Valuation

1. What you'll learn and why it matters

Chapters 6 and 7 taught you to judge whether a business is good. This chapter answers the next question: is the stock worth what the market is asking for it? A wonderful company bought at a silly price can still lose you money for years. You will meet a real case below: Asian Paints in January 2022.

By the end of this chapter you will be able to:

• Explain the difference between price and value, and why valuation gives you a range, not one exact number.
• Move money through time with present value (PV), future value (FV) and net present value (NPV).
• Build a discounted cash flow (DCF) model from scratch: free cash flow to the firm (FCFF), cost of equity from CAPM with a beta you compute yourself, WACC, terminal value, enterprise value, equity value and value per share.
• Run a sensitivity table and apply a margin of safety.
• Run a reverse DCF: find the growth rate today's price is quietly assuming.
• Use relative valuation properly: P/E bands, P/S, P/B for banks, EV/EBITDA, PEG, and know when each one fits.
• Spot the classic valuation traps (garbage-in, terminal-value dominance, cyclicals at peak earnings, one-off gains, look-ahead bias, value traps), each shown with a real Indian example.
• Read the market's overall valuation (Nifty P/E, earnings yield vs the 10-year G-sec) and build a watchlist with buy-zone prices.

The main real examples in this chapter: a full DCF of Pidilite Industries; reverse DCFs of Asian Paints and ITC (plus Pidilite); a dividend discount model of ITC; P/E bands of Pidilite, Asian Paints, ITC and TCS; the cyclical P/E trap at Tata Steel; P/B for five large banks; EV/EBITDA across paints, metals and cement; and the Nifty 50's own valuation. Financials are consolidated, from screener.in pages fetched on 2 Oct 2026 (FY26 annual results plus the June-2026 quarter). Prices are NSE closes on 30 Sep 2026.

2. Price is not value

Price is what you pay: the last traded number on the screen. It changes every second and reflects the mood, flows and news of the moment. Value (also called intrinsic value or fair value) is what the business is worth to an owner, based on the cash it is expected to produce over its life. Value changes slowly, as the business itself changes.

Varsity puts it simply: an investment works best when a great business is bought at a great price, and even a mediocre business can be a good investment if the price is low enough. Valuation is the third stage of equity research. Stage one asks whether you understand and like the business, stage two checks its financial performance (chapters 6–7), and stage three asks whether the price makes sense.

Valuation methods try to estimate value so you can compare it with price:

• Price well below value: possibly undervalued. Worth researching, never an automatic buy.
• Price near value: fairly valued. You earn roughly the business's own return from here.
• Price well above value: overvalued. Future returns depend on things going better than expected.

3. Time value of money: PV, FV and NPV

₹100 today is worth more than ₹100 a year from now, for two reasons: you could invest today's ₹100 and earn a return (opportunity cost), and future money is uncertain. So money received at different times cannot be added up directly. It first has to be moved to the same point in time.

• Compounding moves money forward in time. The result is the future value (FV).
• Discounting moves money backward in time. The result is the present value (PV).

Net present value (NPV)

When an asset produces a stream of cash flows, discount each one separately and add them up. That sum is the net present value. It is the most you should rationally pay today for that stream.

4. Three ways to value a business

Varsity's financial-modelling module groups valuation methods into three families:

Relative valuation is intuitive but has two weaknesses that Varsity points out. The whole sector can be mispriced (think of the dot-com bubble), and no two companies are truly identical. A firm with a 25% profit margin deserves a higher P/S than one with 15%. Absolute valuation forces you to write down every assumption, which is its strength, but the output is only as good as those assumptions.

Enterprise value vs equity value

Rearranging the balance sheet makes the two viewpoints clear. Assets = debt + equity. Split the assets into cash and operating assets and move cash to the other side:

• Value the operating business (all the capital, from lenders and owners together) and you get enterprise value (EV). You discount free cash flow to the firm (FCFF) at the blended cost of all capital, the WACC.
• Value only the shareholders' part and you get equity value. You discount free cash flow to equity (FCFE) at the cost of equity.
• The bridge between them: Equity value = EV − debt + cash (and surplus investments). Divide by shares outstanding to get value per share.

