Sum-of-the-parts valuation

When one ticker is several businesses

Think of Reliance (energy + telecom + retail), Grasim (VSF + chemicals + a listed cement giant + paints), or an Alphabet (search + YouTube + cloud + bets). You can't value a conglomerate with one growth rate and one multiple, because it isn't one thing. The market itself prices it in pieces, and so must you. The method:

1. Cut along economic seams. Use the segment note (Ind AS 108 / ASC 280, from your consolidation work) as the floor, with finer external data where you can get it. Each segment gets the valuation its business model deserves: EV/EBITDA off pure-play comps for the industrial segment, EV/Sales or DCF for the young retail rocket, P/B for a lending arm, and mid-cycle numbers for cyclical segments. That last one is the cyclicals toolkit running inside the SOTP.
2. Value listed stakes at market. Don't build your own SOTP of their SOTP. Apply a haircut (typically 10 to 20%) if the stake is non-control, illiquid relative to size, or carries embedded capital-gains tax on any realistic exit.
3. Subtract the lines amateurs forget:

• Capitalized central costs. Unallocated corporate overhead is a real, permanent negative cash flow, so capitalize it like a segment: annual unallocated cost (post-tax) × a 7 to 8× multiple, subtracted. Skip this line and you overstate NAV by 3 to 8% for most conglomerates.
• Net debt and its location. Parent or subsidiary? Consolidated net debt can double-count what a listed subsidiary's market value already reflects. Subtract only parent-level and unlisted-segment debt when you take stakes at market.
• Minority interest: outsiders' share of the segments you just valued. Where the stake terms are known, take it at value instead of book.
• Contingent liabilities, guarantees, pension deficits (your full-filing lens).

1. Then judge the discount instead of denying it. Your NAV will usually come out above the market price. That gap is a market judgment with causes, and it isn't automatically an opportunity. Name the causes. Decide whether any catalyst can shrink them. Then value the discount's future as well as measuring its present.

What does a sum-of-the-parts sketch look like?

Here's a Reliance-shaped mini-sketch (illustrative, modeled on ~FY25 scale, so pull the actual segment note and rebuild). Segment EBITDA ≈ ₹60,000 crore (O2C) × 7 + ₹65,000 crore (digital) × 12 + ₹26,000 crore (retail) × 25 + ₹20,000 crore (upstream) × 4.5 + new-energy investment at cost. From that, subtract capitalized central costs, net debt + spectrum liabilities, and the ~33%/~11% minority stakes in the digital and retail arms sold to investors in 2020. On these round numbers you come out within ~10 to 15% of the market's price. The market broadly agrees with an SOTP here, because the pieces are separately funded, separately reported, and visibly headed for separate listings. Compare the pure holdcos below, where no such path exists and the discount yawns. All the numbers here are placeholders for the drill. Worked Example 4 does the full SOTP with complete arithmetic.

The holding-company discount: India's most reliable anomaly

A pure holdco owns stakes in other companies and does little else. The stakes are usually listed. Compute NAV = market value of stakes + net cash, and compare it with the holdco's own market cap. In developed markets the conglomerate/diversification discount averages ~10 to 15%. That's the classic Berger–Ofek estimate, contested in later literature, though the sign has held up. In India, pure-holdco discounts of 30 to 70% are normal and persistent. Bajaj Holdings, Tata Investment Corp, Godrej Industries, Kalyani Investment, JSW Holdings and Maharashtra Scooters have all traded in that band for years. Verify current marks, because the discounts breathe with market mood, narrowing in bull phases and in demerger winds.

Run it on a Bajaj Holdings-shaped example (illustrative, mid-2026). The stakes are ~33% of Bajaj Auto (market value ≈ ₹84,000 crore) and ~39% of Bajaj Finserv (≈ ₹1.26 lakh crore), plus other investments of ≈ ₹15,000 crore. That puts NAV at ≈ ₹2.25 lakh crore. The holdco's market cap is ≈ ₹1.4 lakh crore, so the discount ≈ 38%, and that's the narrow end of its own decade range (~35 to 60%).

Why is the holding-company discount so deep in India?

Why is the discount so deep in India? Know all five causes:

1. The cash never reaches you. The holdco receives dividends from the operating companies and passes on only what the promoter chooses. Realistically, the stakes themselves will never be sold. They're instruments of promoter control, and nobody is holding them as a portfolio to distribute. You own a claim on cash flows that structurally don't flow.
2. Tax friction at two gates. Dividends are taxed in the holdco's hands, with Section 80M relief only to the extent of onward distribution. Any hypothetical stake sale triggers capital-gains tax. Unless distribution is intended, the NAV you computed is pre-tax fiction.
3. Agency risk. Holdco treasuries have historically funded promoter priorities (new ventures, group support, pledged-share rescues) ahead of minority shareholders. Your related-party lens from the forensic case files applies with full force.
4. Illiquidity and float: thin trading, high promoter holding, index exclusion.
5. No terminal event. A closed-end fund can be wound up. An Indian promoter holdco, in practice, can't be forced to. With no catalyst the discount has no expiry date, so "it's at a 60% discount!" is a description. You still need a thesis.

Which catalysts have actually closed holdco discounts?

So which catalysts have actually closed discounts? Watch for these, and don't wait for mean reversion on faith: demergers and group simplifications (the 2023–25 Indian demerger wave: ITC's hotels demerger, Vedanta's proposed split, Tata Motors' CV/PV separation), buybacks at the holdco, promoter-family settlements that force asset splits, regulatory pushes (RBI's core-investment-company rules), or a real change in distribution policy.

That makes the SOTP output a conditional statement. NAV ₹X, price ₹Y, discount Z%. Here are the causes. Here is the catalyst, or its absence. Here is what I get paid to wait, which is the dividend yield while waiting. The separation mechanics are covered here and in the M&A and special-situations lab. What happens to the paper after a demerger, meaning the forced selling by index and mandate-bound holders and how to size it, is a drill set in the event-driven special-situations work. Go there once a catalyst is actually announced.