SOTP vs Consolidated Valuation: When Each One Is Right

A consolidated valuation values the group from one forecast and one rate. A sum-of-the-parts (SOTP) valuation prices each division on its own terms and adds up what comes out. Here the parts reach Rs 24,03,30,00,000 while the whole business trades at an enterprise value of Rs 22,40,00,00,000. The Rs 1,63,30,00,000 between them is 6.79 per cent of the parts, and what it means depends on four numbers somebody chose.

Sankalp Industrial Systems Limited, an invented manufacturer of industrial valves, precision castings, and the aftermarket parts and service that go with them, carries every figure below. Both builds are set out in full elsewhere, so each answer is stated in a line, and the one thing neither build can settle takes the rest: which of the two should have been running, and what the difference between them is allowed to mean.

The choice between the two methods is not a choice between a rough answer and a precise one. The choice is about what each method asks an analyst to assume. Valuing the group whole assumes that one set of economics fairly describes the whole business. Valuing it in pieces refuses that assumption and pays for the refusal immediately. Every division now needs a multiple of its own, and somebody has to choose it and defend it. One assumption that can be seen, against several that have to be argued for. One visible assumption against several arguable ones is the trade, and seen that way each method's proper place answers itself.

What is each method actually willing to assume?

A household with one salary and a small shop on the side can be budgeted for as one household: money in, money out, one picture. One picture works while the two sources behave alike. The moment the shop starts growing at a different speed, carrying different risk and swallowing money the salary never needed, one picture stops describing either thing well. The budgeter has not got worse. The household has become two things wearing one label.

A consolidated valuation makes the household assumption. One revenue path, one margin, one level of reinvestment, one discount rate, one answer. Consolidation is not a lazy method. The method carries a single visible premise. Where that premise holds there is exactly one thing to argue about, and one thing to argue about beats four.

A sum-of-the-parts valuation buys the freedom to describe each division properly, and the price is four separate arguments instead of one. Three divisions here means three multiples, none of which is measured. On top of that comes a decision about the cost that belongs to no division. Four judgements, four places for a reader to disagree, and one number at the end that carries all four of them without saying so.

Before any comparison, can the divisions be valued separately at all?

The separability question comes first, and skipping it is how most of the trouble starts. A sum of the parts is only a valuation of something if the parts could exist as things. Three conditions decide it. The comparison that follows assumes all three have already been answered, so each condition is named in a line here and developed properly elsewhere.

First, the divisions have to be genuinely different in their economics. Different margins, different growth, different cyclicalityHow far a business swings up and down in earnings as the wider economy expands and contracts., different appetite for capital. If three divisions all behave the same way, splitting them apart adds arithmetic and no information.

Second, there has to be enough separate reporting to value each one. Separate reporting means disaggregationBreaking a reported total down into the separate businesses that produced it, so each can be looked at on its own. that can actually be relied on, not a revenue split with everything else pooled. Guessing a division margin and then applying a multiple to the guess produces a number with two assumptions inside it and one of them invisible.

Third, the pieces need a believable way of coming apart. A demergerSplitting a division off into a company of its own, with the existing shareholders handed shares in the new one., a sale, a carve-outSelling a slice of a division to outside buyers while the parent keeps the rest of it.. Where nothing like that is available, the pieces total is a careful valuation of something with no way of coming into being, and setting it beside a traded price compares a company against a hypothetical.

So which method is right more often?

The consolidated method is the right one far more often than readers expect, and the reason is that the three conditions have to hold together rather than one at a time. Plenty of companies have segments. Far fewer have segments whose economics genuinely differ. Fewer again report enough for each to be valued honestly. And fewer still could actually be pulled apart. Divisions in a manufacturing group typically share plants, sales teams, engineering and a single set of customer relationships.

A different reason sits underneath that one, and it concerns whoever is building the valuation rather than the business being valued. A valuation done in pieces looks more thorough. The pieces build has more rows, more inputs, and more places where care is visible. Thoroughness of that kind is not evidence, and a reader who cannot see the four judgements inside the answer will read the extra detail as extra confidence. Reading detail as confidence is precisely backwards.