5. Free cash flow: FCF, FCFF and FCFE

A business is worth the cash it can hand back to its owners. Profit is an accounting opinion; cash is a fact (chapter 6). The cash flow we discount is free cash flow: what is left after the company has paid its running costs and taxes and reinvested enough in fixed assets and working capital to keep growing.

The logic behind each piece:

• EBIT × (1 − t), often called NOPAT, is the after-tax operating profit the business would earn if it had no debt. We start from EBIT, not PAT, because interest belongs to the lenders, who are part of "the firm".
• + D&A: depreciation is a non-cash charge. The cash left when the asset was bought, and that purchase is captured in capex.
• − ΔWorking capital: a growing company ties up more cash in inventory and receivables.
• − Capex: spending on plants, machinery and stores is real cash out.

The tax shield

Interest is tax-deductible, so debt reduces a company's tax bill. Varsity's example: EBIT ₹700 cr, interest ₹70 cr, tax 25%. With the interest, tax is ₹157.5 cr. Without it, tax would be ₹175 cr. The ₹17.5 cr saved is the tax shield, and it is why the after-tax cost of debt is \(K_d(1-t)\). If you start from PAT instead of EBIT, add back interest × (1 − t), not the full interest. Both routes give the same FCFF.

6. The discount rate: CAPM, beta and WACC

The discount rate is the return investors require for the risk they take. Lenders take less risk and accept a lower return. Shareholders are paid last and demand more.

6.1 Cost of equity with CAPM

Risk-free rate. Varsity uses the 10-year G-sec yield as the proxy. On 30 Sep 2026 it was 7.18% (countryeconomy.com; Trading Economics shows 7.16%). It rose from 6.96% at the start of September 2026. We use \(R_f = 7.18\%\).

Equity risk premium. This is the extra return investors want for owning stocks instead of government bonds. Professor Aswath Damodaran's January 2026 update puts India's total ERP at 7.08%, and Indian valuation firm Incwert recommends 7.0% (range 6.5–7.5%) from January 2026. We use ERP = 7.0%. Varsity's older chapters instead assume a market return \(R_m\) of 8.5–12%. Either way, write the number down and justify it.

Beta. Beta measures how much a stock tends to move when the market moves. A beta of 1.2 means that, on average, a 1% Nifty move came with a 1.2% move in the stock. Compute it from your own data:

Weekly returns to 30 Sep 2026 vs the Nifty 50. Beta changes with the window: ITC's ranges from 0.45 to 0.65. Use a few windows and pick a sensible middle.

Two refinements you will meet in research reports. (1) Adjusted beta: raw betas tend to drift towards 1 over time, so many data providers report \(\beta_{adj} = 0.67\,\beta_{raw} + 0.33\). For Pidilite: \(0.67 \times 0.84 + 0.33 = 0.89\), which lifts \(K_e\) from 13.06% to about 13.4%. (2) Levered vs unlevered beta: a stock's beta contains business risk plus the extra risk from its debt. To compare companies with different debt levels, analysts strip debt out, \(\beta_U = \beta_L / [1 + (1-t)\,D/E]\), and put the target company's own debt back in. Tata Steel (debt/equity 0.40 at market value) has an unlevered beta of about \(1.24 / (1 + 0.7483 \times 0.40) = 0.95\). Most of its above-market beta comes from leverage. We use raw betas in this chapter to keep the arithmetic visible.

6.2 Cost of debt and WACC

7. Terminal value: Gordon growth and exit multiple

We cannot forecast cash flows year by year forever. The standard approach forecasts explicitly for 5–10 years, then assumes the company settles into steady growth at rate \(g\) forever. The value of all cash flows after the forecast period, measured at the end of year \(N\), is the terminal value (TV).

Intuition: a cash flow of ₹100 next year that grows 5% a year forever, discounted at 12%, is worth \(100 \times 1.05/(0.12-0.05) = ₹1{,}500\) at the end of year N. Nudge \(g\) to 6% and it becomes ₹1,766.7 (+18%). Nudge \(r\) down to 11% instead and it becomes ₹1,750. Small changes in the gap \((r - g)\) cause large changes in value. Keep this in mind for the sensitivity section.

