The fastest route into a company's decision architecture is not the organisation chart but the approval trail behind fifteen contracts signed in the past six months, traced backwards. That trail almost invariably terminates somewhere narrower than the chart suggests: beneath a procurement decision nominally within the purchasing manager's remit sits an email confirmation from the general manager; a pricing flexibility formally devolved to regional heads is exercised only after verbal confirmation from the founder; and an expenditure already approved as a budget line returns, at the payment stage, to the same desk for a second time. Nobody inside the company describes this as a problem, because for everyone involved the system works — decisions get made, work proceeds, delay is not conspicuous. The difficulty is not that the system functions, but that the reason it functions is not the line written in the document.
The same observation appears from the opposite direction. In a substantial share of companies that report having an authority matrix, the document was drafted by an adviser in the year of incorporation, adopted by board resolution, and never opened since. Its monetary thresholds were calibrated to the currency value of that year, with the predictable consequence that, several years on, nearly every purchase falls into the highest approval band; the matrix remains technically in force while doing nothing except routing every decision to the same individual. That the document is never breached is not, in this configuration, evidence of compliance — it is the consequence of the document having ceased to discriminate between decisions at all.
The mechanism operating underneath is not negligence but a cost calculation. Delegating authority imposes three separate burdens on the person delegating: remaining answerable for the outcome regardless, standing up a separate review layer to monitor decision quality, and absorbing the errors that surface during the first months of the transfer. Retaining authority produces none of those three in the short term; it produces only a time cost, and a time cost remains unfelt for as long as it does not appear on the founder's own calendar. This preference is rational under the conditions in which it typically forms — a small company, a decision volume that fits within one person's cognitive capacity, and errors that are reversible. The problem lies not in the preference but in its persistence after those conditions have changed.
A second mechanism is the asymmetric distribution of authority and accountability. The matrix is usually written from the accountability side — who is answerable for which function, who appears as owner of which output — while the decision rights required to carry that accountability are not entered on the same line. What emerges is a manager held to a target whose budget he cannot set, and a unit head answering for team performance without the ability to determine who staffs the team. This configuration produces, in predictable fashion, two behaviours: managers seek approval for decisions rather than taking them, and at the moment of reckoning the line of accountability is pushed upward. The company tends to name this a cultural problem, whereas what is being observed is the natural consequence of leaving the authority column of the matrix blank.
The balance-sheet correspondence of this structure does not appear as a discrete line item; it sits distributed across the working capital cycle, the length of the sales cycle, and the mid-level attrition rate. Where the approval line is effectively locked to a single individual, supplier contracts are signed when that individual's calendar permits, which produces a procurement rhythm governed by the approval window rather than by price advantage. The same lock lengthens the cycle time of proposals awaiting discount approval on the commercial side, registering not as customer loss but as quiet erosion in conversion. Mid-level turnover draws on the same source, since a position stripped of decision rights has limited retention value for a senior professional who has held such rights elsewhere.
At the review table this layer is interrogated through the traceability of the document rather than its existence. The reviewing party examines the date of last approval, the rationale behind interim revisions, and — most determinative — the degree of correspondence between the thresholds defined in the document and the actual approval records. The sample is rarely random: the three largest expenditures by value, two supplier selections that appear anomalous, and a slice of hiring and termination decisions are pulled, and in each case the question asked is who in fact granted the approval. Where the person named in the matrix and the person who approved diverge, the finding is not reported as a documentation deficiency but as an inability to verify decision capacity independent of the founder — and the second formulation is considerably more expensive in the transaction structure.
Measurement is the dimension most frequently left empty here, because an authority matrix is not generally thought of as something measurable. What is measurable, however, is not the matrix but the behaviour it produces: the utilisation rate of delegated authority, the number of approvals escalated above the threshold at which they should have terminated, the average elapsed time between decision request and decision, and the share of exception approvals within total approvals. Tracking those four indicators across a single quarter establishes whether the matrix is a living structure or an archived document more definitively than any management interview. The exception rate is particularly informative; once it rises above a certain band, the matrix has stopped describing the rule and begun describing the route around it.
On the ownership dimension, what is sought is not a name identified as owner of the matrix but whether that name carries the authority to change it. In most companies the document is owned by human resources or the quality function, yet neither of those units can decide whether a given authority is to be delegated; they can only maintain the record of the decision. Ownership remains nominal for as long as the party keeping the record and the party making the decision do not meet on the same line. For that reason, in mature structures the revision authority over the matrix sits with the board or the executive committee while the monitoring and reporting obligation sits with a separate function; the separation of the two is the only structural mechanism that prevents the matrix from eroding silently.
The structural intervention has three components, built in sequence. The first is deriving the current state from the trail rather than the document: the approval records of the past twelve months are scanned, the de facto decision line is mapped, and that map is placed alongside the matrix in force. The second is recalibration of thresholds — where monetary bands are defined as a ratio to annual revenue or to the relevant budget line rather than as absolute figures, the matrix acquires the capacity to update itself, failing which it is rendered inoperative again with every inflationary cycle. The third is keeping the decision record at the moment of proposal rather than at the moment of approval; once it is recorded who brought the proposal, which alternatives were weighed, and on what grounds they were eliminated, delegation ceases to be a question of trust and becomes a traceable process.
BEIREK's intervention in this area typically begins not with drafting a new matrix but with reconstructing the existing decision line backwards, because a matrix written from a blank page converges on the adviser's reference model rather than on how the company actually operates, and is abandoned within six months. Once the mapping is complete, the mechanism we install has three layers: separation of decision classes along the axes of value, reversibility and counterparty risk; fixing, on a single line for each class, the distinction between the decision-maker, the party to be consulted and the party to be informed; and binding those lines to a quarterly review rhythm. The rhythm matters more than the document here — a matrix that is never reviewed detaches from the actual line within a year even if it was accurate on the day it was written.
The second layer is the management of exceptions. No matrix can encompass every decision, and matrices that attempt to do so become unusable; the question is therefore not how to prohibit the exception but how to make it recorded. Under the regime we operate, every exception approval is logged in a single register together with its stated rationale, and the quarterly review examines the distribution of those exceptions: an exception recurring within the same decision class is not a breach of the rule but evidence that the rule has been miscalibrated, and the threshold is corrected accordingly. Once this feedback loop is established, the matrix ceases to be a compliance artefact and becomes an instrument through which the company measures its own decision capacity — which is precisely what is verifiable at the review table.
The genuine test of continuity is what a company can decide, and what it holds in abeyance, during a month in which the founder is unreachable. The list of deferred decisions constitutes, independently of whatever the matrix says on paper, the inventory of authority that was never delegated; and the length of that inventory is the most honest estimate available of the discount an investor will apply for founder dependency. What determines a company's valuation is, more often than not, not the magnitude of past performance but the demonstrability that the decisions producing that performance can be reproduced without the founder — and the authority matrix is either the most concrete evidence of that demonstration or its most visible absence.