How Valuation Ranges Improve Decision Discipline

Discipline comes from what a range forces somebody to say out loud. Sankalp Industrial Systems Limited, invented, values at Rs 21,28,13,79,094 by one route and Rs 27,36,00,00,000 by another. Neither figure is a mistake. The two routes answer different questions, so the Rs 6,07,86,20,906 between them is the finding, and a single number would have buried the choice that produced it.

Underneath that sits a claim about where the content of a valuation actually lives. A value is the output of a set of choices. The choices carry everything; the output carries almost nothing on its own. Strip the choices away and what remains is a figure that looks settled and cannot be examined. A figure like that is a strange thing to hand somebody who has to decide.

The four values below are taken as given. Each one was produced somewhere else, by a method covered in its own place. The subject is not how the four numbers were made but what happens to a decision when all four are put on the same sheet of paper instead of one. Putting all four on one sheet is a question about discipline rather than about arithmetic, and that is the reason the four are worth printing together.

What does a single number actually take away?

Consider a second-hand scooter parked outside a building. Asked what it is worth, three people give three prices. The mechanic who services it prices it as a machine he could keep running for four more years. The neighbour who needs transport by Friday prices it as a problem solved this week. The dealer who will resell it prices it at what he can get for it later, minus his margin. Nobody is lying. The three prices differ because the three questions differ.

Now suppose somebody writes down one of those three prices, drops the other two, and hands over a slip of paper with a figure on it. More than two numbers have been lost. What has been lost is the fact that there were three questions. The slip of paper cannot be argued with. There is nothing inside it to disagree about, and the only two responses left to its recipient are acceptance and refusal.

The whole complaint about a point estimate sits in that slip of paper. The objection is not that a single figure is inaccurate. A point estimate might be a perfectly good figure. The complaint is that a single figure has thrown away every part of the work that a reader could have checked, and kept only the part a reader cannot do anything with.

Sankalp Industrial Systems Limited makes this concrete. Its model figure of Rs 21,28,13,79,094 rests on four visible choices. The cash flows are the ones the business generates as it stands. The rate is the company's own weighted cost of capital. Growth after the fifth year runs at 5.00 per cent forever. New capital put into the business earns 18.00 per cent. Four sentences, four places where a reader can push back. Print the figure alone and all four sentences stay in the drawer.

Why does one figure feel stronger than four?

Because certainty reads as competence. A slip of paper with one number on it looks like somebody did the work and finished. Four numbers look like somebody did the work and could not decide. Readers reward the first and are irritated by the second, and analysts learn that lesson quickly.

The feeling is real and it points the wrong way. A single figure feels stronger precisely in proportion to how much it has hidden, because everything it hides is something a reader might have objected to. Confidence and content pull against each other here: the fewer choices a writer shows, the fewer challenges the writing invites, and the fewer challenges it invites, the less anybody learns from reading it.

A household version of this is easy to feel. If somebody in the house says the flat is worth about eighty lakh, that ends the conversation. If they say it would fetch one figure in a hurried sale, a higher one after two months on the market, and a different one again if the cousin next door bought it, the conversation begins. The second version is less comfortable and enormously more useful, and it is more useful for exactly the reason it is less comfortable.

Is the range saying the value is probably in the middle?

The range is saying nothing of the kind, and this is worth settling before anything else is done with it. A confidence intervalA stated span around an estimate that carries a probability, drawn from a distribution somebody has specified in advance. comes out of a distribution that somebody has specified. A confidence interval carries a probability, and the probability means something because the distribution behind it means something.

Four separate valuations carry no distribution at all. Each of the four is a single deterministic calculation with its own inputs, and the space between two of them is not populated by anything. Nothing sampled it, nothing weighted it, nothing said the middle was likelier than an end. Between the ends there is empty paper, not a hump.

The distinction matters. A reader who quietly treats a range as a probability statement takes one obvious next step, and that step is to move to the middle. The middle then feels like the balanced answer. The middle is not balanced. The middle is a figure that no method produced and no question asked for, and moving to it is the most common way a range gets thrown away.

What are the four numbers actually answering?

The scooter shows the same thing. Once it is clear that the mechanic, the neighbour and the dealer were answering three different questions, the three prices stop competing. The three prices stop being three attempts at one answer and become three answers, each to its own question, and the sensible thing to do with them is to read all three.

Sankalp Industrial Systems Limited has been priced four times over and the same thing is going on. Before a single figure appears beside them, the four questions are worth setting out in plain words.

Read the four questions and the disagreement stops looking like a disagreement. Agreement between the four would be the surprising outcome, not the expected one, because nothing about the four questions makes them likely to have the same answer. A set of four figures that landed within a whisker of each other would be worth opening up, on the reasonable suspicion that three of them had quietly been anchored on the fourth.

