Valuation

Trading comps value a business off how similar public companies currently trade: fast to build and grounded in real, live market prices, but only as good as the peer set and only ever a minority, no-control-premium value. Precedent transactions value a business off multiples actually paid in prior M&A deals: they capture a control premium and real acquirer behavior, but are backward-looking and carry deal-specific noise that a mechanical average cannot filter out. A DCF values a business off its own projected cash flows discounted to the present: independent of current market sentiment and directly tied to the specific company's own fundamentals, but highly sensitive to the growth, margin, and discount-rate assumptions that drive it.

No single method is correct in isolation. Standard practice is to build all three, lay them out side by side as a football field, and use the overlap, together with judgment about which method's assumptions are most defensible in the specific situation, to form a final view.

Comps carry the most weight when a genuinely tight set of public peers exists and the goal is understanding current market value, for example, assessing whether a stock is fairly priced right now. Precedents carry the most weight specifically in an M&A context, since they directly reflect what similar businesses have sold for, control premium included. A DCF carries the most weight when a company's own fundamentals and future trajectory are unusually predictable, or when there are simply no good public or transaction comparables at all, for a genuinely unique business, a DCF may be the only method that does not lean on an imperfect proxy.

A DCF becomes much less reliable for early-stage or highly cyclical companies where long-run cash flow projections are largely guesswork; in those cases bankers lean more heavily on comps and precedents, though a DCF is often still shown for completeness.

In practice the three methods rarely agree exactly, and part of what is being tested, in interviews and on the job, is explaining why they differ and what that implies. A DCF value meaningfully below comps and precedents can mean the market, or recent acquirers, are paying for growth or synergies the standalone cash-flow projections have not captured, or it can mean the projections themselves are too conservative. Comps sitting well below precedents is the expected pattern, driven by the control premium, rather than a red flag on its own.

The final valuation range presented to a client is rarely a single blended number, more often a defensible band informed by all three methods, with the banker able to explain in plain language which method's logic is doing the most work in that specific situation and why.

As a general pattern, not a rule that always holds: precedent transactions tend to produce the highest values because they include a control premium and, often, synergies specific to the acquirer. Trading comps tend to produce a lower, minority-stake value, since public trading prices carry no control premium. The DCF can land above, below, or between the other two, since it depends entirely on the growth, margin, and discount-rate assumptions fed into it, which is exactly what makes it the method most sensitive to the analyst's own judgment.

Interviewers ask this specifically to see whether a candidate understands the mechanism behind the ordering, not whether they have memorized a rule. Stating the typical order without being able to explain the control-premium logic behind it does not demonstrate real understanding.

The typical ordering breaks in specific, explainable situations. A DCF with aggressive long-term growth or a discount rate set too low can produce the highest value of the three. Precedents can land below comps if the deals in the screened set happened during a downturn or involved distressed sellers accepting discounted prices. Comps can spike above precedents if the public market is currently pricing in enthusiasm, a sector trading at unusually rich multiples, that has not yet been tested by an actual transaction.

The strong interview answer walks through the specific assumptions or market conditions driving an unusual ordering in the case at hand, rather than defaulting to the textbook pattern regardless of the facts presented.

This ordering has direct practical use in a deal: a sell-side banker representing a company being sold leans on the higher-value method, usually precedents, given the control premium, to support a higher asking price, while a buy-side banker advising an acquirer leans on the lower-value method, usually comps, to argue for paying less. Both sides generally know exactly what the other is doing, which is why a credible negotiating position requires defending the assumptions behind whichever method is being emphasized, not just citing the resulting number.

This is also why a well-prepared banker builds and genuinely understands all three methods even when one is expected to dominate the final number: being able to explain away an inconvenient method, why the DCF looks low, why a specific precedent is an outlier, is part of defending a valuation in front of a skeptical counterparty.

CAPM estimates the return equity investors require: Cost of Equity = Risk-Free Rate + Beta x Equity Risk Premium. The risk-free rate is typically the 10-year US Treasury yield, chosen because its duration roughly matches a typical DCF's projection horizon and because it is liquid enough to give a clean real-time read on long-term risk-free returns. The equity risk premium is the extra return investors demand for holding stocks over risk-free bonds; it is not directly observable and different sources produce somewhat different estimates, so a commonly cited figure like the Damodaran implied US ERP is treated as a reasonable default rather than a single official number.

