Trading Multiples: Applying the Peer Set to One Company

Six invented peer multiples applied to one invented company give two different answers. Count all six and the median is 7.8 times while the mean is 8.63. Those two statistics put Rs 22,46,40,00,000 and Rs 24,86,40,00,000 of enterprise value on one unchanged figure for earnings before interest, tax, depreciation and amortisation (EBITDA). With the sixth peer switched out, both statistics land on 7.60 times.

Underneath every one of those movements is a single difference between two ways of summarising a set of numbers. A median asks which observation sits in the middle, so an observation far from the rest gets exactly one vote, the same vote as its neighbour. A mean adds up how far every observation sits from everything else, so a distant observation carries its whole distance into the answer. Everything the tool below does is a consequence of that one difference.

What does this calculator hold still, and what does it let move?

One thing moves: whether a given peer is counted. Six switches, six invented companies, and nothing else on the screen changes when one is flicked. The subject company's own figures are frozen throughout, so every movement comes from the peer set and never from the business being valued.

Sankalp Industrial Systems Limited, an invented manufacturer, made Rs 12,00,00,00,000 of revenue in Year 0 and Rs 2,88,00,00,000 of EBITDAEarnings before interest, tax, depreciation and amortisation. The measure leaves out what the funders charge, what the tax authority takes and what past spending is being written down against. Everything left over is close to the trading itself. on it. Those two figures make a 24.0 per cent EBITDA marginThe same EBITDA figure set against the revenue of the identical year, a ratio that lines a small maker and a large one up on one scale., and the margin does not move. Its shares and its borrowings together price the whole business at Rs 22,40,00,00,000. The division is Rs 22,40,00,00,000 over Rs 2,88,00,00,000, or 7.777778 times, printed as 7.78. The traded multiple does not move either.

The tool computes one statistic across whichever peers are switched on, then multiplies that statistic by the frozen Rs 2,88,00,00,000. Every rupee figure the tool prints is the same EBITDA wearing a different peer set's opinion of what a rupee of it is worth. Nothing about Sankalp Industrial Systems Limited is being re-examined between one reading and the next.

Who are the six peers, and which one behaves differently?

Six invented companies, all in industrial equipment, each with a revenue line, a margin, an enterprise value and therefore a multiple. Five of them look like variations on each other. One does not.

Read the last column downwards and the shape jumps out. Peers 1 through 5 run 6.6, 7.1, 7.6, 8.0 and 8.7 times, a spread of 2.1 turns across five companies. Peer 6 sits at 13.8. Nallamala Components Limited is 5.1 turns above the next highest peer in the set, a distance wider than the entire spread of the other five put together.

Nallamala Components Limited got there honestly. Its margin of 30.0 per cent is four points above peer 5 and ten points above peer 1. Its revenue is growing at 18.0 per cent a year against 9.5 per cent for the next fastest. Nallamala Components Limited carries net cashThe subtraction of borrowings against the bank balance landing the other way round, where the money in the account is larger than everything owed to lenders. where every other company in the set carries net debtBorrowings outstanding after the bank balance has been taken off, so a company that raised money and has not yet spent it is not counted as though it had.. And it is small: Rs 5,00,00,00,000 of revenue is 41.67 per cent of Sankalp Industrial Systems Limited's Rs 12,00,00,00,000 and 62.50 per cent of peer 5, the smallest of the others. Whether a company like that belongs in the set at all is a genuine question with two defensible answers, and it is settled elsewhere. The set is taken as given here, and what follows asks what happens to the arithmetic either way.

Why does one distant peer pull a mean and leave a median alone?

Six flats sold in one building over the past year have the same shape. Five went for prices within a few lakh of each other, and the sixth was the duplex on the top floor with the terrace. The duplex went for three times any of them. Because the duplex is one sale out of six, and the middle sale is unaffected by how spectacular the top one was, a broker asked what a flat in that building fetches names the middle price. A spreadsheet asked for the average returns a figure the duplex has dragged somewhere no ordinary flat in that building has ever sold.

The two statistics are not two attempts at the same question; they are two different questions asked of the same six numbers. One asks about position in an ordered line. The other asks about total distance. Peer 6 occupies exactly one position and carries an enormous distance, so it is invisible to the first and dominant in the second.

