06.3 · DCF step by step¶

Why this matters: a discounted-cash-flow model is the only valuation method that forces you to write down every assumption — growth, margins, reinvestment, risk, duration — and see what they add up to. Its output is less important than its discipline. This lesson builds the Kaveri Pumps reference DCF line by line so that you can reproduce ₹320 per share, understand where each rupee of that value comes from, and then build your own for any company.

Learning objectives — after this lesson you can:

Choose between FCFF and FCFE and explain what each discounts.
Build a driver-based forecast (revenue, margins, D&A, capex, working capital, tax) with an explicit period and a fade.
Compute terminal value correctly with the value-driver form (reinvestment = g/RONIC), and explain why an exit multiple smuggles in relative valuation.
Apply mid-year discounting and build the equity bridge (net debt, leases, minorities, investments, dilution).
Reproduce the Kaveri reference valuation with tools/fi/valuation.py and decompose its value by source.

Prerequisites: 06.1 What value is, 06.2 Cost of capital, 04.4 Working capital & cash conversion · Time: ~120 min

1. The architecture¶

flowchart LR
  A[Historicals<br/>5–10 years] --> B[Drivers<br/>growth, margins,<br/>capex, NWC, tax]
  B --> C[Explicit forecast<br/>FCFF years 1–10]
  C --> D[Terminal value<br/>value-driver formula]
  C --> E[Discount at WACC<br/>mid-year]
  D --> E
  E --> F[Enterprise value]
  F --> G[Equity bridge<br/>− net debt − leases<br/>− NCI + investments]
  G --> H[Value per share<br/>÷ diluted shares]
  H --> I[Compare with price;<br/>sensitivities; scenarios;<br/>reverse DCF]

Six steps. Every one has a conventional choice and a trap; the sections below take them in order using Kaveri.

2. Step 1 — FCFF or FCFE?¶

	FCFF (free cash flow to the firm)	FCFE (free cash flow to equity)
Cash flow	NOPAT + D&A − capex − ΔNWC	PAT + D&A − capex − ΔNWC + net borrowing
Discount rate	WACC	Cost of equity
Result	Enterprise value → subtract net debt to get equity	Equity value directly
Use when	Capital structure is stable or changing predictably; most industrials	Leverage is the business (banks — but use dividends/residual income there), or the debt schedule is the point (LBO-style)
Trap	Forgetting the bridge items	Net borrowing assumptions drive the answer and are easy to game

The course uses FCFF for non-financials and equity-side methods for lenders (07.1). The two must give the same equity value if the assumptions are consistent; when they differ, the debt assumptions are inconsistent.

\[\text{FCFF}_t = \underbrace{\text{EBIT}_t(1 − t)}_{\text{NOPAT}} + \text{D\&A}_t − \text{Capex}_t − \Delta\text{NWC}_t\]

Note what is not subtracted: interest (it is in WACC), dividends (a distribution, not an operating flow), and other income (excluded from EBIT; the cash and investments that produce it are added in the bridge).

3. Step 2 — the forecast drivers¶

A forecast is a set of drivers, not a set of numbers. For Kaveri the reference case (valuation date 31-Mar-2026, base year FY26) uses:

Driver	FY27	FY28	FY29	FY30	FY31	FY32	FY33	FY34	FY35	FY36	Reasoning
Revenue growth	10%	14%	14%	13%	12%	11%	10%	9%	8%	7%	FY27 cut below the 15–18% guidance after the weak Q1; recovery as solar dues clear and Hosur ramps; fade toward nominal GDP
EBITDA margin	13.3%	14.2%	14.8%	15.2%	15.5%	15.5%	15.5%	15.5%	15.5%	15.5%	Recovery from 13.8% (FY26) toward the FY24 peak as motors utilisation rises and mix stabilises; capped at the best peer's level
D&A / revenue	3.4%	← constant									FY26 actual 3.6%; slight decline as revenue grows on a fixed asset base
Capex / revenue	3.5%	← constant									≈ D&A + modest growth capex; Hosur at 57% utilisation means no new plant this decade
NWC / revenue	25%	23%	22%	← constant							FY26 actual 27.7%; assumes receivable days fall from 96 toward ~80 as state dues normalise — the single most important operating assumption
Tax rate	25.17%	← constant									Section 115BAA