The exit-multiple method

The alternative assumes the business is "sold" at the end of year N at a market multiple, for example \(TV_N = EBITDA_N \times\) EV/EBITDA multiple. It is popular with practitioners because it ties to market prices. But it imports today's market mood into a "fundamental" model. Always convert an exit multiple into the perpetual growth it implies and ask whether that growth is believable:

8. A complete DCF: Pidilite Industries

Pidilite (Fevicol, Dr. Fixit, M-Seal) is a near-debt-free consumer business with high returns on capital. That suits a DCF, because its cash flows are positive and fairly predictable. Here is the full model, step by step. All figures are in ₹ crore unless marked otherwise.

Step 1 — Growth assumptions and why

Growth rates below are compound annual growth rates (CAGR): \(\text{CAGR} = (\text{End}/\text{Start})^{1/\text{years}} - 1\). For example, sales of ₹5,361 cr (FY16) to ₹14,601 cr (FY26) is \((14{,}601/5{,}361)^{1/10} - 1 = 10.5\%\) a year. Over FY16–FY26 Pidilite's sales grew 10.5% a year and its EBIT 11.3% a year. FCFF grew 15.2% a year (three-year average FY16–18 to FY24–26), partly because margins expanded. For the explicit forecast we assume 12% a year for years 1–5 (a little above sales history and below the FCF history, on the view that margin expansion cannot repeat forever), then 9% for years 6–10 as the business matures, then 5% forever. This follows Varsity's two-stage approach (Varsity used 18%/10% for Amara Raja in 2014), but with growth anchored to the company's own track record.

Step 2 — Project and discount the cash flows

Example row: year 1 FCFF = 1,943.5 × 1.12 = 2,176.7, and its PV = 2,176.7 / 1.13 = 1,926.3. (The table is computed from the unrounded base, ₹1,943.43 cr. Starting from the rounded ₹1,943.5 moves later rows by ₹0.1–0.2 cr and leaves the per-share value at ₹418.2.) Notice how the PVs fall even though cash flows rise: discounting at 13% outruns 9–12% growth.

Step 3 — Terminal value

1. \(TV_{10} = 5{,}269.8 \times 1.05 / (0.13 - 0.05) = \mathbf{69{,}167}\).
2. \(PV(TV) = 69{,}167 / 1.13^{10} = 69{,}167 \times 0.2946 = \mathbf{20{,}376}\).

Step 4 — Enterprise value → equity value → per share

1. Enterprise value = 17,815 + 20,376 = 38,191. The terminal value is 53.4% of it.
2. Add cash and equivalents (299) and investments (4,350), and subtract borrowings including leases (417). Net cash = 4,232.
3. Equity value = 38,191 + 4,232 = 42,423.
4. Per share = 42,423 / 101.45 = ₹418.2.

Step 5 — Cross-check with an exit multiple

Suppose instead that at the end of FY36 the business is valued at 20x EBITDA, half its current ~40x. Grow FY26 EBITDA (₹3,521 cr) on the same path to ₹9,547 cr in year 10. Then \(TV = 20 \times 9{,}547 = 1{,}90{,}950\), its PV is 56,252, and value per share becomes ₹771.8. Now the terminal value is 75.9% of EV. Convert it into implied perpetual growth: \(g = (190{,}950 \times 0.13 - 5{,}270)/(190{,}950 + 5{,}270) = \mathbf{9.97\%}\) a year, forever. Even a "conservative" 20x exit multiple assumes Pidilite grows at almost 10% for eternity. That is how market multiples hide aggressive assumptions.

9. Sensitivity and margin of safety

A DCF is a machine for turning assumptions into a number. To see how fragile that number is, vary the two most powerful inputs, the discount rate and terminal growth, and look at the whole grid.

Stage-1 growth matters just as much. Keeping WACC at 13% and terminal growth at 5%, with stage-2 growth 3 points below stage 1 (minimum 5%), value per share is ₹326 at 8% growth, ₹369 at 10%, ₹418 at 12%, ₹507 at 15%, ₹617 at 18% and ₹705 at 20%.

