How Companies Get Valued

Three ways to answer one question

Trading comparables value a company against what the market pays today for similar listed businesses. Precedent transactions value it against what acquirers have paid to buy similar businesses outright. A discounted cash flow values it on the cash the business itself is expected to generate.

The first two are relative measures: each one prices the business against what someone else is paying right now, so both inherit the market's mood, its mistakes included. A DCF works the other way. It is intrinsic, so the output depends only on your own assumptions about the business, which is exactly why it is the one method that can catch the market being wrong, and the one that can be quietly wrong for a hundred slides without anyone noticing.

No single method wins. In practice you build all three, lay the ranges side by side, and treat the gaps between them as the real analysis: where they disagree is usually where the interesting judgement about this specific company actually sits.

Why the ranges sit where they do

Precedent transactions almost always come out highest, and the reason is control. An acquirer buying the whole company pays a premium over the traded price because it gains the right to direct the business, and often because it expects synergies that a passive shareholder could never realise. Trading comps, by contrast, reflect minority stakes with no such rights.

A DCF can land anywhere. Its output is extremely sensitive to the discount rate and to terminal value assumptions, which is precisely why interviewers ask about those two inputs more than any other part of it.

Three methods, one company, three answers

Ranges on a chart stay abstract until you run the arithmetic once. Take a business with 200 of LTM EBITDA, 400 of net debt and 100 million shares outstanding, and value it three ways.

The listed peer set trades at a median 8.0x EV/EBITDA. Apply it and 8.0 times 200 gives an enterprise value of 1,600. Subtract the 400 of net debt and equity value is 1,200, or €12.00 per share. That is what the market pays today for a minority stake in a business like this one.

Deals in the sector cleared at a median 10.0x. That is not a different opinion about the business. It is the same 8.0x with a 25% control premium on top, because those buyers were acquiring the whole company rather than a slice of it. 10.0 times 200 gives 2,000, and €16.00 per share.

The DCF ignores both and asks what the cash flows themselves justify: 620 of present value from the explicit forecast years plus 1,230 from the terminal value, so 1,850 of enterprise value and €14.50 per share. Note that two thirds of that answer sits in the terminal value, which is why the growth rate and the discount rate get so much attention.

€12.00, €16.00 and €14.50 are not three failed attempts at one number. They answer three different questions: what a share is worth today, what the whole company is worth to a buyer who wants control, and what the underlying cash flows support. The gap between the first two is close to the control premium by construction. If you cannot explain a gap, that is precisely where the analysis needs more work.

Choosing what to lead with

For a stable, mature business with real listed peers, trading comps do most of the work and a DCF supports them. For a business with volatile or negative near term earnings, multiples become close to meaningless and the DCF has to carry more weight. For an early stage business with no profits at all, revenue multiples and scenario analysis are often the honest answer, and pretending otherwise is worse than admitting the limits.

If a company is an acquisition candidate, precedents matter far more than usual, because they show what someone paid rather than what the market thinks in the abstract.

One constraint sits above all of these and is easy to forget in a room: what data exists. A private mid market target in Europe frequently has no listed pure play peer, no disclosed transactions in its niche, and financial statements filed late and in abbreviated form. That is not a reason to pretend a comparables analysis is available. It is the reason to lead with the cash flows and management's own numbers, tested hard, and to say which methods you would have run if the evidence supported them. An answer of the form I would lead with this because that one has no usable inputs here is a stronger answer than any recital of the menu.

When a method is the wrong tool, not just a different one

So far this module has treated the three methods as three answers to one question, and most of the time that is exactly right. Sometimes it is not. A method can be unavailable rather than merely inconvenient, and the difference matters more in an interview than almost anything else in valuation. A candidate who runs a discounted cash flow on a business whose cash flows nobody can forecast has demonstrated that they know the mechanics, not that they know what the mechanics are for. What an interviewer is testing when they hand you an awkward company is whether you can decline a method and say why.

