In a data room, encountering the same EBITDA figure at three different values across three different documents is not an anomaly in mid-market companies but a recurring pattern: the management presentation carries one number, the summary schedule prepared for a lending application carries another, and the figure derived from audited statements produces a third. The divergence rarely originates in accounting error; it originates in the fact that the three documents were built for three purposes at three separate moments, each with a different audience in mind. A bridge between the numbers can usually be constructed, and yet the document that constructs it is typically absent — what exists instead is a person who can construct it on request. The question posed at the review table is therefore not which of the three figures is correct, but whether a written answer exists for why the difference arose at all.
The second and more determinative observation concerns where the calculation actually lives inside the company. Asked which account codes feed the figure, many companies answer by pointing not to a policy but to a working file on a single machine; the file runs, the formulas are sound, the logic is internally consistent, and yet the logic itself is recorded nowhere. Put the same question separately to the finance director, to the external accountant, and to the general manager, and three answers emerge that are adjacent without being identical, diverging most predictably on rent, management fees, consultancy charges, and the items characterized as non-recurring. That divergence is among the first findings a reviewing party records, since a metric carrying definitional ambiguity transmits that ambiguity to every covenant, every earn-out, and every price mechanism resting upon it.
The mechanics of this condition are inherent to the measure itself. EBITDA is not a line item defined in any reporting standard; it is a bridge measure constructed to strip out financing structure, tax regime, and depreciation policy so that businesses carrying different capital structures may be compared, and its definition arrives from contract and internal policy rather than from a standard-setter. That vacancy invites each user to build the metric toward its own purpose: the lender drafts a narrow definition for covenant testing, the sell side constructs a broader one capable of carrying the equity story, and management incentives attach to a third. Rebuilding the definition on each occasion is a rational shortcut to the extent that it lowers the immediate cost of producing a number; the difficulty lies not in the shortcut but in its persistence once conditions change — that is, once a third party begins to see the entire set of documents at once.
Add-backs are where this mechanism generates the densest friction. Treating an expense as non-recurring is not a statement of historical fact but an assertion about the future: this item will not arise again. Unless the evidence supporting that assertion is assembled at the moment the event occurs, it cannot be manufactured retroactively two years later; where the termination correspondence, the settlement decision, and the scope document for a one-time advisory engagement were never filed at the time, the item ceases to be defensible irrespective of its underlying merits. Items such as founder personal expenses, below-market related-party rent, and normalized management compensation present a further difficulty, in that even where each is individually defensible, their aggregation produces a concentration signal that a reviewer weighs separately from the merits of any single line.
The implementation dimension reveals whether the metric lives in the reporting pack or in the operation. Where EBITDA is computed only at year end or when a financing need arises, the measure functions as a presentation instrument rather than a management instrument, and pricing decisions, supplier terms, headcount changes, and commissioning schedules are being taken against cash position or revenue instead. In companies where the metric genuinely governs, the distinguishing markers are consistent: the calculation is produced at each monthly close and at segment level, and variances are carried into the following period's agenda with a written explanation attached. The account code and cost center selected at the moment a journal entry is posted determine whether an adjustment will survive scrutiny two years later, which makes implementation discipline, in substance, the construction of future negotiating leverage in the present.
The first channel through which the institutional cost travels is arithmetic, and it operates quickly. Where valuation is built on a multiple, every disputed unit of EBITDA reaches price magnified by that multiple, so the effect of a rejected adjustment is not its own magnitude but a multiple of it. The second and heavier channel is the credibility discount: when a material share of the presented adjustments cannot be evidenced, the counterparty typically abandons line-by-line negotiation and applies a blanket reduction in confidence across the entire number set, on the reasonable premise that as the count of unverifiable items rises, the care with which the verifiable items were produced becomes a legitimate question in itself. Of the two, the second is the harder to recover in negotiation, attaching as it does to the company's reporting rather than to any particular figure.
The third channel appears in transaction structure, and it tends to move before price does. An EBITDA whose definition has not settled makes the structure heavier rather than the number smaller: additional verification items appear among the conditions precedent, the escrow percentage rises, the representation and warranty package broadens under the heading addressing the accuracy of financial information, and a portion of the consideration migrates into an earn-out. Since the earn-out is itself measured against the same metric, the parties end by committing a future payment to a measure they have never defined identically; post-closing accounting policy changes, allocated corporate overhead, and new cost items imposed by the buyer will, predictably, generate dispute wherever the definition was left unwritten. The working capital peg draws on the same foundation, with the result that definitional ambiguity is invoiced a second time in the post-closing adjustment.
The fourth channel sits on the credit side and frequently operates independently of any transaction. The EBITDA definition in a credit agreement constrains permitted adjustments, the treatment of lease expense, and the pro forma contribution of acquired businesses in explicit terms; where the definition used in management reporting exceeds those constraints, the headroom visible under the covenant heading is apparent rather than real. The ownership dimension becomes decisive at precisely this point: if the record does not state who sets the definition, what approval its amendment requires, and to whom deviations are reported, the definition drifts quietly, and the drift is generally detected on a covenant test date, by which time the room to correct it has already closed. The common outcome of unowned metrics is that responsibility reverts in practice to the founder, and founder dependency is priced as a valuation discount.
The structure that neutralizes this tendency is architectural rather than attentional, and it rests on four separable components. The first is a single EBITDA definition document, approved by the board or an equivalent body, carrying the mapping from chart-of-accounts codes to the metric and subjecting any amendment to the same approval threshold as an accounting policy change, together with an obligation to restate comparative periods. The second is an adjustments register in which each item is recorded in the period it arises rather than at the point of sale, with its supporting evidence attached at that moment. The third is a bridge from statutory financial statements to management EBITDA maintained as a continuous record rather than reconstructed from scratch each month. The fourth is a named owner holding decision authority over the definition, paired with a fixed review cadence at which variances are examined.
BEIREK's intervention in this area begins not with recomputing the figure but with institutionalizing the chain that produces it. We draft the definition document at account-code level, convert the bridge from statutory result to management measure into a standing output of the monthly close, and operate the adjustments register as a record that closes in the month the item originates, with its evidence attached — which largely removes the need for retrospective defense two years later. On engagements close to a transaction, we run a pre-mortem on each adjustment: which counterparty would reject this item, on what stated ground, and what the price consequence of rejection would be, all answered in writing before the item is ever presented.
The measurement and continuity dimensions constitute the testing side of the same architecture. Reporting the metric at segment, product line, or facility level exposes the dispersion that a single consolidated figure conceals and renders forecast accuracy assessable to a reviewer; a series in which variances are tracked with written explanations is a considerably stronger indicator of management quality than any single annual number. The continuity test reduces to one practical question: when the person who currently builds the calculation leaves the organization, can the same figure be reproduced from the same sources, within a reasonable period, and with the same result? Where the answer rests on documents, the metric is an institutional capability; where it rests on an individual, it will not be priced as one.
What an investment review seeks under the EBITDA heading is ultimately not how much the company earns but how consistently it can explain its earnings in a language a third party is able to audit; the two questions are measured by the same figure and remain distinct questions. The multiple a company receives is determined, more often than by performance itself, by the demonstrability of that performance as something explicable and repeatable independently of the founder.