, and it is the single most useful technique in the module.

Learning objectives — after this lesson you can:

Explain expectations investing (Rappaport & Mauboussin) and why price should be read as a forecast.
Solve a DCF backwards for implied growth, implied margin, implied cost of capital or implied terminal growth, using fi.valuation.reverse_dcf.
Identify the "expectation that matters" and state a variant perception.
Run a reverse DCF on Kaveri (₹390 ⇒ ~14.1% ten-year growth) and interpret it.
Run a reverse DCF on a real Indian company from live data with the course tools.

Prerequisites: 06.3 DCF step by step, 06.4 DCF in practice · Time: ~90 min

1. Price as a forecast¶

A conventional DCF goes forecast → value → compare with price. Expectations investing reverses the arrow: price → implied forecast → compare with your forecast. The advantages:

It removes the anchor. You do not have to defend a point estimate of value; you have to defend a view that the market's implied assumption is too high or too low — a narrower, more testable claim.
It exposes what you are betting on. The implied growth, margin or return is the thing you must have a variant view about; if you don't, you have no edge, however good the company.
It handles high multiples honestly. Instead of "35x is too expensive", it says "35x requires 16% growth for twelve years with margins at 18%; here is why I think 11% and 15% are more likely".

Trader's lens

This is implied vol, exactly. The option price is observed; the model (Black–Scholes) is agreed; the unknown (σ) is solved for; and the trade is realised-vs-implied. In a reverse DCF the stock price is observed, the model (DCF) is yours, the unknown (growth, margin, RONIC) is solved for, and the trade is realised-vs-implied fundamentals. The same discipline applies: you are not paid for knowing the "true" vol; you are paid when realised differs from implied in the direction you bet, and you lose when it doesn't — regardless of how beautiful your model was.

2. Mechanics¶

Fix every input except one; find the value of that input that makes the DCF equal the price. fi.valuation.reverse_dcf does the bisection:

from fi.valuation import dcf_from_drivers, kaveri_base_drivers, reverse_dcf
d = kaveri_base_drivers()               # the reference base case; d["cmp"] is not stored — price is ₹390

# 1. Implied uniform revenue growth (all other drivers at base)
implied_g = reverse_dcf(390, lambda g: dcf_from_drivers(**{**d, "growth": [g] * 10})["per_share"])
# → 0.1413  (14.1% a year for FY27–FY36)

# 2. Implied flat EBITDA margin (base growth path)
implied_m = reverse_dcf(390, lambda m: dcf_from_drivers(**{**d, "ebitda_margin": [m] * 10})["per_share"], lo=0.10, hi=0.30)
# → 0.174  (17.4% every year — above the best peer's 16.8%)

# 3. Implied WACC (base operating case) — value falls as WACC rises, so flip the search
implied_w = reverse_dcf(-390, lambda w: -dcf_from_drivers(**{**d, "wacc": w})["per_share"], lo=0.07, hi=0.20)
# → 0.1107  (11.1% — about 110 bps below the reference WACC)

# 4. Implied terminal growth (base operating case)
implied_tg = reverse_dcf(390, lambda g: dcf_from_drivers(**{**d, "g": g})["per_share"], lo=0.0, hi=0.11)
# → 0.083  (8.3% for ever — above nominal GDP; implausible)

Each solve holds the others fixed, so the four answers are alternative ways the price could be right, not simultaneous requirements. Together they map the space:

For ₹390 to be fair, you need…	Implied value	Base case	Plausible?
Revenue growth (uniform, 10 years)	14.1%	10.8% CAGR	Equals the FY23–26 historical CAGR (14.7%) — but that included the solar surge; sustaining it for ten more years needs share gains or a new segment
EBITDA margin (flat, 10 years)	17.4%	13.3% → 15.5%	Above the best peer (16.8%) and Kaveri's own peak (15.4%): no
WACC	11.1%	12.19%	Only if the market's cost of equity is ~11.5% — possible in the Indian flow regime, but then every stock is "cheap"
Terminal growth	8.3%	5.5%	No — above nominal GDP for ever

The most plausible route to ₹390 is the growth one, so that is the expectation that matters: the price assumes Kaveri grows revenue ~14% a year for a decade with margins recovering to 15.5% and receivables normalising. Everything else in the analysis — the say-do record, the incremental ROIC of 4.6%, the industry map showing mid-single-digit volume growth ex-solar, the state-payment risk — bears on whether that is likely.

3. Variant perception¶

A variant perception (the term is Michael Steinhardt's) is a well-founded view that differs from the consensus embedded in the price. Reverse DCF gives it a precise form:

The market expects X. I expect Y, because of evidence Z that the market is under-weighting. If I am right, the value is V; I will know I am right (or wrong) when W happens by date D.