Margin of safety

Because the inputs are uncertain, insist on buying well below your estimate. That cushion is the margin of safety, a term from Benjamin Graham.

• Varsity's rule: build a ±10% band around the DCF value. Below the lower band, consider buying. Inside it, hold. Above the upper band, do not buy. For Pidilite the band is ₹376–₹460.
• A stricter rule uses 25–40% MOS. At 30%, Pidilite's buy price would be ₹293.
• Use a bigger MOS when the business is less predictable: cyclicals, high debt, short history, or governance questions.

10. Reverse DCF: what is the price assuming?

Instead of guessing growth and computing value, a reverse DCF takes the market price as given and solves for the growth rate (or the return) that makes the DCF value equal the price. Then you judge whether that growth is believable compared with the company's history and its industry.

11. Dividend discount model (ITC)

For a mature company that pays out most of its earnings, the cash shareholders actually receive is the dividend. The dividend discount model (DDM) values the share as the PV of all future dividends. With constant growth it reduces to the Gordon formula:

12. Relative valuation: P/E, P/S, P/B, EV/EBITDA, PEG

Multiples are shortcuts: price divided by some measure of what you get. They are quick and they reflect market reality, but they hide the assumptions a DCF makes explicit. Use them to compare like with like: a company with its own history, or with close peers.

12.1 Price to earnings (P/E)

P/E is how many rupees investors pay today for one rupee of annual profit. Its inverse, the earnings yield, lets you compare a stock with a bond. A P/E of 20 is an earnings yield of 5%. A trailing P/E uses the last 12 months' EPS (TTM). A forward P/E uses next year's estimate, which is more relevant but is someone's guess.

Why P/Es differ (the logic): from the Gordon model, a "fair" P/E ≈ payout × (1 + g)/(\(K_e\) − g). P/E rises with growth and with the quality (sustainability) of growth, and falls with risk and with interest rates. A high P/E is not "bad". It is a bet that growth will be high and long-lasting.

P/E band chart. Plot price alongside lines of EPS × fixed multiples. You immediately see whether the stock is trading at the top or bottom of its historical range.

Varsity's personal rule of thumb is to avoid stocks above 25–30x earnings. Today's Indian market routinely pays more for consumer franchises. Treat such rules as starting points, and always ask what growth the multiple implies (section 10).

12.2 Price to sales (P/S)

Useful when earnings are temporarily depressed or negative, for example a cyclical at the bottom or a young company. Always read it next to the profit margin. Pidilite trades at 1,48,628 / 15,399 TTM sales = 9.65x sales with a 17.4% net margin. Asian Paints trades at 2,29,440 / 37,187 = 6.17x sales with a 13.0% net margin. A rupee of Pidilite's sales carries more profit, which partly explains the higher P/S.

12.3 Price to book (P/B), the right lens for banks

A bank's assets (loans) and liabilities (deposits) are financial and carried close to their real value, so book value means something. Its "raw material" is money, which makes EV and FCF awkward. For manufacturing or consumer businesses whose value lies in brands and know-how, P/B means little.

12.4 EV/EBITDA

EV/EBITDA is capital-structure neutral. EV includes debt and EBITDA is before interest, so a heavily indebted company and a debt-free one can be compared fairly. It also ignores depreciation policies. That makes it the standard multiple for capital-heavy, leveraged sectors: steel, cement, telecom, power, infrastructure. Its blind spot is that EBITDA ignores capex, which in these sectors is huge and very real.

12.5 PEG ratio

PEG adjusts P/E for growth. A popular rule of thumb (from Peter Lynch) says PEG ≈ 1 is fair and below 1 is attractive. The catch is that the growth used must be sustainable future growth, and past growth can be badly distorted:

12.6 Which multiple when?

13. Valuation pitfalls, with real failures

13.1 Garbage in, garbage out

A DCF has no opinion of its own. It multiplies whatever you feed it. Our Pidilite value moves from ₹326 to ₹705 just by changing stage-1 growth from 8% to 20%. Defend each input with history, industry context and management commentary. Write your reasons down (Varsity suggests a "working notes" sheet), and update the model when facts change, not when the price changes.

