A recurring pattern appears in monthly close meetings. When the doubtful receivable provision comes up, the justification offered for the figure is rarely a rule; it is the figure from the prior period. The aging schedule is on the table, collection performance is known, sector conditions have been discussed at some length, yet the answer to the question of which threshold generated which rate is typically a statement of continuity with what was booked three months earlier. The same logic runs, in the same meeting, through warranty provisions, inventory write-downs and litigation reserves, each of which carries a structurally different class of uncertainty and each of which is nonetheless adjusted in the same direction by the same reflex toward prudence. Nothing about this is irregular or ill-intentioned; it is simply the lowest-cost way an organization produces consistency in the face of uncertainty it cannot resolve.
Opened again several years later on a due diligence table, the same subject changes shape. The review team does not ask for the provision balance, which it already has from the trial balance; it asks for the provision policy. What arrives, in most cases, is not a policy document but the calculation itself — a spreadsheet with columns broken into aging bands and a total at the foot. An aging schedule, however, is an output, whereas a policy is the separate document specifying who sets the parameters of that output, on the strength of which data, at what frequency, and under whose approval. The collapse of these two artifacts into a single file is the most common condition observed on the existence dimension: the practice is present, the policy exists verbally in the heads of two or three people, and the written, dated, approved version does not exist at all.
The mechanism beneath this pattern arises from the structural position provisions occupy in the income statement. A provision is one of the few items that reduces profit without any cash leaving the business, that therefore determines the reported result directly, and that is fixed by no external invoice, contract or counterparty confirmation. The cost of over-providing is immediate and highly visible — profit falls, bonus pools narrow, bank covenant headroom compresses, and someone is asked to explain a variance in the following week. The cost of under-providing, by contrast, is deferred to a later period and, when it eventually materializes, can be attributed with reasonable credibility to a customer failure, a market shift or a specific dispute. This temporal asymmetry pushes the decision-maker in a predictable direction: anchoring on the prior figure and proceeding by small adjustment is both cheaper to produce and easier to defend than recalibrating from first principles each quarter.
A second mechanism follows from the separation of information and authority. The party best positioned to know whether a receivable will actually be collected is usually the commercial or field team, which sits closest to the customer relationship; the party that records the provision is finance. The commercial estimate remains structurally optimistic to the extent that the role carries an obligation to preserve the relationship — an incentive embedded in the position itself rather than a matter of individual temperament. Finance, unable to observe the relationship directly, falls back on the schedule or on the prior period ratio. Where no formal interface exists to reconcile these two information sets, the provision amount becomes the residual of a negotiation between two functions, and the outcome of a negotiation is, by definition, not reproducible from the same inputs by a different pair of participants.
The first valuation consequence of this configuration surfaces in the normalized EBITDA exercise. In a quality of earnings review, the analyst must decompose period-over-period margin improvement into its sources — pricing, cost, volume, mix — and determine what portion, if any, is attributable to loosening in provision levels. Where the policy is unwritten and its parameters unfixed, that decomposition cannot be performed on the evidence available, and improvement that cannot be decomposed is, predictably, normalized at the most conservative assumption the reviewer can justify to an investment committee. The result is an impact disproportionate to the absolute size of the line: a few hundred thousand of estimation uncertainty, once carried through the multiple, produces a considerably wider band in enterprise value. That disproportion between the smallness of the item and the size of its effect is the channel most frequently recognized late on the sell side.
The second channel runs through the balance sheet. Where the receivable provision is assessed as under-calibrated, the shortfall is resolved either by lowering the working capital peg, which produces a price adjustment at completion, or by treating the gap as a debt-like item charged directly against the purchase price. Warranty and litigation provisions operate differently: an undocumented provisioning approach tends to generate a specific indemnity head within the representations and warranties package, an increased escrow ratio, or a condition precedent requiring an independent actuarial assessment before signing. Where warranty and indemnity insurance is in play, the ordinary outcome of the underwriting process is exclusion of areas lacking a written policy and a supporting history of realized outcomes, in which case the exposure does not disappear but simply migrates out of the policy and back onto the seller.
