Quality of Earnings Before a Sale: What Each Adjustment Must Show

Why an adjusted figure is only readable next to net income

An adjusted figure carries no information by itself. It only means something as a named starting measure plus a specific set of differences applied to it, and "our earnings" is not a named measure. When the SEC staff addresses this for public companies, it is blunt: in its non-GAAP interpretations, "'Earnings' means net income as presented in the statement of operations under GAAP." When a company presents EBIT or EBITDA as a performance measure, the staff says those measures "should be reconciled to net income as presented in the statement of operations under GAAP," and it specifically rules out the more convenient alternative: "Operating income would not be considered the most directly comparable GAAP financial measure because EBIT and EBITDA make adjustments for items that are not included in operating income."

That last sentence explains the whole principle. The adjustments in an EBITDA calculation reach interest, taxes, and other items that sit below operating income. If you start the bridge from operating income, some of those adjustments are adding back things that were never subtracted in the first place, and the reader has no way to see it. The starting measure has to be wide enough to contain every item the differences touch, and net income is the figure that does.

None of this binds you. Regulation G governs public registrants' disclosures, and a private company preparing for sale isn't subject to it. We borrow the reasoning because it's right, and because it names a problem you have regardless of who regulates you: your near end has to be settled before you propose a single difference. That means the legal entity whose result you're starting from, the period the result covers, and the reported measure itself. Change any of those and every row on the bridge silently changes with it, because each difference was measured against a figure that no longer exists. A row that reads correctly against last year's net income for one entity means something else against a consolidated figure or a trailing twelve months, and the row's label won't tell anyone that the ground moved.

Where a company presents seller's discretionary earnings instead, that calculation is a prerequisite handled elsewhere; the point here is only that whatever the starting measure is, it has to be named and fixed.

With the near end fixed, the natural instinct is to present the total distance to the far end: reported figure, adjusted figure, and the difference between them as one number. That instinct is the next problem.

Why each difference belongs on its own line

A net difference tells the person receiving it nothing. It says how far apart two numbers are without saying why, and "why" is the only question anyone reviewing your figures actually has. The staff's standard for public companies puts the point in one clause: the reconciliation from the non-GAAP measure to the comparable GAAP measure "should be in sufficient detail to allow a reader to understand the nature of the reconciling items."

The test isn't that the column foots. A reconciliation can foot with a single line reading "adjustments" and pass arithmetic while failing the test completely. The standard is that a reader who wasn't in the business, who didn't prepare the schedule and can't call you over to explain it, can tell what each line represents.

That standard produces a specific shape. One line per difference. Each line described by what it actually is—the event or obligation behind it—rather than by a category word like "normalization" or "one-time," which describes your conclusion about the item instead of the item. Each amount visible on its own row rather than folded into a neighbor with a related story. Two differences that happen to share a theme are still two events, two periods, and two records, and combining them hides whichever one is weaker behind whichever one is stronger.

Owners often think of a lumped figure as a smaller, more modest disclosure. It's actually the larger ask, because a lumped figure is an unreviewable one. The person receiving it can't ask about a line they can't see, so they're left with two options: accept a number they don't understand, or reject the whole thing and rebuild it from your general ledger. Neither is what you want. Exposing every row lets someone accept the rows that hold while questioning the ones that don't, and that's a better outcome than having the total stand or fall together.

Once a line is exposed, it invites the obvious question about that particular line: what would make this difference defensible?

What makes an adjustment directly attributable and factually supportable

This is the center of the article, and everything after it is a consequence. The clearest language we've found for the test comes from the SEC's Financial Reporting Manual, in the sections governing pro forma information after a business combination. Adjustments there "should give effect to events that are directly attributable to each specific transaction and factually supportable," and "should include those items that have a continuing impact and also those that are nonrecurring." A companion section repeats the phrase almost verbatim: "directly attributable to each specific transaction, factually supportable, and expected to have a continuing impact."

Take the phrase apart, because the two halves guard against two different failures.