What happens when peer 6 comes out?

Switch Nallamala Components Limited off and five multiples remain: 6.6, 7.1, 7.6, 8.0 and 8.7. The median is now simply the third of five, or 7.6 times. The mean is 38.0 divided by 5, or exactly 7.60. Removing one company out of six moves the median by 2.56 per cent and the mean by 11.97 per cent, more than four times as far, and lands the two statistics on the identical number.

Both percentages are computed on the unrounded mean of 8.633333 rather than on the 8.63 that gets printed. Read them with that in mind. Working from the printed figure instead gives 11.94 per cent, just as plausible-looking and wrong by three hundredths of a point. The same slip on the rupee side is larger: the mean applied to Rs 2,88,00,00,000 is Rs 24,86,40,00,000 on the unrounded value and Rs 24,85,44,00,000 on the printed one, a difference of Rs 96,00,000 conjured out of nothing but a display convention.

The coincidence at 7.60 times looks like luck and is not. Measured against their own middle observation of 7.6, the five remaining multiples sit at distances of minus 1.0, minus 0.5, zero, plus 0.4 and plus 1.1. The five distances add to exactly zero. Once the amounts above the middle and the amounts below it cancel, the identity in the box above has nothing left to add, so the mean has nowhere to sit except on the median. Peer 6 was not merely one contributor to the gap between the two statistics; at plus 6.0 turns against a total of plus 5.0, it was the entire gap and then some, with the other five pulling 1.0 turn back the other way.

What enterprise value does each statistic imply?

The arithmetic is one multiplication and the point is what sits on either side of it. Rs 2,88,00,00,000 of Year 0 EBITDA at 7.8 times is Rs 22,46,40,00,000. The same Rs 2,88,00,00,000 at 8.633333 times is Rs 24,86,40,00,000. Rs 2,40,00,00,000 separates those two figures, and every rupee of it came from choosing which statistic to quote rather than from anything anybody learned about the company.

The gap can be reached a second way, and the second route confirms the figure rather than merely repeating it. The mean sits 0.833333 turns above the median, and 0.833333 turns of Rs 2,88,00,00,000 is Rs 2,40,00,00,000 exactly. The two routes agree to the rupee. Notice the size of the gap in ordinary terms: Rs 2,40,00,00,000 is more than five sixths of an entire year of the company's EBITDA, produced by a decision that would take a keystroke.

Once peer 6 is out, both statistics sit at 7.6 and both imply Rs 21,88,80,00,000, and the choice between them stops mattering at all. The collapse onto one number is the honest description of a statisticA single figure taken across a set of observations to stand in for all of them, throwing away by construction everything about the spread it came from. here: when the set is balanced the choice is free, and when it is lopsided the choice is expensive.

How wide is the answer across the whole set?

A median and a mean are both compressions, and a compression by definition throws away the thing it was compressing. The rangeThe lowest and the highest observation in a set, reported together as a pair rather than reduced to one figure between them. is what survives that compression, and on this set it is uncomfortable. Peer 1 at 6.6 times implies Rs 19,00,80,00,000 for Sankalp Industrial Systems Limited. Peer 6 at 13.8 times implies Rs 39,74,40,00,000. Six companies selected as comparables imply a top figure more than double their own bottom figure. The word comparable suggests nothing of the sort to anybody reading a note.

Think of asking six caterers to quote for the same wedding, the same guest count and the same menu. If the cheapest quote is Rs 4,00,000 and the dearest is Rs 8,50,000, nobody would report the middle quote as the price of the wedding and leave it there. The honest report gives the range the quotes covered and then explains why the dearest one was so much dearer. A peer set deserves the same treatment, and reporting the range beside the statistic is usually the more honest output rather than the more cautious one.

Taking peer 6 out narrows the top considerably without touching the bottom. The remaining five run 6.6 to 8.7 times, implying Rs 19,00,80,00,000 to Rs 25,05,60,00,000. Peer 1 was never the problem, so the low end does not move.

How does the subject company sit against its own peer set?