For Kaveri, a bearish variant perception:

The market expects ~14% revenue growth for ten years (implied by ₹390). I expect ~9–11%, because the solar segment that produced the FY23–26 surge is tender-driven, working-capital-heavy and now 27% of sales, and because management has missed its own guidance four quarters running. If I am right, value is ₹280–320. I will know by Q3 FY27 results (Feb-2027): revenue growth below 10% and receivable days above 85 would confirm; a clearance of the two state dues and a return to 15%+ growth would refute.

And a bullish one, for balance:

The market expects ~14% growth; the bull case (15–17% fading to 8%, margin to 16.5%, NWC to 19%) is worth ₹416. I believe it because [state budget allocations for PM-KUSUM have been released / the Hosur plant has won an industrial-motor contract / …]. I will know by … .

The structure forces you to name evidence, a value and a falsification date. Most "theses" fail the test at the first clause: they cannot say what the market expects.

4. Which expectation matters?¶

For a given company one or two drivers dominate the value; the reverse DCF should be run on those, not on all of them. Find them with the sensitivity decomposition from 06.3 §8: the drivers with the largest ₹/share swing per plausible unit of change.

Company type	Expectation that usually matters	Reverse-solve for
Growth company at a high multiple	Duration and rate of growth	Implied growth; implied years of growth before fade
Mature compounder	Return on new capital; fade	Implied RONIC; implied competitive-advantage period
Cyclical at a trough	Mid-cycle margin	Implied normalised margin
Turnaround / loss-maker	Terminal margin and time to reach it	Implied steady-state margin
Lender	Long-run RoE and credit cost	Implied RoE from P/B: RoE = g + P/B × (r − g)
Rate-sensitive (REIT, utility)	Cost of capital	Implied cap rate / cost of equity

The lender case is a one-liner worth memorising. From the justified P/B, $\text{ROE}_{implied} = g + (P/B)(r − g)$. Nirmal Finance at 2.5x book, $k_e$ 13%, g 8%: implied RoE = 8 + 2.5 × 5 = 20.5%, against an actual ~15%. The price assumes Nirmal becomes a materially better lender — that is the expectation to test.

5. A real-company reverse DCF with live data¶

The course tools fetch statements and price data for any NSE/BSE-listed company (disable any VPN first; Yahoo Finance blocks many VPN exits). The steps, without hard-coding today's numbers (which will be stale by the time you read this):

import io, sys
sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding="utf-8")    # ₹ on Windows
from fi.data import fetch_statements, fetch_price_info, revenue_row
from fi.valuation import dcf_from_drivers, reverse_dcf, capm, wacc

t = "ASIANPAINT.NS"                                   # any NSE ticker
stm = fetch_statements(t, period="annual", in_crore=True)
info = fetch_price_info(t)
inc, bs = stm["income"], stm["balance"]

rev = float(revenue_row(inc).iloc[0])                 # latest FY revenue, ₹ Cr
ebitda_margin = float((inc.loc["ebitda"].iloc[0]) / rev) if "ebitda" in inc.index else 0.18   # inspect inc.index for the row name
net_debt = float(bs.loc["total_debt"].iloc[0] - bs.loc["cash_and_cash_equivalents"].iloc[0]) if "total_debt" in bs.index else 0.0
shares = info["shares_outstanding"] / 1e7             # crore
price = info["price"]

ke = capm(rf=0.0707, beta=info.get("beta") or 1.0, erp=0.06)
w = wacc(ke, kd_pre_tax=0.085, tax_rate=0.2517, debt_weight=0.05)

# Base drivers: hold margin, capex, NWC at recent levels; solve for growth
base = dict(base_revenue=rev, base_nwc=0.15 * rev, growth=[0.10] * 10, ebitda_margin=ebitda_margin,
            da_pct=0.03, capex_pct=0.04, nwc_pct=0.15, tax_rate=0.2517, wacc=w, g=0.055, ronic=0.25,
            net_debt=net_debt, leases=0.0, shares=shares)
implied = reverse_dcf(price, lambda g: dcf_from_drivers(**{**base, "growth": [g] * 10})["per_share"])
print(f"{t}: price ₹{price:,.0f} implies ~{implied:.1%} revenue growth for 10 years at a {ebitda_margin:.1%} margin, WACC {w:.1%}")

Read the printed row names of inc and bs before relying on them — Yahoo's labels vary by company (ebitda may be absent; build it from operating income + D&A), and quarterly data for Indian companies is patchy (tools appendix). Then ask the only question that matters: is that implied growth plausible for this company for ten years? Compare it with the last ten years' CAGR, the industry's growth, and the company's reinvestment capacity ($g = \text{RR} \times \text{ROIC}$).