Principles behind the choices:

Anchor on history, then explain every departure. FY26 EBITDA margin was 13.8%; forecasting 15.5% requires a reason (utilisation, mix) and a ceiling (best peer 16.8%).
Fade toward the economy. No company grows 14% for ever; by year 10 growth is at nominal GDP (6–7%). The fade should be visible and gradual.
Reinvestment must be consistent with growth. Capex and NWC drivers imply a reinvestment rate; check it against the growth using \(g = \text{RR} \times \text{ROIC}\) (04.3). A model with 14% growth and 2% capex is claiming a return on new capital that should make you suspicious.
Ten years, not five, for a company whose growth is well above the terminal rate — the fade needs room, and a short explicit period pushes too much value into the terminal assumption.

4. Step 3 — the explicit forecast¶

Running the drivers through the arithmetic (₹ Cr; this is the table on the reference page, produced by tools/running_example/valuation.py):

₹ Cr	FY27	FY28	FY29	FY30	FY31	FY32	FY33	FY34	FY35	FY36
Revenue	1,449.8	1,652.8	1,884.2	2,129.1	2,384.6	2,646.9	2,911.6	3,173.6	3,427.5	3,667.4
EBITDA	192.8	234.7	278.9	323.6	369.6	410.3	451.3	491.9	531.3	568.5
D&A	49.3	56.2	64.1	72.4	81.1	90.0	99.0	107.9	116.5	124.7
EBIT	143.5	178.5	214.8	251.2	288.5	320.3	352.3	384.0	414.8	443.8
NOPAT = EBIT × 0.7483	107.4	133.6	160.7	188.0	215.9	239.7	263.6	287.4	310.3	332.1
Capex	50.7	57.8	65.9	74.5	83.5	92.6	101.9	111.1	120.0	128.4
Net working capital	362.5	380.1	414.5	468.4	524.6	582.3	640.5	698.2	754.0	806.8
Increase in NWC	(2.6)	17.7	34.4	53.9	56.2	57.7	58.2	57.6	55.9	52.8
FCFF	108.5	114.2	124.5	132.0	157.3	179.3	202.5	226.5	251.1	275.6

FY26 NWC is 188.1 + 346.7 + 39.5 − 143.4 − 65.9 = ₹365.0 Cr (27.7% of revenue). FY27's target of 25% releases ₹2.6 Cr of cash even as revenue grows — the working-capital normalisation is doing real work in the early years. Check FY27: 107.4 + 49.3 − 50.7 + 2.6 = 108.6 ≈ 108.5 (rounding). FCFF is ~₹109 Cr against FY26 reported FCF of ₹13 Cr; the whole difference is the assumption that receivables stop growing faster than sales. If that assumption is wrong, so is the valuation — which is why 06.4 stresses it.

from fi.valuation import dcf_from_drivers, kaveri_base_drivers
base = dcf_from_drivers(**kaveri_base_drivers())
print(base["projection"].round(1))          # the table above
print(round(base["per_share"]))             # 320

5. Step 4 — terminal value¶

After FY36 the model assumes steady state: growth \(g\) = 5.5% for ever (≈ 4% inflation + 1.5% real; below nominal GDP, as a perpetuity must be), and a return on new invested capital RONIC = 18% — above the 12.19% WACC, reflecting a modest brand/dealer advantage in the core business (05.3).

The reinvestment needed to grow at 5.5% with an 18% return is \(g/\text{RONIC}\) = 30.6% of NOPAT:

\[\text{FCFF}_{FY37} = \text{NOPAT}_{FY36}(1+g)\left(1 − \frac{g}{\text{RONIC}}\right) = 332.1 \times 1.055 \times 0.694 = 243.3\]

\[\text{TV}_{FY36} = \frac{\text{FCFF}_{FY37}}{\text{WACC} − g} = \frac{243.3}{0.1219 − 0.055} = 3{,}638.7\]

from fi.valuation import terminal_value
tv = terminal_value(nopat_next=332.1 * 1.055, r=0.1219, g=0.055, ronic=0.18)   # 3638.7

Two ways to get this wrong:

Growing FCFF instead of NOPAT. Taking FY36 FCFF (275.6) × 1.055 / (WACC − g) = ₹4,347 Cr assumes the FY36 reinvestment rate (17%) persists — but a 17% reinvestment at 18% return only supports ~3% growth, not 5.5%. The value-driver form keeps growth and reinvestment consistent; the naive form gives a free lunch worth ~₹37/share here (the extra ₹708 Cr of terminal value, discounted ten years).
Exit multiples. "Apply 12x EBITDA to FY36 EBITDA" = 568.5 × 12 = ₹6,822 Cr — nearly double. An exit multiple imports today's market pricing (a relative valuation) into a supposedly intrinsic model, and typically assumes the company is still a growth stock in year 10. Use it only as a cross-check: the reference TV implies an exit multiple of 3,638.7 / 568.5 = 6.4x EBITDA — which is what a 5.5%-growth, 15.5%-margin industrial should trade at.

6. Step 5 — discounting¶

The valuation date is 31-Mar-2026. Cash flows arrive through each year, not at its end, so the mid-year convention discounts year \(t\) at \((1+\text{WACC})^{t−0.5}\):

	FY27	FY28	FY29	FY30	FY31	FY32	FY33	FY34	FY35	FY36
Discount factor	0.9441	0.8416	0.7502	0.6687	0.5960	0.5313	0.4736	0.4221	0.3763	0.3354
PV of FCFF	102.4	96.1	93.4	88.3	93.8	95.3	95.9	95.6	94.5	92.4

Sum of PV of explicit FCFF = ₹947.7 Cr. The terminal value sits at end-FY36 and is discounted a full ten years: 3,638.7 / 1.1219¹⁰ = ₹1,152.3 Cr.

from fi.valuation import dcf_fcff
fcffs = [108.5, 114.2, 124.5, 132.0, 157.3, 179.3, 202.5, 226.5, 251.1, 275.6]
d = dcf_fcff(fcffs, r=0.1219, tv=3638.7, mid_year=True)
print(round(d["pv_explicit"], 1), round(d["pv_tv"], 1), round(d["ev"], 1), round(d["tv_share"], 2))
# 947.7  1152.3  2100.0  0.55

Enterprise value = 947.7 + 1,152.3 = ₹2,100.0 Cr, of which 55% is terminal value. For a company growing 10–14% for a decade, 50–65% is normal; above 75% means the explicit forecast is doing little and the valuation is really a bet on the terminal assumptions.

Enterprise value is the value of the operations. Shareholders get what is left after every other claim, plus any assets the operations do not use:

Item	₹ Cr	Source and treatment
Enterprise value	2,100.0
− Borrowings (non-current + current)	(188.0)	Balance sheet, 31-Mar-2026
+ Cash and current investments	47.5	Excluded from invested capital, so added back here; "net debt" = 188.0 − 47.5 = 140.5
− Lease liabilities	(18.2)	Debt-like; EBIT was struck after ROU amortisation but before lease interest, so leases must come out here
− Non-controlling interests	0	None at Kaveri; if present, subtract at market or book value
+ Non-operating investments (associates, land, stakes)	0	Kaveri's ₹18 Cr land sale is gone; any surplus land or associate stake would be added at value
− Probability-weighted contingent liabilities	0 in the reference	An analyst might subtract ~50% of the ₹38 Cr GST demand (₹19 Cr, ≈ ₹3/share)
− Provisions/unfunded pension	0	If material, treat as debt
Equity value	1,941.3
÷ Diluted shares (Cr)	6.07	6.00 basic + ESOPs (treasury-stock method)
Value per share	₹320	vs ₹390 market price (18-Sep-2026): −18%

from fi.valuation import equity_bridge
eq = equity_bridge(ev=2100.0, net_debt=140.5, leases=18.2)      # 1941.3
print(round(eq / 6.07))                                           # 320

Timing note: the valuation date is 31-Mar-2026 but the price is from September. Rolling the value forward six months at the cost of equity gives ≈ ₹338 — still 13% below the price. The reference page ignores the roll-forward for simplicity; mention it when the gap between valuation date and price date is material.