Beta measures a stock's volatility relative to the overall market: a beta above 1.0 amplifies market moves and demands a higher return; a beta below 1.0 dampens them. Bankers do not typically plug a target's own historical beta into CAPM, since that beta reflects whatever capital structure the company carried historically. Instead, they unlever each comparable peer's observed (levered) beta to strip out that peer's own debt load, take the median across the peer set as a clean read on pure business risk, and relever that median beta at the target's own capital structure: Unlevered Beta = Levered Beta / [1 + (1 - Tax Rate) x (Debt / Equity)], and to go the other direction, Levered Beta = Unlevered Beta x [1 + (1 - Tax Rate) x (Debt / Equity)]. Two companies doing the exact same business but carrying different amounts of debt will show different levered betas purely because debt amplifies equity volatility, but their unlevered betas will be nearly identical, which is exactly the point of the unlever step.

Cost of debt is the rate a company would pay to borrow new money today, not the coupon on debt it issued years ago at a different point in the rate cycle. The cleanest source, when available, is the yield to maturity on the company's existing publicly traded bonds, since a bond's market price already reflects current credit conditions even if its coupon does not. When bond data is not available, bankers fall back on the weighted average coupon across outstanding tranches, or on the risk-free rate plus a credit spread implied by the company's credit rating, which is the standard approach for private companies with no traded debt at all.

Because interest expense is tax-deductible, the after-tax cost of debt, the figure that actually goes into WACC, is Pre-Tax Cost of Debt x (1 - Tax Rate). A company with a 6% pre-tax cost of debt and a 25% tax rate has an after-tax cost of debt of 4.5%; this tax shield, combined with the fact that lenders take on less risk than equity holders, is why debt financing is structurally cheaper than equity.

WACC = (E/V x Cost of Equity) + (D/V x After-Tax Cost of Debt) + (P/V x Cost of Preferred, if applicable), where E, D, and P are market values and V is total capital. Market values, not book values, are essential for the weights: a company's market capitalization can differ enormously from its book equity, and using book values would distort the blend and, by extension, the discount rate applied to every future cash flow in the model. Book value of debt is usually an acceptable stand-in for market value, since most healthy corporate debt trades close to par.

As leverage rises from zero, WACC initially falls, since cheaper after-tax debt is displacing more expensive equity in the blend. Past a certain point, rising leverage starts pushing up both the cost of debt (lenders price in a higher chance of financial distress) and the cost of equity (shareholders demand more for the added risk), and those rising costs eventually overwhelm the cheaper-debt effect, producing a U-shaped WACC curve against leverage with a capital structure at the bottom that minimizes the company's overall cost of capital. WACC is the correct discount rate for unlevered free cash flow precisely because UFCF is the cash available to every one of those capital providers, and WACC is the blended return all of them require together.

Unlevered free cash flow is the cash a business's core operations generate, available to every capital provider before any financing decisions are layered on top. It starts from EBIT rather than net income specifically to exclude the effect of the company's own capital structure: EBIT x (1 - tax rate) gives NOPAT (net operating profit after tax), then D&A is added back since it reduced EBIT without any cash actually leaving the business, then capex is subtracted since that cash genuinely did leave, and finally the change in net working capital is subtracted, since a growing business typically has to fund more inventory and receivables before it collects the cash from customers.

Applying the tax rate directly to EBIT, before any interest deduction, is deliberate: it keeps the resulting cash flow independent of capital structure, since the tax benefit of carrying debt is captured separately, through the after-tax cost of debt inside WACC, rather than inside the cash flow itself. That is exactly why unlevered FCF gets discounted at WACC and produces enterprise value rather than equity value.

Levered free cash flow is what is left specifically for equity holders after debt service. It starts from net income, which already has interest expense subtracted, adds back D&A, subtracts capex and the change in net working capital the same way UFCF does, and additionally subtracts mandatory debt repayments (adding back any new debt drawn, if relevant). Because it already reflects both the cost and the repayment of debt, levered FCF is discounted at the cost of equity, not WACC, and the result is equity value directly rather than enterprise value.

Standard banking DCFs typically build enterprise value through unlevered FCF and WACC rather than levered FCF, but levered FCF is the standard approach in LBO modeling, where the entire point of the analysis is whether the business can generate enough cash after debt service to pay down the acquisition debt on schedule.

Use unlevered FCF discounted at WACC when the goal is enterprise value independent of financing choices, the standard answer to 'walk me through a DCF'. Use levered FCF discounted at the cost of equity when the goal is equity value directly, or when the capital structure itself is central to the analysis rather than incidental to it, as in an LBO.