Directly attributable means the difference is tied to one identified event, not to a general condition of the business. A row that says a particular obligation ended on a particular date is attributable. A row that says the business "carried costs that a buyer wouldn't" describes a condition, and a condition can't be put on a bridge because there's nothing to point at. The question to ask of each row is whether you can name the event in a sentence with a date in it.

Factually supportable means the row rests on historically determined amounts, on something that already exists in a record, rather than on your estimate of what would have happened. The manual draws this line more sharply than any other source we know. Pro forma information "should illustrate only the isolated and objectively measurable (based on historically determined amounts) effects" of a transaction, "while excluding effects that rely on highly judgmental estimates of how historical management practices and operating decisions may or may not have changed." And it names what the excluded material actually is: information about the expected impact of management's actions, "as if management's actions were carried out in previous reporting periods, is considered a projection."

That word does a lot of work for you. The manual's stated purpose for pro forma information is "showing how a specific transaction or group of transactions might have affected historical financial statements." Showing how something changed a result that already happened is one exercise. Predicting what the business will earn is a different exercise with a different name, and the moment a row depends on "we would have run it differently," it has crossed from the first into the second. There's nothing dishonest about a projection. A buyer may build one. But it belongs in a projection, labeled as such, and it doesn't belong on a bridge whose whole claim is that every row traces to something that occurred.

FASB's framework report gives you the image to test every row against. "Just as a cartographer cannot add roads, bridges, and lakes where none exist, an accountant cannot add imaginary items to financial statements without spoiling the representational faithfulness, and ultimately the usefulness, of the information." A bridge row can only point at something that exists. If the thing it points at is a decision nobody made in the period, the row is drawing a road that isn't there.

Scope, stated once: Article 11 and the manual govern registrant filings for business combinations and don't apply to a private sale. The manual itself notes that smaller reporting companies, to whom the specific rules don't apply, "can consult S-X Article 11 for guidance." That's the spirit in which you're borrowing it—as the clearest available statement of what makes a difference intelligible to someone who didn't prepare it.

Applying this test honestly will kill some of the differences you were planning to propose. Which ones, and why, comes next.

Recurring costs and one-period adjustments that make a figure misleading

Two failures recur in owners' bridges, and neither is an arithmetic error. Both produce a column that foots and a figure that misrepresents the result.

The first is removing a normal cost of operating the business. The SEC staff, again writing for public registrants, says that "certain adjustments may violate Rule 100(b) of Regulation G because they cause the presentation of the non-GAAP measure to be misleading," and it gives one example: "Presenting a non-GAAP performance measure that excludes normal, recurring, cash operating expenses necessary to operate a registrant's business is one example of a measure that could be misleading." That much most owners would agree with in principle. The sentence that surprises them is the staff's definition of recurring: "The staff would view an operating expense that occurs repeatedly or occasionally, including at irregular intervals, as recurring."

Irregular timing, in other words, doesn't make a cost non-recurring. "It didn't happen every year" is not, by itself, an argument. A cost that arrives every few years, or unpredictably, is still a cost of running the business if the business can't run without eventually paying it, and removing it tells the reader the business earns something it doesn't. This is the hardest kind of row to defend no matter how well documented the underlying event is, because the documentation proves the event happened, and the event happening is exactly the point. The record supports the amount. It doesn't support the claim that the business wouldn't face that kind of cost again.

So when you sort your list, the rows that remove a cost the business genuinely needs go into their own pile. You should expect to defend each one, and you should expect to drop some. We'd rather see you drop them yourself than have someone else find them, because a row that fails this test casts doubt on the rows next to it.

The second failure is inconsistency between periods. The staff asks directly whether a measure "can be misleading if it is presented inconsistently between periods," and answers yes: a measure "that adjusts a particular charge or gain in the current period and for which other, similar charges or gains were not also adjusted in prior periods could violate Rule 100(b) of Regulation G unless the change between periods is disclosed and the reasons for it explained."

The discipline here runs backward as well as forward. A row you add for this year raises a question about every similar item in the earlier periods on the same bridge. If you adjusted for a type of charge in the current year and left the same type of charge in the results two years ago, the comparison between the two years is now distorted in your favor, whether or not you intended it. The honest resolution is to treat every period on the bridge the same way, or to disclose that the treatment changed and say why.

