GAAP vs. Adjusted Earnings

Two numbers, one company

Every public company in the U.S. files financial statements under Generally Accepted Accounting Principles. GAAP is the rulebook — standardized, enforced by the SEC, and intentionally conservative. It captures the write-down when a division fails, the charge when a lawsuit settles, the expense when employees receive stock grants. Whatever actually happened to the business that quarter shows up in the GAAP numbers, flattering or not.

Adjusted earnings are something companies calculate voluntarily, under no standardized rules, with full authority to decide what to remove. The rationale is defensible in theory: one-time charges — a factory fire, a legal settlement, an acquisition write-down — can distort the picture of how a business performs in a typical period. Strip those out and you might see a cleaner underlying trend.

The problem is that "one-time" is a self-declaration. No regulator approves the classification. No auditor certifies the adjusted figure. A company can strip out any cost it wants, label it non-recurring, and present the result in its earnings release as the headline number most investors see first.

The gap between GAAP profit and adjusted profit is where the real research starts. A company reporting $200M in adjusted operating income and $80M in GAAP operating income isn't presenting two views of the same reality — it's making a $120M argument about what you should ignore. That argument may be correct. Accepting it without reading the reconciliation is skipping the most important part of earnings analysis.

What actually gets removed

Stock-based compensation is the largest single adjustment for most technology and growth-stage companies. GAAP treats it as a real expense because employees receiving stock grants dilute every other shareholder — that dilution is a cost whether or not cash leaves the bank. Companies argue it should be excluded from adjusted earnings because it's non-cash and doesn't affect operational funding or debt service capacity. The counter-argument is simple: if every grant were paid in cash instead, the company couldn't sustain its workforce at current scale. Salesforce excluded more than $4 billion in stock-based compensation from its non-GAAP results in fiscal 2024 — enough to convert what would otherwise be a sizable GAAP operating loss into a profitable adjusted result.

Restructuring charges appear when companies announce layoffs, shutter offices, or wind down product lines. GAAP hits the income statement immediately: severance costs, lease termination penalties, asset write-downs. The company argues these are episodic costs of rationalizing the organization, not the ongoing cost of running it. The challenge is frequency. Restructuring charges appearing in Q4 of five consecutive years — sometimes labeled "workforce optimization," sometimes "cost transformation program," sometimes "strategic realignment" — are not episodic by any reasonable definition.

Acquisition-related amortization has the strongest accounting justification of any common adjustment. When a company acquires another, GAAP requires it to capitalize the acquired intangible assets — customer relationships, patents, trade names, developed technology — and amortize them over estimated useful lives, often 10 to 15 years. That amortization flows through the income statement every quarter as an expense, even though the underlying assets aren't degrading the way equipment does. The cash left at acquisition. The ongoing charge is a schedule, not a recurring drain. Many analysts accept this adjustment because the economic argument genuinely holds.

Asset impairments are write-downs recorded when an asset's carrying value exceeds what it's actually worth. Goodwill impairments — arising when a company overpaid for an acquisition — can be enormous. General Electric recorded a $22 billion goodwill impairment on its power segment in 2018, reflecting an economic loss that had accumulated over years of competitive deterioration. Excluding that charge from an ongoing earnings picture is defensible; the loss was real but historical. The risk is distinguishing a genuine accounting clean-up from a management team using an impairment to reset the earnings baseline and then grow against an easier comparison.

The adjustment audit

The reconciliation table is called the "Reconciliation of GAAP to Non-GAAP Measures." It lives near the bottom of the quarterly earnings press release — usually after the income statement, after the balance sheet, after the segment breakdown. It's also in the 10-Q and 10-K footnotes, though the press release version tends to be formatted more clearly. If a company reports adjusted figures but publishes no reconciliation, stop there. That omission is itself a red flag.

The structure is typically three columns: the GAAP line item, the individual adjustment items (often with sub-lines for each category), and the resulting non-GAAP figure. Reading it line by line is the actual skill. You're tracking exactly what management decided to add back or subtract — and in what dollar amounts — to travel from the reported GAAP result to the adjusted number in the headline.

Two questions resolve most of the calls. First: is this item genuinely non-recurring? Not unusual, not large — non-recurring. A write-down of a joint venture terminated after a decade qualifies. A "transition and integration" charge from an acquisition that closed three years ago does not. Second: does it represent a real economic cost to shareholders, even if no cash changed hands this quarter? This is where stock-based compensation keeps coming back. Issuing shares to employees is real dilution. The adjusted earnings figure that excludes SBC is higher, but shareholders own a smaller percentage of the company's future earnings than they did before the grants. That cost exists whether or not it appears in the adjusted P&L. See our page on stock-based compensation and dilution for how to model it separately.

