Sum-of-the-Parts Valuation

Put an aerospace supplier, a bolt distributor and a subscription software business under one holding company and ask the market what it is worth. Screening on EV/EBITDA, the group looks expensive against distributors and cheap against software. Neither comparison means anything, because the "company" being compared does not exist as a peer of anybody. The blended multiple is an average of things that should never have been averaged.

Sum-of-the-parts — SOTP, or break-up analysis — refuses to blend. Value each segment as if it were a standalone company against its own peers, add them up, then walk carefully from that total down to what a share is worth. Analysts reach for it whenever the parts genuinely trade differently: diversified industrials, media groups, holding companies, and any business with a fast-growing division buried inside a slow one.

The method

1. Value each operating segment. Use the segment's own metric and its own peer multiple, or a segment DCF if the disclosure supports it. This is the step that needs real segment reporting — SOTP dies where the 10-K only gives you one line.
2. Charge for the centre. Head-office costs sit above the segments and are not in any segment EBITDA. Capitalise them and subtract.
3. Bridge to equity. From total enterprise value, subtract net debt, subtract the value of minority interests in consolidated subsidiaries, add the value of stakes in things you do not consolidate, and subtract other claims like an unfunded pension deficit.
4. Divide by diluted shares and compare with the market.

Worked example

A diversified industrial group, all figures in millions. Segment EBITDA and the peer multiple that fits each one:

Segment	Metric	Peer multiple	Segment EV
Aerospace components	EBITDA 320	12.0× EV/EBITDA	3,840
Industrial distribution	EBITDA 180	9.0× EV/EBITDA	1,620
Software and services	Revenue 250	5.0× EV/Sales	1,250
Gross segment value			6,710

Unallocated corporate costs run $45m a year. Capitalising them at the group's blended 10× gives −450, so operating enterprise value is 6,260.

Now the bridge. Net debt is 1,850. The distribution arm is 80%-owned but fully consolidated, so its whole 1,620 of EV is already in the total; the 20% you do not own is worth $0.20 \times (1620 - 120) = 300$ after that subsidiary's own debt, and must come out. A 30% holding in a listed joint venture with a market capitalisation of 900 is not consolidated, so nothing of it is in the segment total; add $0.30 \times 900 = 270$. An after-tax pension deficit of 160 comes out.

$$6260 - 1850 - 300 + 270 - 160 = 4220.$$

With 210m diluted shares, that is $4220 / 210 = 20.10$ per share. The stock trades at 17.20, so the market applies a conglomerate discount of 14.4% — roughly the mid-point of the 10–15% range Berger and Ofek documented for diversified US firms.

Run it backwards: what does the market believe?

The more useful version of the same arithmetic is the reverse. At 17.20, market capitalisation is $17.20 \times 210 = 3612$. Walk the bridge the other way to recover the enterprise value the market is paying: $3612 + 1850 + 300 - 270 + 160 = 5652$.

Now hold the two easy segments at peer value and add back the corporate charge. The residual is what the market is paying for software:

$$5652 - 3840 - 1620 + 450 = 642.$$

That is $642 / 250 = 2.6\times$ sales, against a peer group at 5.0×. Instead of an unfalsifiable claim that "the stock is 14% cheap", you now have a specific, testable proposition: the market either disbelieves the software segment's peer set or expects the parent to keep it. That is something you can research.

One last caution: SOTP is a break-up value, and break-ups are not free. Separating segments triggers tax, loses shared overhead and distribution, and takes eighteen months. A persistent gap to SOTP is not automatically an opportunity; sometimes it is an accurate price for the frictions.