Skip to main content

Normalizing for Comparability

Updated Jun 24, 2026 at 2:35pm

Research Draft Medium 1,248 words

Comparable company analysis (comps) prices a target off the multiples — EV/EBITDA, EV/Sales, P/E — at which similar public companies trade. But a multiple is only meaningful if the numerator (price/value) and denominator (the operating metric) are measured the same way for every company in the set. Normalizing for comparability is the "scrubbing" work that makes a peer table apples-to-apples: stripping out one-time noise, aligning time periods, and harmonizing accounting and capital-structure differences so that a 9x multiple on Company A means the same thing as a 9x on Company B. The core tension is that every normalization is a judgment call — and the analyst who controls the add-backs effectively controls the answer.

What gets normalized

Normalization works on four distinct axes:

1. Non-recurring / one-time items. Reported earnings carry noise that won't repeat: restructuring charges, litigation settlements, asset write-downs and impairments, gains/losses on asset sales, acquisition and integration costs, and genuinely unusual events (Wall Street Prep's examples include hurricane repair costs and one-time deal-integration costs). These are added back or removed to isolate run-rate, sustainable operating performance. In private-deal and middle-market contexts the same exercise extends to owner-specific add-backs — excess owner compensation, personal expenses run through the business — producing "Normalized" or "Adjusted EBITDA."

2. Time-period alignment. Two sub-steps:

  • LTM (last twelve months) rolls reporting periods forward so every company reflects twelve months of actual data rather than a stale fiscal year. Standard formula: LTM = Most-recent full fiscal year + current stub period − prior-year stub period. (Example from ibinterviewquestions.com: FY2024 $400M + Q1–Q3 2025 $320M − Q1–Q3 2024 $280M = LTM EBITDA $440M.)
  • Calendarization aligns peers with different fiscal year-ends to a common calendar period by pro-rating estimates: Calendarized CY = (% of Period A × FY-A estimate) + (% of Period B × FY-B estimate). Without it, a June-year retailer and a December-year retailer are being compared across genuinely different economic windows — a real problem in retail, tech, and global names.

3. Capital structure. Enterprise-value multiples (EV/EBITDA, EV/Sales, EV/EBIT) are used precisely because they neutralize differences in leverage and tax position, letting you compare an unlevered operating result. Equity multiples (P/E) do not, so they require more care when peers carry very different debt loads. EV itself must be built consistently — fully diluted shares (treasury-stock method for options/warrants), net debt, preferred, and minority interest treated the same way across the set.

4. Accounting policy. Differences in revenue recognition, depreciation method/useful lives, inventory (LIFO vs. FIFO), lease capitalization, and especially stock-based compensation (SBC) treatment can make identical businesses look different. The numerator and denominator must also be consistent — diluted-share, balance-sheet, and income-statement conventions matched throughout.

How it's used in practice

Analysts build the comps table, then "scrub" each company's filings to a normalized metric before computing multiples. The practical workflow: pull reported figures, layer in management-disclosed and analyst-identified adjustments, convert to LTM, calendarize to a common period, and recompute EV and the multiple. The dominant convention is to present both a reported and an adjusted figure so the reader can see the size of the bridge.

In M&A, normalized EBITDA directly sets the price — most private-company deals are valued as a multiple of it, so a dollar of defensible add-back is worth a multiple's-worth of enterprise value. This is exactly why buyers commission a Quality of Earnings (QoE) report: an independent accounting-diligence test of whether each adjustment is real, recurring-adjusted correctly, and supported by the books. The QoE is the institutional check on normalization abuse.

Adoption, debate & evidence

Normalization is universal and uncontested as a principle — every valuation textbook (Damodaran), every bank's comps process, and every QoE practice treats it as mandatory. The debate is entirely about which adjustments are legitimate.

The marquee controversy is stock-based compensation. The "add it back" camp treats SBC as non-cash, like D&A, with dilution captured separately in share count; this dominates tech/SaaS peer comparisons. The "keep it in" camp — including Warren Buffett, who in his 2018 shareholder letter said managements that assert SBC shouldn't count as an expense should ask "What else could it be — a gift from shareholders?", and Charlie Munger, who called adjusted-EBITDA earnings "bulls--t earnings" largely because of SBC — argues it is a genuine economic cost, often converted to cash via buybacks to offset dilution. The stakes are large: practitioner sources put SBC add-backs at roughly 15–30% of reported EBITDA for tech/SaaS names, and a Morgan Stanley Counterpoint Global analysis found SBC equal to a much larger share of free cash flow for large-cap software than for the broad market — the precise magnitude is firm-specific and disputed. One practitioner source (ibinterviewquestions.com) cites an illustrative case where adding SBC back moved a multiple from 15x to 8.3x — making a company look meaningfully cheaper with no change in economics. That is the empirical core of the warning: normalization can manufacture comparability that doesn't exist.

The honest base-rate framing: there is no academic study showing "normalized comps predict returns better than raw comps." Normalization improves internal consistency of a peer table; it does not confer a forecasting edge. Its value is hygiene, not alpha. And "Adjusted EBITDA" is the most-abused number in finance — the SEC's Non-GAAP guidance (Reg G / Compliance & Disclosure Interpretations) exists specifically because issuers use add-backs to flatter results.

Strengths & limitations

Works when: adjustments are few, disclosed, and genuinely non-recurring; peers are close substitutes; and you show the reported-to-adjusted bridge. Calendarization and LTM are nearly always net-positive — they fix mechanical period mismatches with little subjectivity.

Fails when: "non-recurring" items recur every year (the classic tell — restructuring charges that appear in five consecutive 10-Ks are operating costs); when SBC is silently excluded from EBITDA but its dilution is also under-counted in shares; or when normalization is applied to the target but not the peers (or vice versa), reintroducing the very inconsistency it was meant to remove.

The #1 misuse: treating add-backs as a lever to hit a desired valuation. Because normalized EBITDA flows straight into price through the multiple, an aggressive add-back schedule is a powerful and easily-disguised way to inflate value. The discipline is to normalize symmetrically across target and peers, and to demand evidence for every adjustment.

Sources

Disputes flagged: SBC add-back treatment is genuinely contested (tech-banking convention vs. Buffett/Munger and GAAP-purist view); the doc takes the skeptical position. The 15–30% SBC magnitude and the 15x→8.3x example come from a single practitioner source (ibinterviewquestions.com) and are presented as illustrative, not as a measured population statistic. The Buffett quote is verbatim from his 2018 letter; the earlier draft's "If compensation isn't an expense, what is it?" was a paraphrase and has been corrected.