Reading the agenda of a weekly management meeting at a fifty-person company, an observer will find that a substantial share of the items reached the same table five years earlier, when the team numbered ten: the payment terms on a supplier contract, whether a particular client request falls inside or outside agreed scope, whether a hiring band should be lifted one notch to close a candidate. The difference is not the nature of the items but their frequency — a few such calls a day then, several dozen a day now. Decision volume in a growing organisation does not scale with headcount; it scales with the number of relationships headcount creates, which rises far faster. The number of people permitted to close those decisions, meanwhile, tends to remain exactly what it was. None of this shows up in the organisation chart, where departments, directors and reporting lines sit in orderly formation; it shows up, undisguised, in the founder's calendar.
A second pattern repeats in the same company. Asked to describe a problem inside the team reporting to them, a director will often describe not the solution but the approval route to the solution — who needs to be looped in, when the next slot with the founder opens, what framing will secure a yes. Even where the formal remit clearly covers the call, the director prices the probability that the decision will later be reversed above the value of moving quickly, and escalates instead. That calculation is not timidity. It is the rational behaviour of someone who has watched a reversed decision erode their standing in front of their own team, and who has concluded that an unreversed absence of decision costs less than a reversed presence of one. The company issued the title; it did not issue the decision right that sits behind the title, nor the guarantee of finality that makes the right usable.
The pattern has a name — management-layer gap: the failure of a middle tier capable of carrying operational decision authority to form, notwithstanding sustained growth in headcount. Its mechanism operates at the intersection of two components. The first is founder decision speed, which in the early stage functions as a genuine competitive advantage, driving coordination cost close to zero and allowing the company to move while better-resourced competitors are still convening; the company grew because of this, not despite it. The second is the location of institutional memory. Which client was flexed under which circumstances, which supplier has a history of late delivery, why a technical decision taken eighteen months ago was quietly abandoned — as long as none of this is documented, it resides in a single head, and the founder's presence at the table becomes an objective condition for the decision being made correctly rather than a matter of preference.
The conditions under which centralisation remains functional are narrow and can be stated precisely: a team small enough that the founder can maintain individual contact within a single working day, a decision volume that stays inside one person's cognitive capacity, and a problem set in which most questions are genuinely novel, requiring contextual judgment rather than the application of a rule. Where those three hold, centralisation reduces coordination cost and increases speed, and any attempt to distribute authority prematurely imports overhead without buying anything. The difficulty lies not in the shortcut but in its persistence once the conditions have dissolved. Past a certain scale the majority of problems are no longer novel; they recur, they exhibit pattern, and pattern is precisely what a written rule captures. From that point onward centralisation no longer accelerates anything — it manufactures a queue.
The first place the queue registers on the financial statements is not the payroll line but the length of the sales cycle and the drift in the delivery schedule. Three days spent waiting for pricing approval on a single proposal is, taken alone, a trivial figure; multiplied across monthly proposal volume, it removes a measurable share of the commercial team's effective capacity, and it does so without ever appearing as a line item anywhere. The same mechanism surfaces on the delivery side as rework: the team, unable to wait, proceeds on an assumption, the assumption fails to hold, and the work is done a second time. Because rework is seldom tracked as a distinct cost category, it never appears in any report under its own name. The company knows the loss exists, does not measure it, and — precisely because it is unmeasured — files it under the ordinary friction of growth.
The second cost accumulates in turnover. A director recruited into the middle tier at above-market compensation and then denied genuine decision authority will typically remain for a period noticeably shorter than the market average for that level, because the alternative available externally is a role carrying the same title with the authority attached. Each departure repays the recruitment cost, the vacancy period, and the successor's learning curve, and the last of these is the largest and the least visible. More expensive still is the interpretive effect: each departure furnishes the founder with one further data point suggesting that middle management does not work in this company, and that data point reinforces the very centralisation which produced the departure. The gap thereby becomes self-sustaining, each cycle of it generating the evidence used to justify the next.
The third cost, and usually the largest, falls due at the moment the company changes hands or takes external capital. The question asked across the diligence table is not what the company has achieved but whether that achievement is repeatable without the founder in the room. A buyer or investor who finds key client relationships, pricing judgment and supplier negotiation concentrated in a single individual does not record this as an operational observation; it is priced as a risk item, and it is priced in three places simultaneously. A step down in the valuation multiple, an earn-out structure that ties a meaningful portion of consideration to the founder's continued presence through a transition period, and an expanded representation and warranty package covering client relationships and key personnel. Aggregated, those three items typically exceed, by several multiples, the several-year cost of having built the middle layer in the first place.
The intervention that closes the gap is not leadership training or executive coaching; those invest in individual capacity, whereas what is absent is not capacity but architecture. An effective intervention comprises three components. The first is an authority threshold table, setting out without exception and in writing which category of decision closes at which role, bounded by monetary, temporal and contractual limits. The second is a decision record kept at the moment of proposal rather than at the moment of approval — requiring whoever proposes a decision to enter the rationale, the alternatives considered and the expected outcome — which makes the decision reviewable after the fact and, more importantly, migrates the institutional memory sitting in the founder's head into documented form. The third is reversal discipline: a decision taken within threshold is not unilaterally overturned from above even where the outcome disappoints, and where reversal is genuinely necessary, the reasoning is recorded and the threshold itself is recalibrated.
In the management mandates BEIREK runs on capital-intensive projects, this architecture is established as part of the project's own governance instrument rather than as a separate exercise in organisational design. Authority thresholds are consolidated into a single table and defined separately across four distinct surfaces — expenditure within budget, change of scope, schedule movement, and contract interpretation — because collapsing those four into one role empties the threshold of practical meaning, whatever the table says. The decision record is maintained in a single register kept independent of approval correspondence, and no entry closes until the proposer, the rationale, the alternatives and the expected outcome have all been populated. The purpose is narrow and specific: when a decision taken in month three is reopened for argument in month nine, the informational basis on which it was originally taken can be reconstructed rather than reconstructed from memory.
The rhythm that operates this record constitutes the second half of the intervention. What is examined in the monthly review session is not the outcome of decisions but the level at which they closed: how many calls that should have remained inside a threshold were escalated, whether the stated reason for escalation was a genuine information gap or an anticipated reversal, and which thresholds consequently require recalibration. That session produces the only objective indicator of whether a middle layer has in fact formed — a distribution of decisions rather than a distribution of titles. It is also the evidence that founder dependency is receding, and it is evidence of a kind that survives external scrutiny, because across a diligence table an organisation chart carries almost no weight while a twelve-month series showing where decisions actually closed carries a great deal.
The threshold indicating that the layer has genuinely formed is not the founder's withdrawal from decisions, which can be announced at any time and reversed just as easily. It is that the founder's written undertaking as to which decisions will not be entered has survived the following quarter unbreached. Where no such undertaking exists in writing, and where breaches of it are not recorded, decision authority remains at the centre however many middle-management titles have been distributed, and the company's growth ceiling stays fixed at one person's weekly capacity. The real size of a company is measured not by how many people it employs but by how many decisions close at their normal cadence during a month in which the founder is unreachable — and that measurement can be performed by the company itself, well before an acquirer or an investor performs it on far less forgiving terms.