13.2 Terminal value dominance

In our base case 53% of Pidilite's value comes from cash flows after 2036, and with an exit multiple it is 76%. If most of your answer depends on what happens more than 10 years away, be humble about precision and generous with margin of safety.

13.3 Cyclicals: the low P/E trap

For commodity businesses, earnings swing far more than the business's long-term value. At the top of the cycle, earnings are huge and P/E looks tiny. At the bottom, earnings collapse and P/E looks enormous. Reading those P/Es naively gets you exactly backwards.

The three lowest P/Es (4.0, 4.8, 5.6) were all followed by price falls. The highest (19.5 in March 2020 on collapsed earnings, and 56.3 in 2025) were followed by gains of 201% and 24%. FY22 EPS of ₹32.88 was a peak: it fell to ₹7.17 the next year and to a loss the year after. For cyclicals, value the business on normalised (mid-cycle) earnings, or use EV/EBITDA on average-cycle EBITDA and P/B, never on peak P/E. Note that these P/Es use each year's full EPS, which the market only learned after the year ended, so they flatter the "low P/E" signal. In real time it would have looked even more convincing, and been even more wrong.

13.4 One-off gains distort P/E

13.5 Negative or tiny free cash flow: DCF does not fit

DMart's three-year average CFO − capex is −₹525 cr, because it keeps buying land and building stores. Titan's FCF yield is only 0.39%, because gold inventory absorbs cash as it grows. A mechanical DCF values both near zero or below. That says nothing useful about their worth: their value is in the returns on reinvestment. For such businesses, model a future steady state (when capex slows) or use P/E and ROCE-based reasoning instead.

13.6 Look-ahead bias: the buy zone that only worked in hindsight

Here is a tempting rule: "buy when P/E is below its 10-year 25th percentile." On Asian Paints, using the full 2016–2026 history to set the threshold (59x), the median 1-year return after an in-zone day was +15.7% vs +8.6% otherwise. Great?

No. That threshold used the future. Redo it honestly: each day, use only the trailing 5 years of P/E to set the 25th-percentile threshold (testable from Oct 2018):

In real time, "below its usual P/E" was followed by worse returns for all four. De-ratings tend to persist: a P/E usually falls because the outlook changed, and it can keep falling. This does not mean P/E bands are useless. It means a low P/E is a reason to research, not a reason to buy, and that any rule you test must use only information available at the time. Chapter 12 covers backtesting discipline in depth.

13.7 Value traps and changed regimes

13.8 Other common mistakes

• Mixing adjusted and unadjusted data. Splits and bonuses (Pidilite 1:1 in 2025, HDFC Bank 1:1 in 2025) change the share count. Use EPS and prices on the same basis.
• Standalone vs consolidated. Use consolidated numbers for groups with subsidiaries. For banks with big subsidiaries (HDFC Bank, Kotak, ICICI), value the subsidiaries separately ("sum of the parts") if you want precision.
• Forgetting dilution: ESOPs, warrants and planned equity raises.
• Discounting nominal cash flows at a real rate (or the reverse). Keep everything in nominal rupees.
• Precision theatre. "Fair value ₹1,237.45" signals false confidence. Say "roughly ₹1,100–1,350".

14. Valuing the whole market: Nifty P/E and bond yields

An index has a P/E too: NSE divides the combined free-float market cap of the 50 stocks by their combined earnings and publishes the Nifty's P/E, P/B and dividend yield daily (Varsity: NSE website → indices → historical P/E, P/B & dividend yield).

Where the market stands (post-Apr-2021 consolidated series): mean 22.57, standard deviation 2.50, range 18.92–33.60. At 19.36 on 30 Sep 2026, only about 1% of weekly readings since April 2021 were lower. A year earlier the P/E was 21.86. The Nifty fell 8.1% over the year (24,611.1 → 22,620.45). The P/E implies trailing Nifty EPS of about 22,620.45 / 19.36 ≈ 1,168.

