Trading Comps Explained: Peer Selection, Calendarization and Normalization (With Worked Examples)

Trading comps look simple. You pick some similar public companies, divide enterprise value by EBITDA, take the median, and you're done.

Except that's exactly where most candidates get caught in interviews. The real skill isn't the division. It's the three steps that happen before it: choosing the right peers, lining up their numbers on the same calendar, and cleaning out the noise. Get those wrong and your multiples are garbage, no matter how nice the spreadsheet looks.

This post walks through all three, with numbers you can actually follow, plus how interviewers tend to test each one.

Quick Refresher: What Trading Comps Actually Do

Trading comparables (or "comps") value a company by looking at how the market currently prices similar public businesses. You calculate valuation multiples for the peer group, then apply a reasonable range of those multiples to your target's metrics.

The most common multiples:

• EV / Revenue: useful for early-stage or unprofitable companies
• EV / EBITDA: the workhorse for most industries
• EV / EBIT: better when capex and D&A differ a lot across peers
• P / E: an equity value multiple, common for banks, insurers and mature businesses

One rule to keep in your back pocket: match the numerator to the denominator. Enterprise value goes with metrics available to all capital providers (revenue, EBITDA, EBIT). Equity value goes with metrics available only to shareholders (net income, EPS). Mixing them is a classic interview trap.

Now, onto the parts that separate a good comps sheet from a bad one.

Step 1: Peer Selection

Why this step matters most

Your output can only be as good as your peer set. If you compare a slow-growing industrial distributor to a high-growth software company, the multiple you get tells you nothing useful. Peer selection is a judgment call, and that's precisely why interviewers love asking about it.

The criteria bankers actually use

Think of it as a funnel. You start broad and narrow down based on how closely each company resembles your target.

Business characteristics (the most important filter):

• Industry and sub-sector: what the company actually sells
• Business model: recurring subscription revenue vs. one-off sales, asset-light vs. asset-heavy
• End markets and customers: selling to consumers vs. enterprises vs. governments
• Geography: where revenue is generated, not just where the company is headquartered

Financial characteristics:

• Size: revenue, EBITDA or market cap in a similar range
• Growth: similar expected revenue growth rates
• Profitability: comparable margin profiles
• Capital intensity and leverage: similar capex needs and balance sheet structure

A worked example of the thinking

Say your target is a mid-sized US company that sells payroll software to small businesses on a subscription basis. Here's how you'd reason through potential peers:

• A large HR and payroll software company selling to enterprises: good business match, but much bigger and possibly different growth. Keep it, maybe in a secondary tier.
• A small-business accounting software company: different product, very similar customer and model. Strong candidate.
• A payroll outsourcing firm that mostly sells human-delivered services: same end function, very different model and margins. Probably exclude, or flag it.
• A generic IT consulting firm: wrong model entirely. Exclude.

Use tiers, not a single list

When perfect peers don't exist (they rarely do), group your comps into tiers:

• Tier 1: closest matches on business model and end market
• Tier 2: relevant but imperfect, like bigger players or adjacent products

You can then weight your judgment toward Tier 1 when picking your final multiple range. This also gives you a great answer in interviews when someone asks, "What if you can't find good comps?"

Common peer selection mistakes

• Picking too few companies. Two or three peers make your median fragile. One weird multiple swings everything.
• Picking too many. Throwing in 25 loosely related names dilutes the signal.
• Ignoring outliers. A company trading at 60x EBITDA because of a takeover rumor shouldn't anchor your range. Note it, then consider excluding it.
• Matching on SIC code alone. Industry codes are a starting point, not an answer. Read the business descriptions.

Step 2: Calendarization

The problem

Not every company ends its fiscal year on December 31. Plenty of retailers end in late January or early February, many tech companies end in June or September, and lots of companies outside the US end in March.

If you compare one company's fiscal year ending March 2025 with another's ending December 2024, you're comparing different time periods. In a fast-changing environment, that mismatch can meaningfully distort your multiples.

Calendarization fixes this by converting everyone's financials to a common period, usually the calendar year ending December 31.

The formula

For a company with a fiscal year ending in month M, the calendar year figure blends two fiscal years:

Calendar Year = (M / 12) × Fiscal Year ending in that calendar year + ((12 − M) / 12) × Next Fiscal Year

It looks scarier than it is. Let's run it.

Worked example: March fiscal year end

Your peer's fiscal year ends March 31. You need its calendar year 2024 EBITDA.

• FY ending March 2024 EBITDA: $400m
• FY ending March 2025 EBITDA: $480m

Calendar 2024 runs January to December 2024. Here's where those months sit:

• January to March 2024 (3 months) fall in the FY ending March 2024
• April to December 2024 (9 months) fall in the FY ending March 2025

So:

• 3/12 × $400m = $100m
• 9/12 × $480m = $360m
• Calendar 2024 EBITDA = $460m

Notice how different that is from just grabbing either fiscal year. Use $400m and you'd overstate the multiple. Use $480m and you'd understate it.