Start with the discounted cash flow, and the first way it breaks. A forecast has to carry information. Take a single asset biotech whose entire value depends on one trial read out: if the drug is approved the business is worth a great deal, and if it is not the business is worth its cash and its remaining programmes. Those are two futures, not one. The honest structure is a probability weighted valuation that values each branch and weights it. Illustratively, 1,200 if approved and 100 if not, at a 30% chance of approval, gives 0.3 times 1,200 plus 0.7 times 100, which is 430. A single forecast built on risk adjusted revenue can land close to 430 and still be useless, because 430 describes a company that exists in neither future. The number is the same and the analysis is not. The same problem shows up in early stage technology, in a business whose value turns on one court ruling, and in anything where the outcome is a switch rather than a dial.

The second way it breaks is definitional. Unlevered free cash flow assumes debt is a financing choice sitting outside the operations. For a bank or an insurer that assumption is false: deposits and reserves are raw material, leverage is the business, and regulation sets how much capital has to be held against it. There is no meaningful enterprise value to discount to, which is why an enterprise value to EBITDA multiple on a bank is not a conservative approximation but a category error. The replacement methods are named in the section on other methods worth knowing, and the point here is narrower: the method did not become hard, it stopped applying.

The third way is subtler and catches people who have built the model correctly. Take a five year forecast producing free cash flow of 80, 85, 90, 95 and 100, discounted at 9% with a 2% perpetuity growth rate. The explicit years are worth 347 in present value and the terminal value is worth 947, so enterprise value is 1,294 and roughly 73% of it sits beyond the forecast. Now move the growth rate to 3% and enterprise value becomes 1,462, up 13%. Then put the growth rate back and instead lift every single explicit year by 10%, which is a large forecasting change: enterprise value becomes 1,328, up 2.7%. One percentage point on a rate nobody can observe is worth nearly five times a 10% improvement across the whole forecast you actually researched. That is not an argument against a discounted cash flow. It is an argument for knowing what you have built, and it is fatal only where the business is one whose near term forecast was guesswork to begin with, because then the model is an elaborate presentation of a single assumption. The terminal value module works the sensitivity through properly.

Trading comparables fail for a different reason: sometimes there is no set. A pure play peer is a listed company whose economics are driven by the same thing as your target's, and for some businesses none exists. A company that is the only listed operator of its kind, a business that is a bundle of unrelated activities, a national infrastructure asset with a regulated return: each of these has an industry label and no peers. The trap here is the fix rather than the problem. When four names are not enough, the instinct is to widen the screen until there are ten, and the set that comes back is a sector average rather than a valuation of your company. Widening is the wrong direction. Either present a small set and argue explicitly where the target sits relative to each name, or demote comparables to a cross check and lead with something else, and say out loud that this is what you have done. There is also a quieter version of the failure, where the peer set exists but is priced for something the target does not have: if half the names are trading on bid speculation or on an option the target lacks, the median measures that option and then charges your company for it.

Precedent transactions fail when the conditions that set those prices no longer hold, and the honest way to see it is through the discount rate. Take the same forecast as before and discount it at 7% rather than 9%, holding the growth rate at 2%: enterprise value is 1,821 instead of 1,294. Two points on the discount rate is 29% of the answer, on a business where nothing changed. Every deal struck in a lower rate environment was priced by buyers whose cost of capital and whose available debt looked like the 7% world, and quoting their multiples today is quoting an answer to a question that has been withdrawn. Then there is the thin set problem, which is common in Europe because so many targets are private and disclosure is lighter. Three deals is not a distribution. If one of them was a carve out sold without its share of central costs, one was a competitive auction and one was a bilateral negotiation between a founder and a long standing partner, a median of the three is arithmetic performed on things that are not alike. Where the set is that thin, present the deals individually with their dates and their circumstances, and let the reader see how little evidence there is rather than hiding it inside a median.