Earnings yield vs bond yield

Compare what stocks "earn" on their price with what a risk-free bond pays. The yield gap = earnings yield − G-sec yield. The bond-equity earnings yield ratio (BEER) = G-sec yield ÷ earnings yield.

Does a low market P/E predict higher returns?

Post-2021 the correlation between starting P/E and next-year return is −0.25: a modest tendency, with huge scatter. In 2016–2021 it was only −0.06, because 2020's collapsed earnings made the P/E spike just before a huge rally (the same cyclical trap as Tata Steel, at index level). With only about five years of comparable data and overlapping windows, treat this as context, not a timing tool. A low index P/E improves your odds over years. It does not tell you next month's direction, so swing traders should still trade price (chapters 3–5).

15. Building a watchlist with buy-zone prices

Valuation pays off when you prepare. Decide in advance, calmly, what you would pay for businesses you like. When the market offers that price (often in a panic), you act on a plan instead of a feeling. Varsity notes that long-term opportunities "linger": a stock can sit in a value zone for months, so there is no need to rush.

A simple two-lens process

1. Quality gate first (chapter 7): only businesses with consistent ROCE, cash conversion and clean governance get on the list.
2. Lens 1, own history: P/E buy level = 25th-percentile P/E (10 years) × TTM EPS. It tells you when the stock is cheap relative to how the market usually values it.
3. Lens 2, cash flows: reverse DCF at a fixed 12% discount rate and 5% terminal growth. Compare the implied growth with the better of the 5-year EPS CAGR and the 10-year sales CAGR. It tells you whether the price needs heroics.
4. Optional DCF value: stage-1 growth = 10-year sales CAGR (capped at 15%), stage 2 = two-thirds of that (at least 5%), terminal 5%, discount 12%. Then apply a 25% margin of safety.
5. Classify: both lenses favourable means research in depth; only one means keep watching; neither means move on. Re-run quarterly after results.

Source: author's calculations (script course/scripts/ch08_figures.py). The classification is the mechanical output of the rule above, shown to illustrate the method. It is not a recommendation.

How to read this honestly:

• All six stocks are below their own P/E buy level. That says more about the 2025–26 de-rating of quality stocks than about bargains. Section 13.6 showed that buying on this lens alone, in real time, has done badly.
• The cash-flow lens separates them. Pidilite and Asian Paints still need roughly 23–26% a year FCF growth at a 12% discount rate. ITC and TCS need less than they have historically delivered.
• "Research in depth" means read the annual report, the last few concall transcripts and the risks. For TCS, ask whether the drop reflects a temporary scare or a permanent change (13.7). For ITC, weigh regulatory and tax risk on cigarettes.
• Even then, phase your buying (for example in three tranches over a few months) and size positions with risk rules (chapter 11). Valuation tells you what is reasonable to buy. It never tells you when the price will turn.

16. Try it yourself (Python)

These snippets need a copy of the course repository and Python with pandas and numpy (setup steps are in its framework/README.md). Run them from the repository's root folder, for example with .venv/bin/python, because the file paths are relative to that folder. Each reproduces a number from this chapter using the course data. The full analysis, every figure and every number is in course/scripts/ch08_figures.py, which writes all values to data/extra/ch08_numbers.json.

1. Beta and cost of equity

import pandas as pd, numpy as np
D = "data/ohlc/"
stock = pd.read_csv(D + "PIDILITIND.csv", index_col=0, parse_dates=True)["Adj Close"]
nifty = pd.read_csv(D + "IDX_NSEI.csv", index_col=0, parse_dates=True)["Close"]
df = pd.concat([stock, nifty], axis=1, keys=["s", "m"], sort=True).loc["2023-09-30":"2026-09-30"].dropna()
w = df.resample("W-FRI").last().pct_change().dropna()        # weekly returns
beta = np.cov(w.s, w.m)[0, 1] / w.m.var()
rf, erp = 0.0718, 0.07
print(f"weeks={len(w)}  beta={beta:.2f}  cost of equity={rf + round(beta, 2) * erp:.2%}")
# -> weeks=156  beta=0.84  cost of equity=13.06%