Calendarizing forward estimates

The same logic applies to projections. If you're building CY2025E multiples, you'd blend the consensus estimates for the fiscal years that overlap calendar 2025. Forward multiples are often what bankers focus on, so this comes up constantly.

A note on precision

This method assumes earnings are spread evenly across the year. That's an approximation. For highly seasonal businesses, like a retailer that earns a big chunk of its profit in the holiday quarter, you'd ideally build the calendar year from actual quarterly figures instead. Mentioning this in an interview shows you understand the tool's limits.

Bonus: calculating LTM

Calendarization's cousin is LTM (Last Twelve Months), which gives you the most recent full year of performance:

LTM = Most recent fiscal year + Current year-to-date − Prior year-to-date

Example: FY2024 revenue of $1,000m, nine months of 2025 at $810m, and nine months of 2024 at $750m. LTM revenue = $1,000m + $810m − $750m = $1,060m.

Step 3: Normalization

The problem

Reported numbers are full of things that have nothing to do with how the business normally performs. A restructuring charge, a lawsuit, a gain from selling a building. If you leave those in, you're valuing a company based on a strange year rather than its true earning power.

Normalization means adjusting financials to strip out non-recurring and non-operating items so every company in your set is measured on the same clean basis.

What typically gets adjusted

Usually added back (they depress reported earnings):

• Restructuring and severance charges
• Asset impairments and write-downs
• Litigation settlements
• Transaction costs from M&A deals

Usually subtracted (they inflate reported earnings):

• Gains on sale of assets or businesses
• One-time insurance proceeds
• Other non-recurring income

The key test: would this happen again in a normal year? If not, it probably gets adjusted. But be careful. A company that "restructures" every single year doesn't really have a one-time charge.

Worked example: adjusting EBITDA

A peer reports EBITDA of $250m. Digging through the filings, you find:

• $30m restructuring charge → add back
• $10m litigation settlement → add back
• $15m gain on sale of a building → subtract

Adjusted EBITDA = $250m + $30m + $10m − $15m = $275m

Now see what it does to the multiple. If the company's enterprise value is $3,000m:

• On reported EBITDA: $3,000m / $250m = 12.0x
• On adjusted EBITDA: $3,000m / $275m = 10.9x

That's more than a full turn of difference. On a target with $200m of EBITDA, one turn is $200m of valuation. Normalization isn't a rounding exercise.

Don't forget taxes when adjusting net income

When you normalize net income or EPS for a P/E multiple, adjust on an after-tax basis. Interest, taxes and other below-the-line items don't affect EBITDA, but they do affect net income.

If you're adding back that $30m restructuring charge and the tax rate is 25%:

• After-tax add-back = $30m × (1 − 25%) = $22.5m

Adding back the full $30m to net income would overstate earnings, because the charge also reduced the company's tax bill.

Normalizing for accounting differences

Clean adjustments aren't only about one-time items. Different accounting rules can make identical businesses look different:

• Leases (IFRS 16 vs. US GAAP): Under IFRS, most lease costs move out of operating expenses and into depreciation and interest, which pushes EBITDA up. Under US GAAP, operating lease costs generally stay in operating expenses. If your peer set mixes both, you need a consistent treatment, for example by including lease liabilities in enterprise value only when EBITDA excludes lease costs.
• Stock-based compensation: Some companies' "adjusted EBITDA" adds back SBC, others don't. Pick one approach and apply it to everyone.
• Company-defined "adjusted" metrics: Management's adjusted figures can be generous. Don't copy them blindly. Rebuild the adjustments yourself from the filings so they're consistent across peers.

Consistency beats perfection. The goal is apples to apples, not the "right" number for any single company in isolation.

How Interviewers Test This

Expect questions like these. Here's the gist of strong answers:

"How would you pick comparable companies?" Start with business characteristics (industry, model, end markets, geography), then filter on financials (size, growth, margins). Mention tiers if close peers are scarce.

"Why do you calendarize?" Because companies have different fiscal year ends, and you need everyone on the same time period so the multiples are actually comparable.

"A company's fiscal year ends in June. How do you get calendar year figures?" Take 6/12 of the fiscal year ending June of that calendar year plus 6/12 of the following fiscal year. Bonus points for flagging seasonality.

"What adjustments would you make to EBITDA?" Add back non-recurring charges like restructuring, impairments and litigation. Subtract one-time gains like asset sale proceeds. Apply the same logic to all peers.

"Why might two companies in the same industry trade at very different multiples?" Different growth expectations, margins, risk, size, or simply unadjusted one-time items distorting the numbers. That last one ties straight back to normalization.

The Takeaway

The math in trading comps is easy. The judgment is where the value lives.

• Peer selection decides whether your multiples mean anything at all.
• Calendarization makes sure you're comparing the same slice of time.
• Normalization makes sure you're comparing true earning power, not one-off noise.

If you can explain all three clearly and walk through a quick example of each, you'll sound like someone who has actually built a comps sheet, not someone who memorized a definition. And in a superday, that difference shows.