In the final fifteen minutes of a proposal meeting, the sentence arriving from the other side of the table tends to follow a familiar structure: the product is broadly suitable, the price sits within an acceptable band, and what remains is a reporting screen reformatted to the client's internal conventions, or an interface adapted to speak to systems already in place. No one present registers this as a product decision. On the commercial side it reads as the last threshold before signature; on the technical side, as a few days of work. Its price in the proposal is usually zero, since raising the number at this stage is understood to put the transaction at risk. Within the same week a comparable deviation is granted to another account, and because the two decisions are taken independently, their combined effect is aggregated nowhere.
The second manifestation of the pattern surfaces not at the sales desk but in the product structure. Stock-keeping unit counts grow visibly faster than either the active client base or revenue; part lists accumulate variant codes defined for a single account; on the software side, versions of the same function, separated by small differences, begin to live alongside one another. When identifying which build a support call belongs to takes longer than resolving the issue itself, the threshold has already been crossed. At that point the company has moved, quietly, from an organization selling one product to many clients into one maintaining a separate product per client — a transition no board resolution ever approved.
The behavior has a name — uneconomic customization, the condition in which the total cost generated by a client-specific deviation exceeds the revenue attributable to it — and its mechanics rest on the invisibility of that cost at the moment of decision. Because the additional work is absorbed into gaps in an existing team's calendar rather than through new headcount, it produces no cash outflow, and a burden producing no cash outflow does not appear as a discrete line in conventional cost accounting. What is actually paid is opportunity cost: the same engineering hour not spent on the next release of the standard offering, or on an improvement touching the entire installed base. To the extent that opportunity cost is recorded nowhere, the institutional grounds for declining a deviation request cannot be produced either.
Recognizing that the same tendency is entirely functional under a specific condition is a precondition for managing it correctly. Before product-market fit has been established, a client-specific request is not a cost but a source of information; discovery financed by a customer reveals which function genuinely changes a purchasing decision more reliably than any theoretical roadmap. The difficulty lies not in the behavior itself but in its persistence after the condition has changed: from the moment the standard offering begins to sell repeatably, the same conduct produces proliferation rather than discovery. Because this transition has no date, only a gradient, it is typically recognized late.
The manner in which complexity compounds is the principal reason the scale of each decision is misjudged. Every new variant adds a single row to the product structure while simultaneously adding a dimension to the test matrix, the documentation set, training material, spare-parts planning, upgrade procedures and support scenarios. Managing five variants imposes not five times the burden of managing one, but a burden proportional to the intersections among those layers — and that burden does not decay as the product matures; it resurfaces at every version upgrade. A second effect accompanies it: institutional memory cannot be distributed. The number of people who know which deviation was made for which client, and why, is typically one or two, and unless written down, that knowledge remains locked in a single calendar.
The incentive architecture completes the mechanism that makes the behavior durable. The unit approving a deviation is typically sales, measured on contracted value; the unit carrying its cost is engineering and operations, measured on delivery schedule. So long as the party granting approval and the party absorbing the cost are not the same, the decision is not an individual lapse but a predictable output of the incentive structure. That separation does not close in the absence of job-level profitability measurement, because as long as average gross margin appears healthy, the loss-making jobs at the tail of the distribution dissolve into the average.
The accumulation registers on the balance sheet less often as a discrete cost line than as a lengthening of the working capital cycle. Parts specific to a single client sit as slow-turning inventory; collection slips to the extent that it is tied to bespoke acceptance conditions; and work-in-progress balances swell independently of revenue growth. The sales cycle stretches in the same direction, since each new client can be expected to request at least as much latitude as the one before, which raises proposal preparation time and pre-sales engineering load with every cycle. On the personnel side the effect shows in rework rates and in the turnover of key technical staff; an environment producing continuous bespoke work has difficulty retaining people who want to see progress on the standard product.
How this structure reads at the diligence table differs systematically from how it reads to the founder. The question a buyer's diligence process asks is not the average of gross margin but its distribution across clients and jobs; where the jobs at the tail cannot be explained, the unexplained portion is classified as a structural issue rather than a pricing one. A second question follows on repeatability, since whether what is being sold is a product or a service redefined on each occasion determines the band within which the multiple is discussed. The third is dependency — where the rationale for deviations resides in the memory of the founder or a single technical lead, the condition tends to be priced into the closing structure as a widened earn-out share, a higher escrow ratio, or a deeper representations and warranties package.
The mechanism that neutralizes the tendency does not run through asking the commercial team to exercise more discipline; it runs through building an architecture that carries information into the moment of decision. That architecture has four components: first, a deviation log in which every departure is recorded at the moment it is requested rather than at the moment it is approved, carrying its rationale, its estimated engineering load and the account to which it belongs; second, an allocation of authority that moves approval above a defined load threshold out of sales and onto a table where product and operations sit together; third, job-level margin accounting that loads direct cost and engineering hours onto the individual job, making concealment behind the average impossible; and fourth, a re-standardization rhythm under which accepted deviations are periodically reviewed and either absorbed into the standard product or discontinued.
Without a pricing and ownership layer alongside these components, the mechanism remains incomplete. Charging for a deviation serves not only to recover cost but to measure the genuine priority behind the request; how much of a priced request survives is information an unpriced request will never disclose. In the same way, where the contract does not establish plainly who retains the intellectual property in bespoke development and whether the result may be sold to other clients, the company has transferred part of its own roadmap into the exclusivity of a single account. The gap between the exclusivity period written into the contract and the period actually observed is, more often than not, a burden nobody tracks and which therefore never expires.
The intervention BEIREK builds along this line in complex, capital-intensive projects rests on positioning scope discipline as an operating rhythm rather than a document. In practice this means that every request falling outside scope — however minor it may appear technically — enters a single deviation log; that the log is maintained with columns for engineering hours, procurement impact and schedule impact; and that items above a defined threshold advance only with technical and operational signature alongside the commercial one. The purpose of the record is not to reject individual items but to make accumulation visible, since thirty decisions, each defensible on its own terms, produce a structure that cannot be defended at all when none of them is examined against the others.
The second line of intervention consists of translating that record periodically into valuation language. In a scheduled review, accepted deviations are separated into three categories — those to be absorbed into the standard product, those to be sustained on a priced basis, and those to be discontinued — with an owner, a timetable and a cost impact committed to writing for each. This rhythm opens the distribution hidden behind the average before it reaches an investment committee or a buyer's table; a scope history explicable independently of the founder ceases to function as grounds for discount during diligence and becomes instead evidence of management quality.
A company's capacity to customize is, in itself, neither a virtue nor a defect; what is determinative is whether the organization can distinguish the point at which that capacity produces information from the point at which it produces only burden. Where the distinction is drawn, customization operates as negotiating leverage. Where it is not, the answer to the question of what the company actually sells grows a little less definite with each new client.