2. A DCF with a sensitivity grid

def dcf(fcf0, growth, r, g, net_cash, shares):
    """growth = list of yearly growth rates; returns (value per share, TV share of EV)."""
    pv, fcf = 0.0, fcf0
    for t, gt in enumerate(growth, start=1):
        fcf *= 1 + gt
        pv += fcf / (1 + r) ** t
    tv = fcf * (1 + g) / (r - g)                 # Gordon growth terminal value at year N
    pv_tv = tv / (1 + r) ** len(growth)
    ev = pv + pv_tv
    return (ev + net_cash) / shares, pv_tv / ev

# Pidilite, Rs crore: 3-yr avg FCFF, 12% x5 then 9% x5, WACC 13%, g 5%
value, tv_share = dcf(1943.5, [0.12] * 5 + [0.09] * 5, 0.13, 0.05, net_cash=299 + 4350 - 417, shares=101.45)
print(f"value/share Rs {value:.0f}, terminal value = {tv_share:.0%} of EV")   # Rs 418, 53%

for r in [0.11, 0.12, 0.13, 0.14, 0.15]:          # sensitivity row by row (g = 4%, 5%, 6%)
    print(f"WACC {r:.0%}:", [round(dcf(1943.5, [0.12]*5 + [0.09]*5, r, g, 4232, 101.45)[0]) for g in (0.04, 0.05, 0.06)])

3. Reverse DCF (ITC)

from scipy.optimize import brentq

def equity_value(fcf0, g10, r, g_term, net_cash):
    pv, fcf = 0.0, fcf0
    for t in range(1, 11):
        fcf *= 1 + g10
        pv += fcf / (1 + r) ** t
    pv += fcf * (1 + g_term) / (r - g_term) / (1 + r) ** 10
    return pv + net_cash

fcf0, net_cash, mcap = 15193, 38738, 327863      # Rs crore, price Rs 262.75
for r in (0.10, 0.1075, 0.12):
    g = brentq(lambda x: equity_value(fcf0, x, r, 0.05, net_cash) - mcap, -0.3, 0.8)
    print(f"discount {r:.2%}: price implies {g:.1%} FCF growth for 10 years, then 5%")
# -> 3.7%, 5.5%, 8.3%

4. A simple P/E history (annual EPS only)

import pandas as pd
px = pd.read_csv("data/ohlc/ASIANPAINT.csv",
                 index_col=0, parse_dates=True)["Close"].loc["2016-06-01":]
eps_fy = {2016: 18.19, 2017: 20.22, 2018: 21.26, 2019: 22.48, 2020: 28.20, 2021: 32.73,
          2022: 31.59, 2023: 42.81, 2024: 56.92, 2025: 38.23, 2026: 45.09}   # screener.in, consolidated
eps = pd.Series({pd.Timestamp(f"{y}-05-31"): v for y, v in eps_fy.items()})  # known by end-May
eps = eps.reindex(px.index.union(eps.index)).ffill().reindex(px.index)
pe = px / eps
print(pe.describe().round(1))
print("25th pct P/E:", round(pe.quantile(.25), 1), "-> buy level on FY26 EPS:", round(pe.quantile(.25) * 45.09))
# median 66.8, 25th pct 58.5 -> 2,639. (The chapter uses quarterly TTM EPS from 2024, so its numbers differ slightly.)

5. Earnings yield vs G-sec

import pandas as pd
X = "data/extra/"
v = pd.read_csv(X + "ch08_nifty_pe_pb_dy_weekly_craytheon.csv", parse_dates=["Date"], index_col=0)
y = pd.read_csv(X + "ch08_india_10y_gsec_monthend_countryeconomy.csv", parse_dates=["Date"], index_col=0)
m = pd.concat([100 / v.pe.resample("ME").last(), y.Yield10y.resample("ME").last()], axis=1,
              keys=["earn_yield", "gsec"], sort=True).dropna()
m["gap"] = m.earn_yield - m.gsec
print(m.loc["2021-04":].describe().round(2))   # gap mean -2.36, range -2.93 to -1.73
print(m.tail(3).round(2))                      # Sep 2026: 5.17 vs 7.18, gap -2.01

Exercises with the code: (a) change Pidilite's WACC to 11% and stage-1 growth to 15% and see how far you get towards ₹1,465; (b) run the reverse DCF for Asian Paints (FCF 3,940, net cash 4,207, market cap 2,29,440); (c) compute beta over 5 years instead of 3 for ITC and see how \(K_e\) changes.