6. Limits of the method¶

The model is still yours. Implied growth depends on the margin, reinvestment and WACC you held fixed; report the assumptions alongside the implied number, and run the solve on two or three drivers.
Implied ≠ consensus. The price may embed expectations that no analyst holds — a takeover premium, forced buying by index funds, a squeeze. Reverse DCF tells you what the price requires, not what people think.
Non-operating value. For holdcos, cash-rich companies or those with large investments, strip the non-operating assets before solving, or the implied operating growth will be understated.
Distribution, not point. The implied growth is a single number; your view should be a range with a probability. The scenarios of 06.4 are the way to compare.

India notes

Indian sell-side "target prices" are usually built as forward EPS × a chosen multiple, not by DCF; their implied expectations are therefore hidden inside the multiple. When a report says "we value at 30x FY28E EPS", run the justified-P/E decomposition from 06.5 or a reverse DCF to see what growth and returns the 30x needs — the analyst rarely states them.

Common mistakes

Solving for one implied variable and calling it "the" expectation — run the map (growth, margin, WACC, g).
Forgetting that implied growth depends on the margin and reinvestment held fixed.
Comparing implied growth with last year's growth rather than a ten-year sustainable rate.
Using reverse DCF to justify the price ("well, 14% is only what they did last three years").
No falsification date: a variant perception without a "by when" is an opinion, not a thesis.

Key terms¶

Term	Meaning
Reverse DCF	Solving a DCF backwards for the input value that makes model value equal the market price
Expectations investing	Reading price as a forecast and betting on revisions to that forecast
Implied growth / margin / WACC	The value of that input required by the price, others held fixed
Expectation that matters	The driver whose implied value the price most depends on and about which a view can be formed
Variant perception	A well-founded view that differs from the expectation embedded in the price
Falsification test	The observable outcome and date that would prove the variant perception wrong
Implied RoE (lenders)	$g + (P/B)(r − g)$: the RoE a P/B multiple requires
Consensus	The average of published analyst forecasts; not the same as price-implied expectations

Check your understanding¶

Kaveri's implied growth at ₹390 is 14.1%. At what price would the implied growth equal the base case's 10.8% CAGR?

Answer

At ₹320 the *uniform* implied growth is ~11.1% — slightly above the base path's 10.8% CAGR because a flat 11.1% path front-loads less growth than the base path's 14% years and so needs a marginally higher average to reach the same value. Any price above ₹320 implies more growth than the base case (with base margins etc.).

Compute the implied RoE for a bank at 3.2x book with $k_e$ 12.5% and g 7%. What would you check next?

Answer

RoE = 7 + 3.2 × (12.5 − 7) = 7 + 17.6 = 24.6%. Check the bank's actual and historical RoE; if it is 17%, the price assumes a large, sustained improvement — ask where it comes from (NIM, cost ratio, credit cost, leverage) and whether that is plausible under RBI capital rules.

Why does the implied WACC of 11.1% not by itself mean Kaveri is fairly priced?

Answer

Because it assumes the base operating case is certain. An 11.1% cost of capital is a statement about the whole market's risk pricing, not about Kaveri; if you accept it, you must apply it to every stock, which makes the peers look cheaper still. The implied WACC tells you how much of the price is a discount-rate regime rather than a company view.

Write a variant perception for Nirmal Finance at ₹402 (2.6x FY26 book) using the implied-RoE formula.

Answer

"At 2.6x book, $k_e$ 13%, g 8%, the price implies a sustained RoE of 8 + 2.6 × 5 = 21% versus the current ~15%. I expect RoE to stay 14–16% because credit costs are rising (1.3% → 1.9% over three years) and cost-to-income has little further to fall; if I am right the justified P/B is ~1.4–1.6x, i.e. ₹215–245. I will know by FY27 results: credit cost below 1.5% and RoE above 17% would refute me."

An investor says "the market expects 20% growth, I expect 25%, so the stock is cheap". What is missing?

Answer

Evidence for 25% that the market lacks; a value at 25% growth (is the gap worth much after discounting?); the *other* implied drivers (maybe the price also assumes margin expansion the investor hasn't checked); and a falsification test. Also the base rate: how many companies sustain 25% growth for a decade?

Go deeper¶

Alfred Rappaport & Michael Mauboussin, Expectations Investing (revised edition, 2021) — the method, with the "price-implied expectations" framework and turbo charts.
Michael Mauboussin, "What Does a Price-Earnings Multiple Mean?" (Credit Suisse, 2014) — connects multiples to implied expectations.
tools/examples/kaveri_dcf_and_reverse_dcf.py — the code behind this lesson's Kaveri numbers.
11.3 Writing an investment memo — where the variant perception goes.

← Previous: 06.5 Relative valuation & multiples · Module index · Next: 06.7 Other valuation methods →