None of which means the method is exotic. A genuine conglomerateOne company running several businesses that have little to do with each other, reporting them under a single board. with an infrastructure arm, a consumer arm and a lender inside it cannot honestly be valued from one forecast, and there the pieces method is not optional. The point is that the case has to be made rather than assumed, and Sankalp Industrial Systems Limited is a borderline instance rather than an obvious one: three related industrial businesses sharing customers, sitting at margins of 23.0, 25.0 and 30.0 per cent.

What do the two methods say for this company?

Each division is priced on its earnings before interest, tax, depreciation and amortisation (EBITDA). The parts are stated once here, and everything after this points back at them.

Two things in that table are worth holding on to. The three divisions add to Rs 12,00,00,00,000 of revenue, exactly the consolidated figure, and nothing is missing on the top line. But segment EBITDA adds to Rs 2,97,00,00,000 while the group reports Rs 2,88,00,00,000, and that Rs 9,00,00,000 difference is the head office. Head office is real cost, it belongs to no division, and it has to go somewhere.

The consolidated side is one line. The market values the whole business at an enterprise value of Rs 22,40,00,00,000, or 7.78 times the same Rs 2,88,00,00,000 of EBITDA. One business, one multiple, one number.

So the parts say Rs 24,03,30,00,000, the whole says Rs 22,40,00,00,000, and Rs 1,63,30,00,000 sits between them. Both figures are enterprise values of the same invented company on the same day, built on the same Year 0 EBITDA, so they can be set against each other without any further adjustment. The traded figure sits inside the range this business is valued at by the four standard methods.

The gap is Rs 1,63,30,00,000. Per cent of what?

Two enterprise values are available here, so a percentage has two bases and they are not the same size. Divided by the parts, Rs 1,63,30,00,000 is 6.79 per cent. Divided by the traded value it is 7.29 per cent. Half a percentage point does not sound like much until two people spend twenty minutes disagreeing about a discount and discover they were quoting different denominators.

A discount is conventionally quoted against the figure it is being taken off, and that figure is the higher one. So 6.79 per cent of the sum of the parts is the standard form. Neither number is wrong. Printing either of them without the base in the same sentence is wrong, and the missing base is the single easiest defect to fix in any valuation note.

There is a third way to say the same thing that avoids the base problem altogether, and it is worth having. The parts build implies a multiple of 8.34 times on group EBITDA. The market pays 7.78 times. The difference is 0.57 of a turn, and 0.57 of a turn on Rs 2,88,00,00,000 of EBITDA is Rs 1,63,30,00,000. Stating a discount in turns rather than in per cent tells the reader immediately what would have to move for it to close.

Which four numbers is the whole gap resting on?

Now the part that matters. Three multiples were applied, being 7.5, 6.5 and 14.0 times. One decision was taken about the Rs 9,00,00,000 of head office cost, being to capitalise it at 7.8 times. Not one of those four numbers is a fact about Sankalp Industrial Systems Limited. Every one of them was chosen by whoever built the valuation.

Set that against the other side of the comparison. The traded enterprise value of Rs 22,40,00,00,000 is an observed number. The meaning of that number, whether it fairly reflects the business, and what it will do next are all open to disagreement. The number itself is not. The asymmetry is the whole of the comparison: a difference between an observed number and an assembled number is not a discovery about the observed one.

Where does the aftermarket multiple sit against the evidence?

Look at the aftermarket division on its own. Aftermarket parts and service brings 15.00 per cent of group revenue. Its share of segment EBITDA is 18.18 per cent. Its share of the gross value on the parts side is 30.56 per cent. Industrial valves, holding half the revenue, supplies 41.84 per cent of that value, and precision castings supplies 27.59 per cent. The smallest of the three divisions supplies almost a third of the value, and it does so entirely because of the multiple placed on it.