What are the four values, and why does each one sit where it does?

Here they are. All four are an enterprise valueTake the operating business as one object and price that, leaving until later the question of who has a claim on it., and all four are measured against one Year 0 earnings before interest, tax, depreciation and amortisation (EBITDA)Profit taken before interest, before tax and before anything is charged for wear on the assets. figure of Rs 2,88,00,00,000, so they can be set beside each other with nothing further done to any of them. Each of the four was produced somewhere else and is restated here.

Five words carry that last column and every one of them belongs to a subject settled elsewhere, so they are worth hovering over rather than re-teaching. The lowest reading is a standaloneMeasured with nobody else's plans folded in: the business as it runs today, under the people running it today. one. The second is priced off what changes hands in a minority stakeA holding too small to direct anything, so whoever holds it rides along with what the controllers decide.. The third comes from a financial buyerA purchaser that funds heavily with borrowing and expects to sell again, rather than folding the business into one it already runs.. The highest comes from purchasers who took controlThe power to appoint the board and set the plan, which whoever buys the whole business acquires and a small holder never does., and some of them were also paying for synergyValue a purchaser expects only because two businesses would sit together, which neither could produce on its own.. Attach those five words to the four figures and the set stops being a scatter, because every entry now says what it was measuring.

Two of the distances inside that column are worth naming out loud. From the lowest reading up to the second is Rs 1,18,26,20,906, or 5.56 per cent measured against the lowest. The whole of that step is the difference between pricing the machinery yourself and reading a price off what similar businesses change hands at. From the second up to the highest is 1.70 turns of Year 0 EBITDA, being 21.79 per cent measured against that second reading, and in rupees that second step is over four times the first.

Two separate things are sitting inside the larger step and they are not the same thing. Purchasers in the precedent set ended up in charge of what they bought. Several of them also expected the business to be worth more inside their own group than outside it. How much of the step belongs to each is recorded nowhere, and any figure put on either half would be invented. So the step stays whole, and a reader who wants it divided has to go and find evidence rather than an arithmetic convention.

One more figure belongs here and is then left alone. Sankalp Industrial Systems Limited is listed, so its shares carry a price, and that price run through the balance sheet gives an enterprise value of Rs 22,40,00,00,000. The figure falls between the lowest of the four readings and the highest. Where it falls is the observation; whether it belongs there is a separate question.

When is a spread information, and when is it a defect?

Four numbers that differ are not automatically a range. Four figures become a range at the moment somebody writes a sentence beside each of them saying what that number assumed. Before those sentences exist, a reader has four figures and no way to tell whether the spread is telling them something about the business or something about the modelling.

Two readings can look identical in print and mean opposite things. An outlierA reading that sits well away from the rest of its set, which is a description of position and not of quality. in a set with reasons attached is informative. The same figure in a set with no reasons attached is a loose end. The arithmetic is the same in both cases; what separates them is whether anybody wrote down what each figure assumed.

What does the width itself say?

Most readers scan a range for its ends and skip the distance between them. The distance is the part worth reading first. The width measures how much of this company's value depends on which question was asked, and on this company that measurement is large.

Rs 6,07,86,20,906 separates the two ends. Against the low end that is 28.56 per cent. Against the high end the same rupees are 22.22 per cent. Both figures are right, they describe the same gap, and the difference between them is only the denominator. A percentage width printed without naming the end it was divided by hands the reader a figure they cannot place.

Now read what the width means rather than what it measures. Nearly three tenths of the lowest reading separates the price of this business running as it runs from the prices purchasers have paid to end up in charge of businesses like it. Put plainly: on this company, a large slice of what somebody might pay depends on wanting to be in charge and on what they expect to do once they are, rather than on the cash the machinery throws off. Either end on its own hides that, and on a decision of any size it is the most useful line in the file.

A narrow width would have been a finding too, and a different one. A narrow width would have said that the standalone reading and the control reading nearly agree. On most businesses that agreement would prompt somebody to ask whether the control reading had really been taken from buyers of control at all.

What does a range do to an argument?

Two people looking at one number can only argue about the number. One says it is too low, the other says it is about right, and both are stating a conclusion. Neither can produce evidence, because a conclusion is not a thing evidence attaches to. The argument runs until somebody gets bored or somebody outranks somebody.

Put a range in front of the same two people with a reason written beside each end, and the shape of the argument changes completely. The quarrel moves onto one of the reasons. Should a valuation of this business include what a buyer taking control has paid, or is the exercise about the business as it stands? The choice between those two has an answer, the answer depends on the purpose of the valuation, and two reasonable people can settle it in a few minutes.

What does a range cost the person carrying it?

Plenty, and pretending otherwise is how the argument for ranges gets dismissed by people who have to run meetings. A single figure is fast. A single figure fits in a subject line, it survives being repeated by somebody who did not read the file, and nobody asks a follow-up question about it in a corridor. Four figures with four reasons attached do none of that.