The most common interview mistake is mismatching cash flow and discount rate, discounting unlevered FCF at the cost of equity or levered FCF at WACC, which either double-counts or ignores the effect of debt financing and produces a nonsensical value. The pairing to have cold: unlevered FCF pairs with WACC and produces enterprise value; levered FCF pairs with the cost of equity and produces equity value.

The Gordon growth method assumes cash flows grow at a constant, modest rate forever after the explicit forecast period ends: TV = Final Year UFCF x (1 + g) / (WACC - g). The growth rate g has to be conservative and sustainable in perpetuity, typically 2% to 3% for a US company, anchored to long-run nominal GDP growth or inflation expectations, since no company can outgrow the entire economy forever without eventually becoming implausibly large relative to it.

The method is extremely sensitive to the spread between WACC and g: as g approaches WACC, the denominator shrinks toward zero and terminal value explodes toward an implausible number, which is why a candidate should always be ready to sanity-check that the assumed growth rate sits comfortably below the discount rate and is economically defensible, not just mathematically convenient. The resulting TV is expressed in final-year dollars and still needs to be discounted back to the present using the same number of periods as the explicit forecast, not one period further, since the formula already folds every future cash flow into a single value sitting at the end of that final year.

The exit multiple method values the terminal year by applying a valuation multiple, usually EV/EBITDA sourced from current trading comps, to the final projected year's metric: TV = Final Year EBITDA x Exit Multiple. This implicitly assumes the company could be sold, at the end of the forecast period, at a multiple similar to how mature comparable companies trade today, which ties the terminal value back to observable market pricing rather than a purely theoretical growth assumption.

The tradeoff runs the other way from Gordon growth: the exit multiple is easier to source and defend, since it comes straight from comps, but it also imports current market sentiment into what is supposed to be an intrinsic, cash-flow-based valuation, and it implicitly assumes market multiples will not shift meaningfully by the time the terminal year actually arrives. In practice this is the method bankers lead with more often, precisely because a market-based multiple is simpler to explain in a client meeting than an abstract long-run growth rate.

Because terminal value often represents 60% to 80% of a DCF's total value, best practice is to run both methods on every DCF and check that they roughly agree. Dividing the Gordon-growth terminal value by final-year EBITDA backs out the exit multiple that growth assumption is implicitly assuming, and that implied multiple should sit reasonably close to where comparable companies actually trade; a Gordon-growth TV that implies a 15x multiple against a peer set trading at 7x to 9x is a clear signal the growth assumption is too aggressive. Running the check the other direction, algebraically solving the exit-multiple TV for the constant growth rate it implies, should land near long-run GDP growth of roughly 2% to 5%; an implied growth rate well above that range signals the chosen exit multiple is too rich.

A large, unresolved divergence between the two methods means one of the two sets of assumptions needs to be revisited before the DCF is presented, since terminal value moves the total valuation more than almost any other single input in the model, which is exactly why it is the first place a DCF gets sensitized.

The standard answer follows a fixed shape: (1) project unlevered free cash flow for an explicit forecast period, typically 5 to 10 years, built from revenue growth and margin assumptions down through EBIT, taxes, D&A, capex, and the change in net working capital; (2) calculate a terminal value for everything beyond that period, using either the Gordon growth or exit multiple method; (3) calculate WACC and use it to discount both the explicit-period cash flows and the terminal value back to present value; (4) sum those present values to arrive at enterprise value; (5) bridge from enterprise value to equity value (subtract net debt, plus preferred and minority interest if present) and divide by diluted shares to get an implied share price.

Delivering this cleanly, in order, without skipping a step and without padding it with unrequested detail, is itself most of what is being tested. Interviewers are checking for a candidate who has genuinely internalized the mechanics well enough to compress them, not one reciting a memorized paragraph at length.

Not every input matters equally: near-term revenue growth and margin assumptions drive the explicit forecast period, but WACC and the terminal value assumption together typically drive the majority of the total valuation, since terminal value alone usually makes up 60% to 80% of the output and a discount rate compounds over every year in the model. A strong answer proactively names this, framing the DCF's output as only as reliable as its WACC and terminal value inputs, rather than presenting it as a precise, objective number.

This observation is also the natural bridge into the standard DCF follow-up questions: an interviewer who hears a candidate flag WACC and terminal value as the highest-leverage assumptions will often go straight into stress-testing exactly those two.