8. Where the value comes from — a decomposition¶

A DCF is more useful taken apart than whole:

Source	₹ Cr	Share of EV	Per share
PV of FCFF, FY27–FY31 (the "visible" years)	474.0	23%	₹78
PV of FCFF, FY32–FY36	473.7	23%	₹78
PV of terminal value	1,152.3	55%	₹190
EV	2,100.0
Less net debt and leases	(158.7)		(₹26)
Equity	1,941.3		₹320

And by assumption (each one changed alone, everything else at base):

If instead…	Value / share	Δ
NWC stays at 27.7% of revenue for ever (receivables never normalise)	≈ ₹297	−₹23
EBITDA margin plateaus at 14.0% not 15.5%	≈ ₹277	−₹43
RONIC 12.19% (= WACC; no moat in the terminal)	≈ ₹280	−₹40
WACC 12.74% (today's 7.07% G-sec)	≈ ₹293	−₹27
Growth path 3 pp higher every year	≈ ₹389	+₹69

d = kaveri_base_drivers()
print(round(dcf_from_drivers(**{**d, "nwc_pct": [0.277] * 10})["per_share"]))
print(round(dcf_from_drivers(**{**d, "ebitda_margin": [0.133, 0.138, 0.14] + [0.14] * 7})["per_share"]))
print(round(dcf_from_drivers(**{**d, "ronic": d["wacc"]})["per_share"]))
print(round(dcf_from_drivers(**{**d, "wacc": 0.1274})["per_share"]))
print(round(dcf_from_drivers(**{**d, "growth": [g + 0.03 for g in d["growth"]]})["per_share"]))

Margin and the moat assumption (RONIC) are the largest levers, then the cost of capital, then working capital — roughly the order in which Modules 04–05 told you to spend research time. A DCF that ends with one number has told you less than one that ends with this table.

Trader's lens

A DCF is a pricing model, and like any pricing model its value is in the Greeks, not the price. The decomposition above is the equivalent of vega, delta and theta: how much value moves per unit of each input. Trade the inputs you have a view on (receivables normalisation, margin recovery); hedge or avoid the ones you don't (WACC drift). And remember the model's convexity: value is convex in growth and in (WACC − g), so symmetric input errors produce asymmetric valuation errors — sensitivity tables, not point estimates.

India notes

Indian annual reports give you the March balance sheet; the valuation date is normally 31-March. Roll forward to the current date at the cost of equity, or restate net debt from the latest quarterly balance sheet (only half-yearly balance sheets are mandatory under LODR — verify what the company publishes).
Lease liabilities under Ind AS 116 are in the bridge; do not also add lease payments back to FCFF.
Dilution: use the ESOP note (SBEB disclosures) for the treasury-stock count; for companies with promoter warrants outstanding, add those shares and the exercise cash.
Contingent liabilities (tax disputes) are unusually large for many Indian companies; decide explicitly whether to probability-weight them into the bridge.

Common mistakes

g ≥ WACC in the terminal (infinite value), or g above long-run nominal GDP.
Terminal FCFF grown from the last explicit FCFF without resetting reinvestment to g/RONIC.
Exit multiples as the primary terminal method.
Forgetting mid-year discounting (understates value ~6%), or applying it to the terminal value twice.
Subtracting gross debt but forgetting cash; or netting off cash that is not really available.
Ignoring leases, minorities, or dilution in the bridge.
A five-year explicit period for a company growing 14% — the fade never happens and the terminal carries 80%.
Presenting a point estimate without the decomposition and sensitivities.

Key terms¶

Term	Meaning
DCF	Discounted cash flow: value = PV of forecast cash flows plus PV of terminal value
FCFF / FCFE	Free cash flow to the firm (all capital providers) / to equity holders
Explicit forecast period	The years modelled driver by driver before steady state
Fade	The gradual decline of growth and returns toward long-run levels across the explicit period
Terminal value	Value at the end of the explicit period of all subsequent cash flows
Value-driver terminal	\(\text{NOPAT}_{n+1}(1 − g/\text{RONIC})/(\text{WACC} − g)\)
Exit multiple	Terminal value set as a multiple of year-n EBITDA/EBIT — a relative-valuation shortcut
Mid-year convention	Discounting each year's cash flow by \(t − 0.5\) years to reflect intra-year receipt
Enterprise value (DCF)	PV of FCFF + PV of terminal value; the value of operations
Equity bridge	EV − net debt − leases − NCI − other claims + non-operating assets
Treasury-stock method	Diluted share count from options: shares issued minus shares repurchasable with the exercise proceeds
Roll-forward	Adjusting a valuation from its date to today at the cost of equity
Value decomposition	Attribution of value to periods and to individual assumptions

Check your understanding¶

Reproduce the FY30 FCFF from the drivers (revenue ₹2,129.1 Cr).

