A company with three genuinely different divisions gets one share price, one earnings figure and one multiple, and that multiple is correct for none of the three businesses inside it. If one division is a slow-growing distributor worth seven times profit and another is a software operation the market would pay eighteen times for standalone, the blended number the group trades on tells you almost nothing about either.
Sum-of-the-parts is the fix and it is not difficult. It is an afternoon with the annual report, a page of arithmetic, and one honest decision at the end about how much to knock off for the fact that these businesses are stapled together. What follows is the whole bridge with numbers attached, because the shape of it is much easier to see when it is not abstract.
The caret, and what it hands back
Whether the expanded report on any given name carries divisional detail is something to check by opening it rather than to assume, and it will vary with what the company itself discloses. The reliable source, and the one this bridge is built on, is the segment note in the annual report. Every company reporting under either of the main accounting frameworks has to publish one, and it is the only place where the group breaks itself into the pieces management actually runs.
Find it by searching the annual report for the word segment. It is usually deep in the notes, and it usually gives revenue and an operating profit measure for each division, sometimes with assets and capital expenditure as well.
The one line that is always missing
The segment note has a gap in it that is deliberate and that most people paste straight past.
Add up the divisional profits and the sum is almost always larger than the group operating profit. The difference is the unallocated corporate cost, meaning the head office, the group finance function, the board, the listing costs and whatever central spending is not pushed down into a division. It usually appears as a reconciling line called corporate, unallocated, or other, with a negative number beside it.
That line matters more than its size suggests, because it is a permanent annual cost that has to be paid out of the divisional profits and has no revenue of its own. Value it the way you would value anything else, by capitalising it at a multiple, and subtract it from the sum of the parts. Skipping this step is the most common way an amateur sum-of-the-parts comes out twenty percent too high.
Two other reconciling items to note while you are in there. Inter-segment revenue, which is one division selling to another and which must not be counted twice. And any segment reported at a profit measure that differs from the others, which happens when a company reports one division before depreciation and another after, in which case put them on the same basis before applying multiples to either.
Picking a multiple for each part, the boring way
For each division, find three or four listed companies that do only that thing, and take their enterprise value to operating profit multiples. Not the median of a sector index, and not a multiple you remember. Three or four named companies you could list out loud.
Then apply two haircuts, mechanically, so that you are not arguing with yourself about them later.
A size haircut where the division is much smaller than the pure plays you compared it to, because smaller businesses trade at lower multiples for reasons that are structural rather than temporary. Ten to twenty percent is the usual range.
A quality haircut where the division's margin sits well below the peers', since a peer multiple assumes peer economics. If the peers run at eighteen percent margins and this division runs at eleven, you are applying their multiple to a worse business, and the multiple has to come down or the margin has to be argued up. Do one or the other, not neither, and never both at once.
The bridge, in round numbers
An invented company, three divisions, figures kept clean.
Division A, industrial distribution, revenue 1,200 at a twenty percent operating margin, so profit of 240. Pure play distributors trade around twelve times, giving 2,880.
Division B, contract manufacturing, revenue 800 at eight percent, profit of 64. Comparable manufacturers trade around seven times, giving 448.
Division C, software and services, revenue 400 at thirty percent, profit of 120. Comparable software businesses trade around eighteen times, giving 2,160.
The parts sum to 5,488. Now the subtractions, which are where the number becomes real.
Unallocated corporate cost of 60 a year. Capitalised at a blended ten times, that is a deduction of 600, leaving 4,888. Net debt of 900, taken from the balance sheet, leaves 3,988. Other claims ahead of shareholders, in this case a pension deficit and a minority interest totalling 100, leave 3,888.
Then the conglomerate discount. Fifteen percent is a defensible starting point for a group with no announced plan to separate anything, and the reasoning is that the market has to fund a head office it did not ask for and cannot buy the good division on its own. Fifteen percent of 3,888 leaves 3,305.
Divide by 250 million shares and the bridge lands at 13.22 a share against a market price of, say, 10.50. That is roughly twenty six percent, and more usefully, you can now see exactly which assumption it depends on. Division C alone accounts for 2,160 of the gross value on 400 of revenue. If the eighteen times multiple on that division is wrong by a third, most of the upside disappears.
Where this bridge will mislead you
Four things to hold in mind before you act on a number like that one.
The discount is a decision, not a market observation. Anyone can produce upside on any conglomerate by choosing ten percent instead of twenty. Write the discount down before you compute the total, and if you find yourself revising it after seeing the answer, you are no longer valuing, you are negotiating with yourself.
The gap needs a reason to close. A sum-of-the-parts discount can persist for a decade, and frequently does, particularly where a family or founder controls the votes and prefers the structure as it is. If nothing is announced and no shareholder is pushing, treat the upside as a slow release at best rather than a repricing.
Segment profit is not segment cash flow. A division can report good operating profit and consume every cent of it in working capital and capital expenditure. Where the segment note gives divisional capital spending, check it against divisional profit, because a business spending more than it earns is not worth a peer multiple on that profit.
And a separation costs real money. Advisory fees, duplicated head office functions in the new entity, sometimes tax leakage on the transfer. If your thesis is that the group splits, take a few percent off the value you would receive, because the shareholders pay for the transaction out of the same pot the upside comes from.