The costs are real and they are worth listing rather than talking around. Reading a range takes longer, because a reader has to hold four objects in mind instead of one. Carrying a range into a meeting invites the demand to narrow it, and a range narrowed on demand is the failure set out below. And a range cannot be acted on directly. Acting needs one course of action and a range is four readings, so the rule that converts a range into an action is a separate matter treated in its own place.

A range buys back a disagreement somebody can put a finger on, and every other benefit of a range is a consequence of that single one. The reason a lender can hold to the lower reading and say why, the reason two analysts can finish an argument, and the reason a file is still readable a year later are all the same reason: somebody wrote down what each figure assumed. A single figure buys speed by giving that up, and on a decision anybody will still be living with in a year, that is a bad trade.

Notice that the trade runs the other way on small, frequent, reversible decisions. Nobody prices a routine order four ways. The case for a range gets stronger as a decision gets larger, slower to reverse and harder to explain afterwards, which is exactly the shape of the decisions where somebody will later ask how it was reached.

Why not just average the four?

Because the average answers something none of the four methods was ever asked. Averaging the four values for Sankalp Industrial Systems Limited gives Rs 23,89,63,44,774, and that figure is rounding a half rupee upward, because the four sum to Rs 95,58,53,79,094 and an odd rupee count divided by four lands exactly on a half.

Look at the result and try writing the sentence that justifies it. Which conditions would put the value of this business at Rs 23,89,63,44,774? No sentence comes. The average is not the standalone value, it is not what the market pays for a minority stake, it is not what a debt-funded buyer could afford, and it is not what buyers taking control have paid. The average carries no assumption at all, so there is nothing in it to check and nothing in it to disagree with.

The midpoint of the two ends is a different figure again, Rs 24,32,06,89,547, and it sits Rs 42,43,44,774 above the average once that half rupee is rounded up. The two differ because the four readings bunch toward the lower part of the scale instead of spreading evenly along it. Neither is a valuation, and anyone using one of them has to say which. The average and the midpoint are not interchangeable and they are not close.

Why does the range have to be written before the decision?

A range written before a decision and a range written after one are different documents, even when they hold the same figures. The first is a record. The second is a reconstruction, and hindsight has already been through it.

The difference only shows up later. Twelve months on, somebody opens the file to work out whether the thinking was sound. Soundness of the thinking is a separate question from whether the outcome was good. A range dated before the decision, with a reason beside each end, lets them do that: they can read what was assumed, compare it with what happened, and say which assumption held and which did not. A range assembled afterwards cannot answer that question at all, however carefully it was built. The outcome was in the room while it was being written.

Reviewability is why the reasons matter as much as the figures. A file holding four numbers and a date is thin. A file holding four numbers, four reasons and a date is evidence about how a decision was actually reached, and it stays useful long after everybody has forgotten what they thought at the time.

So what does a written range actually contain?

Less than people expect, and the missing parts are usually the reasons rather than the figures. A file holding four numbers and a date is thin. A reader a year later can see what was believed and not why. Here is what a range has to hold if it is going to survive being read by somebody who joined afterwards.

The second and fifth rows are the ones that get left out, and they are the two that make the file worth keeping. Figures and dates get recorded because a form asks for them. Reasons get recorded only when somebody decides they matter. So many files end up holding a range that cannot be interrogated by anybody who was not there when it was built.

Who actually reads a range this way?

A lending committee does, and it reads the width before it reads the ends. A loan is repaid out of the business as it stands, so the standalone reading is the one that matters to a lender and the control reading is somebody else's problem. When the two are far apart, the lender learns that the collateral value they might have leaned on is being held up by what buyers of control would pay, and that is a support which disappears the moment nobody is buying.

The household version of that is a house valued three ways. The bank valuing it for a loan asks what it would fetch in a forced sale next month. The insurer asks what it would cost to rebuild. The people living in it ask what the cousins next door would pay. All three figures are correct and they will not agree, and the one that matters depends entirely on which of the three questions the household is living through this week.

An investment committee reads a range for a different reason. A single figure gives it nothing to interrogate, so the meeting turns into a debate about whether the analyst is any good. A range with reasons gives the meeting something to work on: it can accept the standalone reading, reject the control reading as irrelevant to what it is doing, and record why. The written reason is what makes the decision reviewable later, and reviewability is the whole reason a committee writes anything down at all.

A credit analyst working on Sankalp Industrial Systems Limited would note something specific from this set. The distance between the lowest reading and the highest is 28.56 per cent of the lowest. A gap that size measures how much of the value on offer depends on somebody wanting control, and a lender who has to be repaid out of ordinary operations has good reason to hold to the lower end of the four and say so in writing.