How far out is that multiple? The peer median for businesses of this kind is 7.8 times. Industrial valves at 7.5 times sits 0.3 of a turn below it. Precision castings at 6.5 times sits 1.3 turns below. Aftermarket at 14.0 times sits 6.2 turns above, or 1.79 times the median and 1.68 times the multiple the finished build implies for the group. Two of the three chosen multiples sit close to something observable. The third does not sit near anything.

None of that argues that 14.0 times is wrong. Aftermarket parts and service really do behave differently from the machines they serve, and a business at a 30.0 per cent margin with recurring demand may well deserve a multiple well clear of a valve plant. The argument is narrower and harder to dodge: the one multiple carrying the most weight in the answer is also the one with the least outside evidence behind it, and a reader looking at the finished Rs 24,03,30,00,000 cannot see that.

What happens when that one multiple moves?

Hold everything else exactly where the record locks it and move only the aftermarket multiple. Every single turn is worth that division's own EBITDA, Rs 54,00,00,000, so the relationship is a straight line with no curvature in it anywhere.

Two turns off the aftermarket multiple, from 14.0 to 12.0 times, take Rs 1,08,00,00,000 out of the answer, or two thirds of the entire discount. Six turns across the whole table span Rs 3,24,00,00,000, and somewhere inside those six turns the discount stops being a discount.

The relationship is a straight line, and a straight line teaches nothing the rows above do not already show. A calculator that moves all three division multiples together, where what happens genuinely is hard to predict, is set out separately.

The wrong lesson is easy to draw here, so take one clarification first. A turn on the aftermarket line is worth less than a turn anywhere else: Rs 54,00,00,000 against Rs 1,38,00,00,000 for valves and Rs 1,05,00,00,000 for castings. Measured per turn, aftermarket is the least powerful of the three. The aftermarket multiple is the fragile input for a different reason: distance, not leverage. That multiple is the only one of the three sitting far from any observable comparison, and a reasonable person could pick a very different number there and still be reasonable. Valves would have to give up 1.18 turns to close the gap alone and castings 1.56 turns, but neither has 6.2 turns of daylight between the chosen figure and the peer median.

At what multiple does the discount cease to exist?

Solve it directly. Industrial valves and precision castings contribute Rs 10,35,00,00,000 and Rs 6,82,50,00,000, and the head office line takes Rs 70,20,00,000 away, so those three together come to Rs 16,47,30,00,000. For the parts to equal the traded Rs 22,40,00,00,000 exactly, the aftermarket division has to bring Rs 5,92,70,00,000. Over Rs 54,00,00,000 of EBITDA that is 10.98 times.

A second route, sharing no step with the first, confirms it. The gap is Rs 1,63,30,00,000 and every turn on the aftermarket line is worth Rs 54,00,00,000, so the gap is 3.02 turns. Taking 3.02 turns off the chosen 14.0 lands on 10.98. Nothing about Sankalp Industrial Systems Limited changes between a 6.79 per cent conglomerate discount and no discount at all; one number that somebody picked moves by three turns.

Where does the whole gap actually come from?

The ladder shows one input moving. There is a better question underneath it: what produces the gap in the first place? The answer breaks apart exactly, and the way it breaks apart is worth more than the ladder above it.

Start with a thought experiment. Suppose all three divisions carried the same multiple, the peer median of 7.8 times, and the head office line was capitalised at 7.8 times as well. Then the parts would come to 7.8 times the group EBITDA of Rs 2,88,00,00,000, or Rs 22,46,40,00,000. Applying one multiple to every division reproduces the consolidated multiple valuation exactly. The pieces method can only ever produce a different answer to the extent that the chosen multiples differ from a single group multiple. None of that is a property of this company. The identity is arithmetic, and it shows what a sum of the parts is actually measuring.

So the gap is nothing more than the sum of three departures from that median, plus a small residue for the difference between the median itself and what the market pays. Valves at 0.3 of a turn below the median, on Rs 1,38,00,00,000 of EBITDA, takes Rs 41,40,00,000 off. Castings at 1.3 turns below, on Rs 1,05,00,00,000, takes Rs 1,36,50,00,000 off. Aftermarket at 6.2 turns above, on Rs 54,00,00,000, adds Rs 3,34,80,00,000. Net, the three chosen multiples add Rs 1,56,90,00,000. The peer median of 7.8 times against the traded 7.78 times on group EBITDA accounts for the remaining Rs 6,40,00,000, and the two together are Rs 1,63,30,00,000 to the rupee.