17. Practical checklist

• Has the business passed the quality checks (chapters 6–7) before I value it?
• Am I using consolidated numbers, adjusted for splits and bonuses, on the same basis as the price?
• Did I scan the last four quarters for one-off gains or losses before using TTM EPS?
• Base-year cash flow: did I average 3 years and check for working-capital swings and acquisitions?
• Discount rate: current 10-year G-sec yield (verified today), beta from my own data over 2–5 years, a stated ERP. Is WACC well above terminal growth?
• Growth: is each stage defended by history and industry context? Is terminal growth at most ~5–6%?
• What share of value comes from the terminal value? If over 60%, have I widened my margin of safety?
• Did I run a sensitivity grid (WACC × g, and stage-1 growth) and look at the whole range?
• Reverse DCF: what growth does today's price imply, and is it believable versus history?
• Is my multiple right for the business type (P/B for banks, EV/EBITDA for leveraged cyclicals, never peak P/E for cyclicals)?
• Am I comparing multiples within the same sector, or against the stock's own comparable history?
• Is this a de-rating that might continue (value trap)? What changed in the business?
• Index context: where is the Nifty P/E versus its post-2021 range, and the yield gap versus its range?
• Buy level and margin of safety written down before the price gets there; buying plan phased; position size set by risk rules.
• Thesis and "what would make me sell" written in two lines each; model updated after every quarterly result.

18. Key takeaways

19. Quiz and exercises

20. Sources

Zerodha Varsity, Module 3, Fundamental Analysis: Ch. 11 The Financial Ratio Analysis (Part 3): valuation ratios and index valuation · Ch. 13 Equity Research (Part 1) · Ch. 14 DCF Primer · Ch. 15 Equity Research (Part 2): FCF, terminal value, NPV, share price, intrinsic value band · Ch. 16 The Finale: DCF limitations, margin of safety, when to sell.
Zerodha Varsity, Module 13, Integrated Financial Modelling: Ch. 14 Valuation Overview · Ch. 15 FCFF and FCFE · Ch. 16 Risk Premium and Tax Shield · Ch. 17 WACC and Terminal Growth · Ch. 18 Discounted Cash Flow Analysis. All concepts are explained in the course's own words.
Rates and premiums (checked Oct 2026): 10-year G-sec yield 7.18% on 30 Sep 2026, countryeconomy.com (month-end history saved to data/extra/ch08_india_10y_gsec_monthend_countryeconomy.csv); cross-check 7.16%, Trading Economics. Equity risk premium: Damodaran, Country Default Spreads and Risk Premiums (updated 5 Jan 2026): India total ERP 7.08%; Incwert, Assessment of India ERP, Jan 2026 (7.0%). Corporate tax 25.17% (section 115BAA of the 1961 Act; section 200 of the Income-tax Act, 2025 from tax year 2026-27): caclubindia (AY 2022-23 to 2026-27), TaxTMI, section (new) 200 / (old) 115BAA, incometaxindia.gov.in.
Nifty 50 P/E, P/B and dividend yield (weekly, Oct 2016 – Oct 2026): NSE index data as compiled by craytheon.com (saved to data/extra/ch08_nifty_pe_pb_dy_weekly_craytheon.csv); official source niftyindices.com historical P/E, P/B & Div. Yield. Switch to consolidated earnings on 31 Mar 2021: Upstox Learning Center.
Company financials: screener.in consolidated pages fetched 2 Oct 2026 (data/extra/screener_ch07/). Prices: NSE daily OHLC via Yahoo Finance to 30 Sep 2026 (data/ohlc/, data/extra/ch07_extra_closes_yf.csv). Calculations and figures: course/scripts/ch08_figures.py; all numbers in data/extra/ch08_numbers.json.