The third channel is the absence of measurement, and it operates most quietly. A provision is by definition an estimate; what converts an estimate into an accounting estimate rather than a preference is the discipline of comparing it against what subsequently occurred. Where a company does not regularly measure how much of the provision booked actually converted into loss, how much was released, and in which customer segment or product family the deviation concentrated, the estimate degrades into an opinion held with varying conviction. This is substantially what the expected credit loss approach under IFRS 9 and the broader measurement provisions on liabilities require of the company: a historical loss rate bound into a matrix that is updated with forward-looking information and tested retrospectively against outcomes. Where that testing is absent, a gap persists between the compliance opinion in the audit report and the verifiability an investor is looking for.
The ownership and continuity dimensions typically emerge toward the end of a review, condensed into a single question: who determines this amount, and what happens in the period when that person is unavailable. In a structure where proposal and approval reside with the same individual, where thresholds are not written down and departures from practice leave no trace, the answer inevitably resolves to a name rather than to a process. The buy side does not price this as accounting risk; it prices it as key-person dependency, and the pricing of key-person dependency follows a familiar set of instruments — post-closing retention conditions tied to the individual, an extended earn-out period, a wider escrow with a longer release schedule. At that point the provision policy has ceased to be an accounting matter and has become a component of deal architecture.
The intervention that neutralizes this tendency is architectural rather than attentional, and it rests on four components. The first is a dated policy document, approved by the board or the competent organ, that defines the parameters, data sources and calculation steps separately for each provision class, since receivables, warranty, inventory and litigation carry different uncertainty structures and should not share one prudence reflex. The second is the historical realization file on which those parameters rest, together with a defined frequency at which the file is refreshed. The third is an authority matrix that separates proposal from approval and ties escalation to amount thresholds. The fourth is a back-testing exercise that compares provisions booked against losses realized on a stated cadence. Each component looks unremarkable in isolation; constructed together, they form the machinery that makes the estimate independent of the person producing it.
BEIREK's intervention in this area does not begin with drafting the policy text; it begins with changing the moment at which the decision enters the record. Provision decisions are typically documented at the point of approval, whereas the substance of the judgment forms at the point of proposal, which means the approval file records an outcome without preserving the reasoning that generated it. The record we build fixes, at proposal, the data relied upon, the threshold applied and the assumption adopted, with approval landing on top of that record rather than replacing it. Alongside this we maintain every departure from policy in a separate exception log — amount, rationale and approver together — because the document a review team actually reads is not the policy text but the exception file, and a policy with no recorded exceptions across several years reads as a policy that was never applied.
The second line of intervention is rhythm. On a quarterly cadence, provisions booked are compared against losses realized at segment level, a tolerance band is defined in writing before the comparison is run rather than after, and items falling outside the band trigger a parameter update with a documented rationale. The durability of that rhythm is then tested in a single way: the same raw data is handed to a financial controller reading the policy for the first time, and the question is whether the amount produced falls inside the defined band without recourse to the person who ordinarily prepares it. Where that test is passed, the provision has moved out of an individual's judgment and into the company's capacity, and capacity — unlike judgment — is the thing that can actually be demonstrated to an investor during a transfer process.
The question posed on the review table, in the end, is not whether the provision amount is correct, since no buyer can prove the accuracy of an estimate before completion and no seller can disprove a conservative one. The question is whether the mechanism that produced that amount would produce a comparable amount without the person who produced it, under a different close calendar, in a different market condition, with a different controller reading the same file. The answer to that question is not found in the accounting records, where every provision looks equally deliberate once it has been posted; it is found in the documents that show how the decision was constructed — the parameters fixed in advance, the realization file that supports them, the threshold that triggered escalation, and the exceptions recorded rather than absorbed.