Why do you normalize a peer's EBITDA for one-time items before computing its multiple? Give two specific examples of items you'd adjust for.

What the interviewer is actually testing

Comps questions test process discipline more than any single formula, because the entire value of the analysis comes from how carefully it was built.

1. Do you know the five-step process and can you name selection criteria beyond "same industry"? A weak candidate says "you pick companies in the same industry and look at their multiples." A strong candidate names the specific dimensions: business model, size, growth profile, margin structure, geography, and explains that industry label alone is a poor filter (two companies can share a GICS code and have completely different economics).

2. Do you understand why "cleaning" financials matters, and can you give a concrete example? Interviewers love asking "what would you adjust for?" A strong answer names specific normalizations: stripping one-time restructuring charges, litigation settlements, impairments, or gains on asset sales from EBITDA/EBIT/net income, and being consistent about how stock-based compensation is treated across peers (since companies disclose and expense it differently). Being able to say why skipping this step corrupts the multiple (it makes an unusually good or bad year look like ongoing performance) is what separates a mechanical answer from a real one.

3. Can you exercise judgment about which comps to trust, and when the whole methodology breaks down? The advanced layer is judgment: which peers to weight more heavily, how to handle an obvious outlier, and when comps simply don't work (no good public peers, a company deliberately unlike anything else trading, an entire sector temporarily mispriced). A candidate who says "I'd exclude that peer because it's rumored to be an acquisition target and its multiple is inflated by takeover speculation" is showing real judgment, not a formula.

Signals of mastery: naming multiple selection criteria unprompted; giving a concrete example of a cleaning adjustment; distinguishing "closest comps" from the "broader set"; noting that comps and a DCF should corroborate each other, not that one replaces the other. Red flags: describing comps as picking "any company in the same industry," being unable to explain why you'd adjust reported earnings, or presenting a comp median as if it were a precise, unquestionable value.

Common traps

Trap 1: Selecting peers by industry label alone. Two companies can share a sector classification and have completely different business models, margin structures, and growth profiles. Picking peers mechanically off a GICS or SIC code without checking the underlying economics produces a comp set that looks rigorous but isn't.

Say it out loud: "I wouldn't rely on an industry classification alone: I'd check business model, margin structure, growth, and size, because two companies in the same sector code can have very different underlying economics."

Trap 2: Using reported (unadjusted) EBITDA, EBIT, or net income without normalizing for one-time items. A peer that took a large restructuring charge or a litigation settlement this year will show an artificially depressed EBITDA and an artificially inflated multiple, or the reverse, if it booked a one-time gain. Using unadjusted numbers bakes noise into every multiple derived from them.

Say it out loud: "Before computing multiples I'd normalize each peer's earnings for one-time items: restructuring charges, litigation, impairments, gains on divestitures, so the multiples reflect ongoing operating performance, not a single unusual year."

Trap 3: Mixing LTM and NTM across peers, or LTM for some and NTM for others. Comparing one peer's trailing multiple to another's forward multiple isn't a fair comparison: high-growth peers look artificially cheap on NTM versus a slower peer's LTM multiple. The comparison is only valid if every peer uses the same time basis.

Say it out loud: "I'd apply the same time basis (either LTM or NTM) across every peer in the set, since mixing them makes higher-growth companies look artificially cheap relative to slower ones."

Trap 4: Using basic shares instead of fully diluted shares for equity value. This understates every peer's equity value and enterprise value, and therefore every multiple. It's a mechanical error but a common one under time pressure.

Say it out loud: "I'd calculate diluted shares under the treasury stock method for every peer, since using basic shares would understate equity value and every multiple built on it."

Trap 5: Averaging in an obvious outlier without flagging it. A peer trading at a distorted multiple (due to acquisition rumors, a recent scandal, or a temporary earnings collapse) will skew a small peer set's mean if included uncritically. The multiple isn't "wrong" for that company; it's just not representative of normal trading conditions and shouldn't be blended into a "typical" range without comment.

Say it out loud: "If a peer's multiple looks like an outlier (say, inflated by takeover speculation), I'd flag it and consider excluding it from the summary statistics rather than letting it skew the median or mean."

Trap 6: Treating the comp-derived valuation range as more precise than it is. Comps are a market-sentiment snapshot, not a physics measurement. Presenting "the median peer multiple times our EBITDA" as a single precise answer, rather than a range with judgment applied for where the subject company should sit within (or outside) that range, overstates the method's precision.

Say it out loud: "I'd treat the comp output as a range, not a point estimate, and I'd position the subject company within that range based on how its growth and margins compare to the peer set, not just apply the flat median.