Sankalp Industrial Systems Limited is priced by its shares and its borrowings together at Rs 22,40,00,00,000. Set against Rs 2,88,00,00,000 of EBITDA, that price is 7.777778 times. The median of the six invented peers implies Rs 22,46,40,00,000. Rs 6,40,00,000 separates the traded figure from the median-implied one, or 0.29 per cent of the traded enterprise value. The subject sits close to the middle of this set on this one measure.

One detail there is worth pausing on, and rounding of this kind quietly corrupts a comparison. Subtracting the printed cells gives 7.8 less 7.78, or 0.02 turns. Computing from the unrounded values gives 0.022222 turns. The two answers agree to two decimal places on this occasion, and the agreement is luck rather than method: 0.022222 turns of Rs 2,88,00,00,000 is Rs 6,40,00,000, and the same subtraction done on rounded cells anywhere else in this set would not land on the same rupee. Take every spread from the unrounded figures and treat the printed column as a display.

The closeness licenses no verdict. The subject sits near the median of an invented set on one ratio, and that sentence is the whole of the claim. A company priced away from its peers is either priced for something an EBITDA multiple does not capture, growth, capital intensity, the durability of a margin, or it is not, and this ratio has no way of telling those apart.

What does a working analyst actually do with these three numbers?

An equity analyst writing a note does not pick one of the three. The line that goes into a valuation section reads as a median, a mean and a range together, with the count of companies and the names of any excluded ones sitting in the same sentence. A colleague reading it six months later can then rebuild the set. A statistic that arrives without its set behind it cannot be checked by anybody. A statistic nobody can check is a claim rather than a measurement.

A lender looking at the same seven companies uses hardly any of this. A lender's question is whether the interest gets paid, so the column that matters is net debt against EBITDA, where Sankalp Industrial Systems Limited sits at 1.67 times and peer 5 at 0.6 while peer 6 carries cash rather than borrowings. An enterprise value multiple tells a lender what somebody would pay for the business, a figure that matters only in the situation the lender is trying to avoid.

An investor buying a few hundred shares on an exchange is buying a minority stakeA holding too small to direct what the business does, which is what a parcel of shares bought on an exchange usually amounts to., and a trading multiple measures exactly that: what the market currently pays for a small slice of a similar business, on the day it is measured, with no control and nobody's plan attached. A small slice priced on one day is a narrower thing than what a business is worth, and the narrowness is the point rather than a weakness. The narrowness also explains why the return on invested capitalOperating profit after the tax charge, set against the money actually tied up in the business to produce it. of each peer moves in step with its multiple: peer 1 earns 11.0 per cent on its capital and trades at 6.6 times, peer 6 earns 24.0 per cent and trades at 13.8. The multiple is not arbitrary; it is compensating for something the multiple itself does not name.

What is the tool not saying?

Look at the pine line under the readouts. The line does not print a bare multiple; it prints the statistic, the count, the names left out, which multiple was used and which period it covers, all on one line. An output line built to carry its own exclusions means a figure copied out of the tool arrives somewhere else with its peer set still attached to it.

The tool still cannot settle whether the six companies belong together, whether Year 0 was a normal year for any of them, whether one of those EBITDA figures was flattered by something that will not repeat, or whether the market that priced all six was pricing the same future the reader has in mind. All four are questions about the set, and the set is taken as given here.

What is a reasonable way to report all of this?

Write the set down first, in full, with a reason beside each name. Compute the median, the mean and the range together and print all three. Say how many companies are in the set and name any that were considered and left out, with the reason for leaving them out written at the same time as the decision rather than afterwards. State the multiple and the period. Everything on that list exists to make the figure reproducible by somebody who was not in the room. Reproducibility is the only property that separates a measurement from an assertion.

Then state the questions the figure leaves open. On this invented case, six comparables produce a middle answer of Rs 22,46,40,00,000, an average answer of Rs 24,86,40,00,000 and a spread from Rs 19,00,80,00,000 to Rs 39,74,40,00,000, all against a traded Rs 22,40,00,00,000. Every one of those is an output of the same method with a different setting, and the reader is entitled to see all of them rather than the one that happened to get quoted.

Sankalp Industrial Systems Limited, Aravalli Flow Controls Limited, Satpura Engineering Works Limited, Kaimur Industrial Limited, Girnar Precision Limited, Shivalik Systems Limited and Nallamala Components Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.