The last one is not a failure of the method but of what people do with it. An ability to pay analysis solves an LBO backwards from a required return to produce the highest price a financial sponsor could justify. That is a genuine and useful number, and it is not a valuation, because it answers what a particular buyer with a particular hurdle can afford rather than what the business is worth. Two things go wrong. The first is circularity: an ability to pay run that assumes the sponsor exits on the same multiple it paid has assumed that today's price is the right price, so the output tells you about the leverage and the growth and nothing at all about value. Anchor the exit to where the peer set trades instead, and be ready to defend that choice, because it is the assumption that moves the answer most. The second is that it is only a floor where the business has genuine debt capacity. A cyclical business with volatile EBITDA, or one still burning cash, supports very little debt, so the sponsor's maximum price collapses toward what an all equity buyer would pay and sits far below anything a strategic buyer would offer. Presenting that as a floor implies a level of support that is not there.

None of this means refusing to do the work. When you are asked for a method that does not fit, the answer that lands has four parts: name the method, name the input that does not exist, name the substitute, and name what the substitute still cannot tell you. For the biotech that is: a discounted cash flow needs a forecast, this forecast is a coin flip, so weight the branches, and accept that the answer is now only as good as the probability, which is itself a judgement rather than an observation. That last clause is the part most candidates leave out, and it is the part that sounds like someone who has done the work rather than read about it.

Other methods worth knowing

The three above cover most interviews, but three more come up often enough to name. Sum of the parts values a diversified business by valuing each division separately, on the methodology that fits it, then adding them together. It matters whenever a company mixes businesses that would trade on very different multiples on their own, since a single blended multiple hides that entirely, and a discount to the sum is often the whole investment case for a conglomerate.

Asset based valuation values the business as the sum of what it owns, adjusted to fair value, minus what it owes. It carries little weight for a healthy operating business, but becomes the primary method in liquidation, in real estate and investment holding companies, and in distressed situations where going concern cash flows are not credible.

The LBO implied value already appeared in the football field above as the lowest band. It deserves its own mention because it answers a different question from the other three: not what the business is worth, but the highest price a financial sponsor could pay and still hit its required return. That makes it a floor a strategic buyer can usually beat, not a valuation in its own right.

One more belongs on the list because the standard three do not apply to it at all. Banks and insurers cannot be valued on enterprise value, since debt is their raw material rather than a financing choice, so there is no capital structure neutral figure to discount to. The practical replacements are a dividend discount model, which values the stream distributable to shareholders after the regulatory capital the business is required to hold, and a price to tangible book multiple read against return on tangible equity, which is the relative version of the same idea: a bank earning more on its equity than its cost of equity should trade above book, and one earning less should trade below. The financial institutions module builds both properly.

Which method leads, and who is asking

Nobody in practice assigns percentage weights to methods and takes a weighted average. What happens instead is that one method leads and the others test it, and what decides the lead is the decision being made and the person making it. This is the real content behind triangulation, which is often described as if it meant averaging. It does not. It means reaching one answer by independent routes, so that agreement is evidence rather than construction. Two methods sharing an assumption do not corroborate each other, they repeat each other.

Start with a target board that has received an offer. The question in the room is not what the company is worth in the abstract but whether this price beats staying independent, so the analysis that leads is what other sellers achieved: precedent multiples and a premiums paid analysis, which measures each deal's offer against the target's undisturbed price rather than against the price on the day, since by the day before an announcement the shares have often already moved. The discounted cash flow does not disappear, it changes role: it becomes the value of the plan the board would otherwise be asking shareholders to back. In the UK this is not only good practice. Rule 3 of the Takeover Code requires the board of an offeree to obtain competent independent advice on the offer and to make the substance of that advice known to shareholders, which is why the standalone case has to be defensible rather than decorative.

A financial sponsor is asking a different question and gets a different lead. Its constraint is a required return rather than a view on value, so the leveraged buyout leads and everything else tests the assumptions inside it. Comparables and precedents do not compete with the answer, they discipline the exit multiple, which is the input the whole return depends on.