Keep the staff's own hedge in view. Whether a presentation is misleading "depends on a company's individual facts and circumstances." This is a way of thinking about your rows, not a verdict on any particular one. Where you want to work through a specific expense, whether a particular expense belongs in an adjustment schedule is decided on its own page.

Having passed that filter, the surviving figure still needs a name, and you're probably about to give it the wrong one.

When your number should not be called EBITDA

The label is part of the representation. The staff's interpretation is short: "Measures that are calculated differently than those described as EBIT and EBITDA in Exchange Act Release No. 47226 should not be characterized as 'EBIT' or 'EBITDA' and their titles should be distinguished from 'EBIT' or 'EBITDA,' such as 'Adjusted EBITDA.'"

This isn't etiquette. A name that overstates what the figure is undoes the work of exposing every line, because the person receiving the figure now believes something different from what the rows show. If your bridge removes items that a standard EBITDA calculation would keep, and you call the result EBITDA, the reader who trusts the label will assume a calculation you didn't perform. The rows say one thing and the title says another.

The practical consequence is that the label has to travel with the starting measure and the reconciliation. The name and the bridge are one disclosure. A distinguishing word—"adjusted," or something more specific to what you did—is a cheap fix for a claim that would otherwise have to be defended, and it costs you nothing you actually have.

Every line now ties, survives the test, and is named correctly. This is where most owners conclude they're finished.

What quality of earnings cannot settle even when the math agrees

A bridge that foots and ties is still not a settled answer, for two reasons that no amount of arithmetic addresses.

The first is omission. Everything so far has been about the rows you wrote. Nothing in a bridge flags the row you didn't write, and FASB's framework report treats that gap as damage to the figure's meaning. Completeness, in the report's words, is "the inclusion in reported information of everything material that is necessary for faithful representation of the relevant phenomena"—in plain terms, that nothing significant has been left out. The report explains why this belongs inside faithfulness rather than beside it: "if financial statements are to faithfully represent an enterprise's financial position and changes in financial position, none of the significant financial functions of the enterprise or its relationships can be lost or distorted." And then the sentence that should unsettle anyone who has just finished a clean schedule: "Relevance of information is adversely affected if a relevant piece of information is omitted, even if the omission does not falsify what is shown."

Read that against your bridge. Every row can be true, every amount can trace to a record, and the figure can still mislead because of what isn't on it. An error announces itself when someone checks the arithmetic. An omission doesn't, because there is nothing wrong to find. Completeness is therefore a question you have to ask deliberately, of your own bridge, rather than one the column answers for you: what happened in these periods that isn't represented here, and would the person receiving this figure want to know?

The second reason is that agreement isn't verification. The framework report says "the reliability of a measure rests on the faithfulness with which it represents what it purports to represent, coupled with an assurance for the user, which comes through verification, that it has that representational quality." Faithfulness is the first half, and it's the half you can work toward yourself. Assurance is the second half, and by the report's own definition it comes through someone else checking. You can't supply that about your own work, however carefully you did it. This is what the outside quality of earnings review adds: the verification a preparer can't give about their own figures. It doesn't replace your bridge. It examines it.

So here is what you can do with your list of proposed differences. Sort it into three piles. The first holds rows you could hand to anyone: a named event, a named period, and something outside the spreadsheet showing it happened. The second holds rows that rest on how the business would have been run, which belong in a projection rather than on a bridge. The third holds rows that remove a cost the business genuinely needs, which you should expect to defend and may have to drop. What survives, named honestly against its starting measure, supports one claim: this is traceable. Buyers, lenders, and appraisers will still apply their own procedures and reach their own conclusions, and nothing on this page says which of your rows any of them will accept or what the resulting figure is worth. That's a limit of the article, not a gap in your bridge. "This is traceable" is a real and defensible thing to say, and a smaller statement than "this is what the business earns"; knowing the size of the claim is most of what quality of earnings means before a sale.

An earnings bridge that holds is the precondition for the next question rather than the end of preparation. Where the customer, contract, billing, and collection support behind your revenue lives is the decision that follows.