In a live interview, the answer needs to be deliverable in under two minutes without notes, in a clear logical sequence, while staying ready for an interruption at any step, 'why unlevered rather than levered FCF here?' or 'why that discount rate?'. The best way to prepare is not to memorize a script verbatim but to practice explaining each of the five steps in your own words enough times that you can pick the walkthrough back up cleanly after any follow-up, rather than losing the thread.

A candidate who can smoothly restart mid-explanation after a probing question demonstrates real understanding in a way a memorized, uninterruptible recitation cannot, which is exactly why interviewers interrupt in the first place.

The most common DCF follow-up asks the candidate to reason directionally about how a change in an assumption moves the output: a higher WACC lowers enterprise value, since future cash flows are worth less today when discounted at a higher rate; a higher terminal growth rate raises enterprise value, and does so more than proportionally as growth approaches WACC, since the Gordon growth denominator (WACC minus g) shrinks toward zero. A strong answer states the direction confidently and explains why in plain terms rather than hedging or guessing at it.

Interviewers ask this specifically because a candidate who can only recite the DCF mechanically, without an intuitive feel for how the pieces move each other, has not actually internalized the model. Being able to reason about direction without recalculating anything is the real signal of understanding.

A second common line of attack challenges a specific input directly: why that growth rate, why that margin trajectory, why that terminal value method. The strong response defends the assumption with a specific, external reason, comparable companies' historical growth, industry analyst estimates, management guidance, macro context, rather than a circular justification like 'it seemed reasonable'. Equally important is being willing to say an assumption is uncertain and explain what would need to be true for it to be too high or too low, which shows judgment rather than false precision.

This line of questioning is really testing whether a candidate treats DCF inputs as considered judgment calls grounded in evidence, or as arbitrary numbers plugged into a formula, the difference between a real analyst and someone who has only memorized the mechanics.

A well-prepared candidate can name the model's weaknesses unprompted: it is highly sensitive to assumptions that are inherently uncertain, especially for long forecast periods or terminal value; it produces a single point estimate, or a narrow range, from what is genuinely a wide range of plausible outcomes, which can convey false precision; and it is difficult to apply to early-stage or highly cyclical companies where near-term cash flows are volatile or hard to forecast with any confidence, or to businesses like banks where debt is operational rather than a financing choice layered on top.

Naming these limitations proactively, rather than only when pressed, signals that a candidate understands the DCF as one useful tool among several, alongside comps and precedents, not as a black box that outputs 'the' value of a company.

A sensitivity table shows how a DCF's output, typically implied share price or enterprise value, changes as two key inputs vary simultaneously, most commonly WACC and terminal growth rate, or WACC and exit multiple. The output is a grid: one input varies across the rows, the other across the columns, and each cell shows the resulting valuation for that specific combination, letting a banker present a defensible range instead of a single, falsely precise figure.

Building one well in a model requires structuring the core DCF calculation to run off a small number of clearly labeled input cells, rather than values buried or hardcoded inside formulas, so the two chosen inputs can be flexed cleanly across the grid without breaking anything downstream.

Scenario analysis goes a step further than a two-variable sensitivity table, bundling a full, internally consistent set of assumptions, revenue growth, margins, capex, WACC, into named cases (typically base, upside, downside) that reflect coherent stories about how the business might actually perform, rather than varying two inputs in isolation while holding everything else fixed.

A credible downside case is not just 'lower growth' pasted into one cell; it should reflect a specific, articulable reason (a competitive threat, a market slowdown, margin pressure from input costs) and flow consistently through every line that story would actually affect, which is what separates real scenario analysis from an arbitrary haircut applied to the base case.

Standard DCF discounting assumes each year's cash flow arrives as a lump sum at year-end, which understates present value slightly for businesses that actually generate cash more evenly throughout the year. The mid-year convention corrects this by discounting each year's cash flow as though it arrived at the midpoint of that year, using discount periods of 0.5, 1.5, 2.5, and so on instead of 1, 2, 3, which modestly increases the present value of both the explicit-period cash flows and, with a corresponding half-year adjustment, the terminal value itself.

The convention is worth naming in an interview to show precision, but its effect on the final valuation is usually a few percentage points at most, small compared with the swings caused by WACC or terminal growth assumptions, which is exactly why it gets mentioned as a refinement rather than treated as a major driver of the output.