A lender is asking a third question, and this is the one candidates least expect. A credit committee does not want a value per share. It wants to know whether interest gets paid and principal comes back if the business underperforms, so the analysis that leads is a downside case, and the outputs are leverage, coverage and headroom rather than a price. Asset value matters here in a way it does not in an equity valuation, because it is what remains if the cash flows do not arrive. A candidate who answers a lending question with a football field has answered the wrong question competently.

Then there are the cases where the method is chosen for you. Under IAS 36 an asset or cash generating unit is impaired when its carrying amount exceeds its recoverable amount, which is the higher of fair value less costs of disposal and value in use. Value in use is a discounted cash flow with rules attached: it is built on the asset in its current condition, so it excludes the benefit of future restructurings the entity is not yet committed to and of capital expenditure that improves the asset rather than maintaining it, and it is measured on a pre tax basis. The practical consequence is worth carrying into an interview. The same asset can be worth more to a buyer than its value in use, entirely legitimately, because the buyer is allowed to price its own plans and the impairment test is not. An impairment charge is not a statement that a disposal at a higher price is impossible.

Continental practice has its own version of this. Where a German squeeze out or a domination agreement is challenged in an appraisal proceeding, the valuation follows the profession's own standard, IDW S1, and the workhorse is the capitalised earnings method, which discounts sustainable distributable earnings to the shareholder rather than unlevered cash flow to the firm. It sits closer to a dividend discount model than to a banker's enterprise value build, and the forum, not the analyst, decides that. The general lesson generalises well beyond Germany: before choosing a method, find out whether anyone has already chosen it for you.

Equity research is the last common case and the most familiar. A twelve month target price is a forecast of where a stock trades, and where a stock trades in twelve months is mostly a function of what the market pays for the sector, so comparables lead and the discounted cash flow supplies the longer view and the argument for why the multiple should change. That is the opposite emphasis from an appraisal and the same toolkit.

All of this comes together on the page where the ranges are stacked, and the part worth learning is what to do when there is no overlap band. The instinct is to treat it as a presentation problem and quietly widen the bars until they touch. It is the opposite: non overlap is the most informative thing the exhibit can tell you, and each specific gap points somewhere. Comparables sitting entirely above the discounted cash flow usually means your growth or margin assumptions are below what the market is pricing, or that the sector is expensive, and the way to tell is to back out the multiple your terminal value implies and put it next to where the peers actually trade. A discounted cash flow sitting entirely above the precedents usually means the terminal assumptions are heroic, or that the market has derated since those deals were struck. Precedents far above everything is normally just control and synergies doing their job, and becomes a problem only when the set is old. And a bar that is supposed to be the floor sitting above the comparables is telling you about the credit market rather than about the company.

That last one is worth the arithmetic, because it is the least intuitive. Take a target with 120 of EBITDA, 240 of existing net debt and 50 million shares, whose peers trade at 8.0x, giving 960 of enterprise value and €14.40 per share. A sponsor expects EBITDA of 150 in five years and, disciplined about the exit, values that at the peer multiple of 8.0x, so 1,200. In a tight market it raises 4.5x, or 540 of debt, and repays 150 over the hold, leaving 390, so exit equity is 810. At a 2.5x money hurdle it can put in 324 today, which supports 864 of enterprise value, 7.2x, and €12.48 per share. Now open the credit market: 6.5x of debt is 780, more of the cash goes to interest so only 100 is repaid, exit debt is 680 and exit equity is 520, which at the same 2.5x hurdle supports an equity cheque of 208 and an enterprise value of 988, or 8.2x and €14.96 per share. The sponsor writes a cheque a third smaller, clears exactly the same return, and pays 124 more for the company. Nothing about the business moved. The floor is now above the trading comparables, which is precisely the signal that the price in that market is being set by financing rather than by the asset.

So the discipline on a page with no overlap is: do not average, do not widen, and do not present the disagreement without a sentence explaining it. Say which bar you are recommending from, which method does not support it, and what would have to be true for that method to be right. A recommendation that names its own weakest point is far more credible than one where every bar mysteriously agrees, and interviewers who have sat on the other side of that page know it.