Among the scenes that recur most reliably at a diligence table is the one in which a single question, put to three different people, produces three different answers: at precisely what point is revenue recognised. The managing director points to the moment the work was completed, the finance manager to the date the invoice was issued, and the system record to the date of dispatch. Each answer is internally coherent, and each has operated without friction for years; the difficulty is that the mutual incompatibility of the three becomes visible only when an outside party attempts to rebuild the picture from the ledger upward. What is being discussed at that point is no longer an accounting detail, but the definition of the base to which a multiple will be applied.
The second observation is quieter, and it is that this question is one the company has never put to itself. The first heading in the folder opened by the reviewing party is usually accounting policies, and what comes back is, more often than not, either a generic text lifted from the notes to the audit report or nothing at all. That absence does not mean the company has no policy; it means the policy is unwritten, and the distance between those two conditions is precisely what the review is measuring. A policy set is in fact being applied every day, but it has come into existence as an accretion of habits rather than as a body of decisions.
The mechanics of that accretion are quite predictable. When a new transaction type appears for the first time — a multi-year maintenance contract, a development expenditure, a customer advance, a warranty commitment — whoever posts the entry that day reaches for the nearest available analogue, makes a choice, and the work moves on. Because the choice was not taken as a policy decision, no rationale is recorded alongside it, and a choice without a recorded rationale becomes progressively unexaminable, since examining it would first require knowing which alternatives were set aside. Every subsequent transaction of the same kind is then matched to the original entry, and what emerges five years later is a chain of practice that looks consistent while its justification sits nowhere. The chain is not an error; the cost of re-evaluating the measurement basis on every transaction is genuinely high, and leaning on precedent lowers that cost. The problem lies not in the shortcut itself but in the shortcut persisting after scale, contract architecture, and counterparty profile have all changed.
A second layer of mechanism concerns the purpose for which the bookkeeping architecture was built in the first place. In companies that financed their growth internally, the accounting infrastructure is typically assembled to satisfy filing obligations, and useful life assumptions, provisioning practice, inventory valuation method, capitalisation thresholds, and expense classification together form a system tuned to managing the periodic taxable base. Under that lens, policy questions are not questions of representation but questions of consequence — which choice produces which tax effect. What the investor side is looking for runs in the opposite direction: which choice represents the economic reality of the business, and with what fidelity. The fact that two lenses have operated on the same ledger for years surfaces the need for normalisation at the earliest stage of review, and what gets priced is not the magnitude of the normalisation but its unpredictability.
In companies where the policy has in fact been written down, the review shifts to a different layer: the distance between what is written and what is done. Policy manuals are commonly produced under the pressure of an audit cycle or a credit application, approved in the form in which they were drafted, and thereafter left untouched by the daily flow of operations. Testing the application dimension therefore rests not on the existence of the document but on tracing how the same transaction type was recorded across different periods and by different people. A sample-based test of that kind will usually reveal, within the first few hours, either that thresholds defined in the manual are not used in practice or that an undefined treatment has quietly become the system default.
The first channel through which all of this reaches valuation runs through the quality of earnings work. Every valuation built on a multiple rests on the assumption that the base to which the multiple attaches can be defined with stability; in a ledger where policy choices are undocumented and variable, that assumption weakens directly. A one-year shift in the recognition point carries a material volume of revenue between periods on multi-year contracts and changes the slope of the growth curve; a choice at the capitalisation boundary for development expenditure moves operating profit and the amortisation charge simultaneously. Uncertainty of this kind is answered not by pricing the mid-point of the adjustment but by pricing its conservative end, and the acquirer’s committee reads the width of the range itself as a risk premium.
The second channel, less frequently discussed, is contractual. Working capital adjustment clauses stipulate that the closing balance sheet will be prepared using the same accounting principles as the reference balance sheet and consistently with past practice, and that language silently assumes the existence of an identifiable past practice. Where no such definition exists in writing, whichever team actually performs the calculation after closing also establishes the definition, which creates a one-sided negotiating surface that opens only after signature. The same asymmetry is sharper still in earn-out structures: once post-closing performance begins to be measured under the acquirer’s group policy, the threshold the seller was targeting can move away without a single operational variable changing. Attaching the policy set to the agreement as a schedule is therefore not a gesture toward compliance but a direct mechanism of price protection.
A third channel opens where the ownership and continuity dimensions intersect. In a great many companies the policy resides not in a document but in a person who has held the same role for more than a decade; transaction types are classified in that individual’s memory, exceptions are managed by that individual’s judgment, and consistency is sustained by that individual’s continued presence. The arrangement is efficient for as long as it holds, but to a reviewing party it reads as a control environment dependent on a single point of failure, which is what founder dependency looks like on the accounting line. The consequence takes the form of broader representations and warranties, a higher escrow percentage, or the conversion of part of the exposure into a condition precedent; and where the exposure is put to an insurer instead, the list of matters excluded from cover draws heavily on exactly this territory.
What neutralises this tendency is not individual diligence but an architecture with three components. The first is the company’s actual transaction taxonomy: not a list drawn from a standard template, but one derived from two years of posted entries and reflecting the transaction types the business genuinely produces. The second is that, for each transaction type, the recognition point, the measurement basis, and the presentation choice are written down together with the reasoning by which alternatives were eliminated; a policy recorded without rationale is nothing more than the oral policy transcribed onto a page. The third is a separate judgment register maintained for items requiring estimation and discretion — doubtful receivables, warranty provisions, useful lives, percentage of completion — recording which data each assumption rests on and under what conditions it is to be revisited. Add a materiality threshold and a protocol for policy changes, and the structure becomes auditable.
BEIREK’s intervention in this area begins not with the delivery of a manual but with rebuilding the points at which policy connects to operations. A policy inventory is derived from the transaction taxonomy extracted out of the ledger itself, each line is recorded with its rationale and rejected alternatives, and estimation items are collected into a distinct judgment register; the period close is then bound to a calendar with defined steps, defined owners, and a defined sign-off chain, so that ownership settles onto an approval point rather than onto a job title. Policy changes are logged at the moment they are proposed rather than after they take effect, which is on its own the least expensive mechanism available for preventing the rationale behind a decision from disappearing.
The second phase tests the measurement and continuity dimensions. Whether the policy has stayed on paper is tracked not by the presence of a manual but by the movement of a handful of indicators: the elapsed time to complete the close, the number and value of post-period adjusting entries, the aging of balances that remain unreconciled, and the dispersion among entries made by different people for the same transaction type. Continuity is measured by a handover test — whether, once the person who wrote the policy steps away from the desk, a newly encountered transaction type is recorded on the same reasoning and to the same result. A rehearsal quality of earnings exercise run across two periods brings the greater part of the reviewing party’s questions to the table while negotiating leverage still sits on the sell side.
A company’s accounting policy, however carefully maintained, generates no value on its own; left undefined, however, it transfers to the counterparty the authority to determine where value will be measured, and that transfer cannot be reversed once the transaction is signed. What the diligence table looks for is not a set of correct choices, but demonstrable evidence that the choices exist as a body of decisions and that they can be reproduced independently of any particular person.