Answer

EBITDA = 2,129.1 × 15.2% = 323.6; D&A = 3.4% × 2,129.1 = 72.4; EBIT = 251.2; NOPAT = 251.2 × 0.7483 = 188.0; capex = 3.5% × 2,129.1 = 74.5; NWC = 22% × 2,129.1 = 468.4, up 53.9 from FY29's 414.5; FCFF = 188.0 + 72.4 − 74.5 − 53.9 = 132.0 ✓.

Recompute the terminal value with g = 5.5% but RONIC = 12.19% (no excess return). Why does the per-share value fall to ~₹280?

Answer

Reinvestment = 5.5/12.19 = 45.1% of NOPAT; FCFF₃₇ = 350.4 × 0.549 = 192.3; TV = 192.3 / 0.0669 = 2,874 (vs 3,639); the PV of the terminal falls by ~₹242 Cr → ≈ ₹40/share less, i.e. ≈ ₹280. With RONIC = WACC, growth adds nothing — the terminal is worth 1/WACC × NOPAT.

What is the implied exit EV/EBITDA multiple of the reference terminal value, and is it reasonable?

Answer

3,638.7 / 568.5 = 6.4x FY36 EBITDA. For a mature industrial growing at 5.5% with 15.5% margins and RONIC 18%, 6–8x is reasonable; today's 22x is a growth multiple that cannot persist into a perpetuity.

Kaveri's FY26 FCF was ₹13 Cr; the model's FY27 FCFF is ₹109 Cr. Name the two assumptions that account for the jump and how you would test them.

Answer

(1) NWC falling from 27.7% to 25% of revenue (releases ~₹2.6 Cr instead of absorbing ~₹90 Cr): test via receivable days each quarter and the >6-month ageing bucket. (2) Capex at 3.5% of revenue (₹51 Cr) vs FY26's ₹52 Cr with revenue 10% higher — modest. Almost all of the jump is working capital; if FY27 receivable days stay at 96, FCFF is closer to ₹40–50 Cr.

Why does the bridge subtract lease liabilities even though the model already charged ROU amortisation?

Answer

Because EBIT (and hence FCFF) is before lease *interest* — leases are treated as debt financing, consistent with Ind AS 116. The amortisation is the operating cost of using the asset; the liability is the financing. Subtracting it is the same logic as subtracting borrowings.

Explain in one sentence why the reference DCF says ₹320 while the price is ₹390, without using the word "wrong".

Answer

The price requires either faster growth (~14% for ten years), a higher terminal return, a lower cost of capital or a faster working-capital recovery than the base case assumes — and the next lesson's sensitivity and the reverse DCF in [06.6](06-reverse-dcf-and-expectations.md) say exactly which.

Go deeper¶

McKinsey, Valuation, Part Two (Core Valuation Techniques) — the reference for every step here.
Aswath Damodaran, The Dark Side of Valuation — DCFs for difficult companies (young, cyclical, financial).
tools/running_example/valuation.py and tools/examples/kaveri_dcf_and_reverse_dcf.py — read the code; every number on this page comes from it.
Module 10 — building the same forecast as a full three-statement model.

← Previous: 06.2 Cost of capital · Module index · Next: 06.4 DCF in practice →

06.3 · DCF step by step¶

1. The architecture¶

2. Step 1 — FCFF or FCFE?¶

3. Step 2 — the forecast drivers¶

4. Step 3 — the explicit forecast¶

5. Step 4 — terminal value¶

6. Step 5 — discounting¶

7. Step 6 — the equity bridge and per-share value¶

8. Where the value comes from — a decomposition¶

Key terms¶

Check your understanding¶

Go deeper¶