Materiality matters here too. A $4M adjustment in a company with $900M in revenue is background noise. A $180M adjustment in a company with $400M in revenue is the story. When a single adjustment line is large enough to swing the company from a GAAP loss to an adjusted profit, that item deserves a dedicated paragraph in your analysis, not a passing mention.

Never examine a single quarter's reconciliation in isolation. The table shows what management removed this period. It doesn't show whether that category has appeared in every period for four consecutive years. The trend across multiple reconciliation tables is almost always more informative than any single result. Pull the last eight quarters and lay them side by side. What keeps showing up?

When adjusted earnings are actually more useful

The acquisition amortization case is the clearest example of a GAAP distortion that adjusted earnings legitimately correct. Broadcom provides a useful illustration. After its string of major acquisitions — CA Technologies, Symantec's enterprise security business, VMware — Broadcom carries tens of billions of dollars in acquired intangible assets that it amortizes every quarter under GAAP. That amortization hits the income statement as an expense regardless of whether the acquired technology is growing or shrinking. An analyst modeling Broadcom's capacity to service its debt, pay its dividend, or fund the next deal would get a distorted picture from GAAP earnings alone. The cash earnings power of the business is materially higher than the GAAP net income figure implies, and the difference is almost entirely accounting amortization from prior transactions.

Young software companies with heavy equity grants present a different legitimate case. In the early years of a SaaS business — before revenue outgrows the fixed cost base — GAAP losses are often dominated by stock-based compensation. If the question you're trying to answer is whether the unit economics work at all, adjusted operating income strips out the SBC and lets you see whether customer acquisition costs recover in a reasonable period, whether gross margins support a viable model at scale, whether the business would be profitable if it stopped growing as aggressively. That's a real analytical question. The correct approach is to track the SBC separately — monitoring dilution per share over time — rather than treating it as invisible.

A practical test for any adjustment you're considering accepting: would a sophisticated buyer of the whole business care about this? An acquirer evaluating Broadcom's earnings power would adjust out intangible amortization from the assets they just purchased — the cash is already spent. But that same acquirer evaluating a company with five years of restructuring charges would price those charges into the business, not wave them away. If an adjustment fails that test, it belongs in the earnings figure you use to value the company.

The broader principle is that some adjusted figures genuinely correct for accounting conventions that don't track economic reality well. The issue isn't that non-GAAP metrics exist — it's that companies exploit the legitimate cases to build credibility, then extend the same treatment to items that don't qualify. Understanding where adjusted earnings are actually useful makes it easier to spot the line being crossed.

Serial adjusters and the multi-year test

Pull four or five years of both GAAP and adjusted earnings for any company you're studying. Build a simple table: fiscal year, GAAP net income, adjusted net income, the spread, and the spread as a percentage of adjusted earnings. The figures themselves matter less than what the spread does over time. If the gap is roughly constant — adjusted earnings running $300M to $400M above GAAP every year for five years — then the items management classifies as non-recurring are, by definition, recurring. The label is wrong even if each individual charge has a new name.

Restructuring charges appearing every Q4 for five or six years are not exceptional costs of rationalizing the organization. They are an operating cost of running it. The management team either cannot build a stable structure, or is using the annual charge as a mechanism to reclassify real operating expenses into the adjusted-out column. Either conclusion produces the same response: those charges belong in the earnings figure you use to value the stock.

"Transition and integration costs" from acquisitions follow a similar pattern. One year post-close, integration costs are plausible. Two years is a stretch. Three years means the integration is not a temporary project — it's the operating model. Several large industrial and technology mergers have run "acquisition-related integration costs" for four and five consecutive years, each year's charge appearing in the reconciliation as a line item, each year labeled exceptional. Summed across five years, the cumulative excluded costs in some cases exceed the original acquisition premium paid.

The multi-year test works in the other direction too. A company where the GAAP-to-adjusted gap is clearly narrowing — where the one-time charges are actually going away — is demonstrating that the adjustments were real and temporary. That's the reconciliation doing what it's supposed to do. Finding that convergence pattern is as informative as finding the divergence pattern, just in the opposite direction.

None of this requires specialized data. Annual reports are public. Building the comparison table takes twenty minutes using the press releases. The question it answers — whether the spread between GAAP and adjusted earnings is shrinking, stable, or widening — is usually worth more than anything in the earnings call transcript.