The decomposition changes the meaning of the discount, so read it once more. The aftermarket departure alone is Rs 3,34,80,00,000, more than double the gap being reported. The two give-backs are Rs 1,77,90,00,000 together. A reader handed a 6.79 per cent conglomerate discount is being handed the residue left after three large and independent judgements partly cancel each other out. A residue of three judgements is not a signal about the market. The residue is a signal about the judgements.

Is the head office line driving the result?

The head office line is the honest weak point of the method, and it deserves testing rather than defending. The Rs 9,00,00,000 of allocation basisThe rule chosen for spreading a shared cost across the units that benefit from it, such as revenue or headcount. problem is genuine: the cost is real, it belongs to no division, and not one of the three sets of division economics can say what multiple ought to be put on it. Capitalising at 7.8 times is a decision taken because something had to be decided.

So test the alternatives instead of arguing about them. Three defensible treatments of the same Rs 9,00,00,000, each applied to the same three division values.

The three treatments span Rs 4,75,45,455, or 2.91 per cent of the Rs 1,63,30,00,000 gap, so the head office decision is not what is producing the discount. The narrow span is worth knowing in both directions. It clears one of the three tests a real discount has to pass, and it also shows where the effort should not go. The allocation feels like accounting and therefore feels solvable, so the temptation with a sum of the parts is to polish it. The multiples are the part that decides the answer, and they are the part that cannot be solved, only argued.

The third treatment is worth a sentence on its own. The build itself implies 8.34 times for the group, so capitalising the head office cost at 7.8 times while the divisions carry 7.5, 6.5 and 14.0 times is quietly inconsistent. Making the deduction agree with the answer it feeds costs Rs 4,75,45,455 and removes an inconsistency a careful reader will find. The change touches no conclusion, and that is exactly why it is safe to do.

What would have to be true for the gap to be a real discount?

Three things, and all three are about evidence rather than arithmetic. None of them is satisfied by working the numbers more carefully, and that is what makes them uncomfortable.

First, each division multiple has to be defensible against something outside the model. Outside the model means pure playA listed business that does one kind of thing only, so its price prices that one activity and nothing else. companies doing only that one thing, or transactions in which only that kind of business changed hands. Here two of the three multiples sit close to the peer median and the third sits 6.2 turns above it with nothing named to support the distance. Until that is fixed, the largest single component of the gap is unevidenced.

Second, the head office treatment has to be shown not to be driving the result. The test has been run above and it passes: three treatments span 2.91 per cent of the gap. Running it and printing the answer is the whole discipline. Assuming it would pass is not.

Third, separation has to be more than a thought experiment. Somebody has to be able to take one of these divisions away. Separation is the condition people skip, and everything about whether the comparison means anything rests on it. If the three divisions share plants, engineers and customers so completely that no buyer could take one, then a valuation of them separately prices a company nobody could ever own, and its distance from the traded price is a fact about a hypothetical.

A discount that passes all three is a finding worth reporting; one that fails any of them is arithmetic. Note also what none of these tests asks. None of them asks whether the discount is large. Size is not evidence, and a 6.79 per cent figure that survives all three tests carries more weight than a twenty per cent figure that survives none.

What does a discount not establish on its own?

A conglomerate discount, even a well evidenced one, is a difference between two numbers. The discount recommends nothing, passes no verdict on the people running the company, and predicts nothing about what the gap will do next. Set beside it the holding company discountThe gap between what a parent company is itself valued at and the value of the stakes it holds in other companies., a related idea measured on a different structure. Both figures describe a distance and neither describes a cause.

Why such a distance persists, once it has been properly established, is a separate subject with its own causes: what a group discloses, who holds the shares, how the pieces are followed by the people who follow them. The causes are covered elsewhere, and they take the 6.79 per cent as a given rather than rebuilding it.