Where the weight of discussion settles during an investment committee session tends to be more informative than the content of the presentation itself; the pattern typically observed is that the bulk of the session is spent on the EBITDA line and its year-over-year growth rate, while the replacement capital figure sits in the appendix as a single-line percentage assumption. In that same session the refinancing date of the debt, together with the interest terms likely to prevail on that date, is generally treated as a separate agenda item belonging to the financing team, so that the table debating valuation and the table debating debt service look at two faces of the same asset without either seeing the other. The decision is then taken not on the sum of the two faces but on the one that appears higher and cleaner. This is less a failure of attention than a consequence of how the architecture of the presentation directs attention.
The same pattern is quieter inside the annual budget cycle. When a scheduled overhaul on a production line is postponed by a year, current-period EBITDA rises, and that rise is reported as operational improvement; the cost of the deferral appears later, in a future maintenance budget, in downtime hours attributable to unplanned failure, or in the technical condition report produced when the asset is eventually brought to market. To the extent that incentive and bonus mechanisms are tied to EBITDA, the system rewards the deferral not once but in every subsequent budget cycle. Measurement and incentive intersecting on the same adjusted figure determines not the preference of a single manager but the default behavior of the institution.
The name that has attached itself to this behavior is EBITDA blindness — the exclusion of interest, taxes and capital investment requirements from the performance discussion — and the point worth holding onto is that it constitutes not an accounting error but the use of an instrument outside the purpose for which it was designed. EBITDA was developed as a normalizing device, one that allows assets with different capital structures, different tax regimes and different depreciation policies to be placed side by side on a comparable scale; where the operating performance of a levered asset must be compared against that of an unlevered one, this normalization genuinely produces information. That is the condition under which the shortcut lowers cost, and under that condition using it is entirely rational.
The difficulty begins when the shortcut persists after the condition that justified it has dissolved. The three excluded items do not belong to a single category: interest is a cash outflow tied to capital structure and repriced at each refinancing date; tax is a cash outflow that jumps between years depending on incentive regimes and the depreciation shield; depreciation, by contrast, is not a cash outflow at all, though the replacement investment it stands for certainly is. When reading habits collapse these three distinct items into a single cash proxy, three separate time horizons are compressed into one annual figure. In a capital-intensive asset this compression is unusually costly, since the physical integrity of the asset is a precondition of the revenue itself; maintenance capital is not a discretionary growth expenditure but the condition under which the revenue stream continues.
Three thresholds indicate when the underlying condition has shifted, and all three are observable. The first is a shortening of average debt maturity or the approach of a refinancing date; the second is the average age of the asset base crossing into the second half of its technical life; the third is a narrowing of the validity window for the tax shield or the applicable incentive regime. Once any one of these thresholds is crossed, the gap between EBITDA and free cash widens not gradually but in steps, such that the same business, with the same revenue and the same margin, can present an entirely different debt service profile across two consecutive years. Because the measurement system registers the step only after it has occurred, the response typically arrives a full budget cycle late.
The way this gap surfaces at the diligence table is close to standardized. When a quality of earnings exercise is run during a transaction process, among the first questions raised is how maintenance capital has been separated from growth capital; in most companies the two sit merged under a single capital expenditure line, so the portion of spending presented by the sell side as growth that in fact goes toward preserving existing capacity emerges only through an asset-level analysis. The inability to make that separation drives the buy side to normalize from the conservative end, and conservative normalization translates directly into the multiple. The question the company never asked itself returns to the table as a justification for discount.
The second surface is the contractual text. As the EBITDA definition in credit documentation widens through one-time charges, restructuring costs, synthetic synergies and comparable add-backs, the distance grows between the figure used in the covenant calculation and the figure that actually generates cash; leverage may look compliant on a debt-to-EBITDA basis while the cash items entering the coverage calculation carry no such comfort. In periods when the two ratios point in opposite directions, the breach generally arrives not from the leverage covenant but from the debt service coverage ratio. Even where each add-back is individually defensible, their cumulative effect is a question of how the definition as a whole has been calibrated rather than of any single line item.
The third surface is valuation itself. Like any performance element that cannot be shown to be repeatable independently of the founder or the incumbent management team, maintenance capital discipline, where it cannot be demonstrated, is priced through the transaction mechanics: a multiple discount, a technical condition precedent to closing, an expanded scope of representations and warranties, or an earn-out structure that ties maintenance obligations into the post-closing period. Each of these instruments distributes the same uncertainty across different parties, yet none of them removes it; what removes it is a maintenance capital record grounded in the asset register and verifiable retrospectively. The working capital cycle deserves the same reading, since a shift in inventory or receivable levels can carry the cash position for an entire quarter without appearing anywhere in EBITDA.
The mechanism that neutralizes this tendency lies not in individual vigilance but in the architecture of measurement, and it typically separates into four components. The first is defining maintenance capital as a line distinct from growth capital and anchoring that line to the asset register — equipment, technical life, date of last overhaul — so that the figure ceases to be a percentage assumption and rests instead on a physical justification. The second is that EBITDA never appears alone in any management report, being accompanied by a mandatory second line from which cash interest, cash tax and maintenance capital have been deducted. The third is that the add-back record is kept at the moment the item is first proposed rather than at the moment it is approved, with its rationale tested retrospectively in the following period. The fourth is separating the role that advocates for these items from the role that approves them, meaning that the party carrying the technical maintenance case does not report into the finance function during budget negotiation.
The mechanism BEIREK establishes on capital-intensive projects rests on operating these four components. When a project or portfolio is taken over, the first record built is a maintenance capital calendar derived from the asset inventory, carrying for each equipment group its technical life, remaining life and the date of the next major overhaul; that calendar is the source of the capital expenditure line in the financial model, not the other way around. Within the same model the EBITDA line never stands alone, sitting above a line for cash available to owners from which cash interest, cash tax and calendar-derived maintenance capital have been deducted, and management reporting carries the two lines side by side.
The second layer is rhythm. In monthly project reviews, add-back proposals are recorded at the proposal stage rather than at approval, and at each quarter end the prior period's add-backs are compared against realized cash; a rationale that is not systematically confirmed is removed from the definition in the following period. In credit agreement negotiation, whether the EBITDA formula in the covenant definition and the debt service coverage calculation draw on the same assumption set becomes a separate control item, since two calculations running on different definitions make the origin of any breach invisible in advance. The record this rhythm produces becomes, in a later diligence process, the primary evidence of maintenance discipline.
The distance between the profit an asset generates and the cash that reaches its owner is a product of management choice rather than accounting technique; EBITDA does not measure that distance, having been designed precisely not to. The question worth asking, therefore, is not which metric is correct, but which metric the institution rewards, and whether that reward points in the same direction as the physical future of the asset.