A defensible peer set is screened on five dimensions, not just industry: industry and sub-sector (a high-growth SaaS company is not a peer for a legacy on-premise vendor even if both sit in the same broad software code), geography (companies in the same region share tax regimes and macro conditions, so US targets default to US peers unless the domestic set is too thin), size (screened on revenue or EBITDA rather than EV, since screening on value would artificially narrow the multiples the set produces), financial profile (growth rate and margins, since those are what actually drive the multiple), and business model (a subscription software company and a perpetual-license vendor monetize differently enough that the market prices them differently even in the same sub-sector).

The target is a peer set of five to ten companies: fewer than five and one or two data points swing the median too much; more than ten and the set starts absorbing names that do not really belong, widening the range without adding real signal. Bankers typically keep a tighter core of three to five closest peers that drive the actual range, with the rest providing context.

Once the peer set is chosen, each company's financials get pulled onto a comparable basis. That means computing both LTM (trailing twelve months, the sum of the four most recent reported quarters) and NTM (next twelve months, from consensus estimates) figures for revenue, EBITDA, and net income at minimum, adjusting each company's numbers for one-time items like restructuring charges or legal settlements so the multiple reflects ongoing operating performance rather than a noisy year, and calendarizing every company onto the same fiscal-year timeline so a peer with a March year-end is not compared directly against one with a December year-end without adjustment.

Comps generally lead with the NTM multiple rather than LTM, since a valuation is inherently a forward-looking exercise, what investors are willing to pay today reflects what they expect the business to earn, not what it has already booked. LTM stays alongside as the audited, non-estimate-based backstop, and it becomes the primary basis specifically in one context: LBO entry pricing, where lenders size acquisition debt against results that have already actually happened rather than a set of projections.

Once multiples are calculated across the peer set, bankers record the full spread (minimum, 25th percentile, median, 75th percentile, maximum) and generally anchor on the median rather than the mean, since a single outlier deal or company can drag an average without actually reflecting the peer group as a whole. The target's own implied value comes from applying that range, not a single number, to the target's LTM or NTM metric, which is why comps output is always presented as a range rather than a point estimate.

Comps reflect current public market sentiment, a strength when the market is efficient and a weakness when it is not, and the values they imply are minority-stake, no-control-premium values, since public share prices reflect what it costs to trade a small block of stock, not what a buyer would have to pay to acquire the whole company outright. That gap is exactly why precedent transactions, covered next, typically produce a higher valuation range than trading comps do.

Precedent transactions value a company off multiples paid in prior, completed M&A deals for similar businesses, on the theory that real transactions reveal what acquirers have actually been willing to pay, control premium included. Selection criteria mirror comps (industry, geography, size, deal type) with one extra dimension that matters more here than in comps: recency. A deal struck several years ago reflects a rate environment, financing availability, and competitive backdrop that may no longer hold, so most screens stay inside roughly a 2 to 3 year window and flag or drop anything older, even when it is otherwise a close comparable.

Because acquired targets, especially private ones, rarely have reliable analyst-covered forward estimates, precedent analysis runs almost entirely off LTM figures at the time of announcement, unlike trading comps, which usually lead with forward multiples.

The control premium is the amount an acquirer pays over the target's unaffected share price, the price the day before deal news moved the stock, to take majority ownership: Control Premium = (Offer Price / Unaffected Price) - 1. Typical premiums run 20% to 40% over the unaffected price, and can go higher when multiple bidders are competing for the same target.

Three things drive the premium: synergies the buyer expects to realize post-close that the target could not achieve standing alone; the strategic value of full control, the right to set strategy, replace management, or restructure operations, none of which a minority shareholder can do; and competitive bidding, since even a single credible rival bidder pushes the winning price above the unaffected level. Without some premium, existing shareholders have no reason to sell, since they could already capture the unaffected price simply by selling into the open market.

Precedent multiples carry noise trading comps do not: a strategic acquirer paying up for a rare asset, an auction that pushed price above fair value, or a distressed seller accepting a discount all get baked into a single data point with no way to separate the noise from the underlying valuation logic. Disclosed deal terms and target financials also are not always complete, particularly for private targets, which is part of why precedent screens are usually harder to build than comp sets in the same sector.

Because of this, precedents work best as a check and a bracket around a valuation range built from comps and a DCF, rather than as a standalone answer, and a banker should be able to explain, deal by deal, what drove an outlier multiple rather than treating the whole precedent set as a mechanical average.