When the channel table is opened in a quarterly marketing review, the budget attached to the top row is preserved almost without argument, and negotiation begins several rows below it. The logic behind the table is straightforward: the channel through which a converting customer first entered the system has been recorded, and the entire revenue associated with that customer has been written into that channel’s column. Should the same customer have been retargeted before closing, opened two nurture emails, returned twice to a comparison page, and finally typed the brand name directly into a search box, none of those interactions appear anywhere in the table. What the meeting debates, therefore, is not what actually happened, but the cross-section that the measurement rule renders visible, and the budget is allocated according to the outcome of that debate.
The same pattern emerges more sharply on the reduction side. In a spending-compression exercise, the first line items removed are typically those with no revenue written in their own name — brand visibility, content production, mid-funnel retargeting, industry events, comparison and evaluation material. Defending such items requires a manager to argue without a number in hand, and an argument without a number loses predictably in a corporate setting. The consequence of the cut, however, does not appear within the same quarter; first-touch volume holds at roughly the prior level for some time, because the demand pool fed by the removed activity continues to drain for several months. That lag in the causal chain produces a misleading confirmation that whoever made the cut was right.
The pattern has a name — first-click attribution bias, the practice of assigning the full conversion value generated by a purchase journey to the touchpoint that opened it. The tendency begins as a technical measurement preference and then hardens into an institutional convention, largely because the first touch is the cheapest point to observe, the least contested to assign, and the simplest to report. The channel through which a user enters the system is fixed by a single record, whereas binding every intermediate interaction on the path to closing to the same identity is both expensive and fragile in a measurement environment where identity resolution has fragmented. The organisation prefers, understandably, to model what it can measure rather than to measure what it cannot model.
In certain conditions that preference is entirely defensible, and the mechanism cannot be understood without acknowledging as much. In low-value, low-consideration purchases completed within a single session, the first touch and the last touch largely coincide, and in such a portfolio the cost of constructing a more elaborate attribution model exceeds the decision improvement it would deliver. The second function of the first-touch rule is not measurement at all but governance: a single-touch convention terminates the contribution dispute among channel owners decisively, which makes it closer to a settlement document than to an analytical method. The difficulty lies not in the shortcut itself but in the shortcut persisting after the character of the purchase cycle has changed.
As the cycle lengthens, the carrying capacity of the rule declines. In a sale where consideration extends over weeks or months, where several decision-makers participate, and where technical approval sits in a different function from purchasing approval, the first touch marks only the moment the door opened; what determines the close is the trust constructed in between. Under such a structure the first-touch rule systematically overfunds the channels positioned at the top of the funnel and systematically underfunds those operating in the middle. Over time the portfolio drifts toward a mix that initiates a large number of journeys and completes comparatively few, and while acquisition cost appears unchanged, conversion rate erodes quietly.
The first balance-sheet expression of that erosion appears not in the marketing expense line but in sales-cycle length and in the realised, as opposed to attributed, cost of customer acquisition. Attributed acquisition cost reads low to the extent that a portion of mid-funnel spend has been booked to brand budget or to general overhead, while the true cost accumulates in the hours the sales team spends per close, in the number of times proposals are reissued, and in the discount concessions granted at the end. Incentive structures on the agency and media side reinforce the tendency, since for a supplier whose performance is assessed on first-touch volume, the most rational behaviour is to generate new visitors irrespective of conversion quality. Once annual media commitments have been contracted on that logic, the correction window extends to the length of the contract term.
The second and more expensive expression surfaces at the diligence table. What determines a company’s valuation is rarely past growth in itself, but whether that growth can be demonstrated to be repeatable independently of a particular person, channel, or set of period-specific conditions. When the gap between the attribution model and realised cash becomes visible during quality-of-earnings analysis, the discussion ceases to be about a reporting difference and becomes a question of whether the growth engine is genuinely understood. At that point the typical counterparty response is not to reduce the headline price directly but to shift the risk into the structure: an earn-out indexed to an acquisition metric, a post-closing undertaking on channel performance, or an expanded representation and warranty covering marketing statements.
The third expression is organisational and the slowest to be recognised. The team owning the channel that generates first touches enters every quarterly review with a revenue figure written in its own name and accumulates budget, headcount, and decision rights accordingly; teams whose contribution never surfaces in the table remain in a defensive posture and, over successive periods, either shrink or rewrite their objectives around the visible metric. The organisation consequently loses its institutional memory of what actually worked, since the only place that memory is held is the attribution table, and the table records solely what its own rule permits it to see. The talent effect runs in the same direction: turnover among staff working on unmeasured activity rises relative to those working on measured activity.
This tendency is managed through decision architecture rather than individual awareness, and the architecture has four separable components. The first is fixing the attribution policy in writing before the period opens and leaving it unchanged throughout; where the rule is selected after results are observed, the selection will inevitably be the rule that vindicates the line item someone wishes to defend. The second is placing incrementality measurement, rather than the attribution model, in the position of arbiter — geographic or segment-based control groups, planned spend pauses, and reading windows declared in advance of those pauses. The third is splitting budget authority in two: a primary allocation distributed according to attributed performance, and a separate reserve, subject to its own approval threshold, for activity that cannot yet be attributed. The fourth is maintaining the decision record at the moment of proposal rather than at the moment of approval, since the only thing that breaks retrospective justification is an expectation written down before the outcome.
BEIREK’s intervention in this area is not to recommend a measurement tool but to reconstruct the decision ground on which measurement rests. In the engagements we run, the attribution policy is first fixed in a single document — which interaction is recorded and how, within which window it counts, and under which conditions the record is treated as void — and that document is approved at investment committee or board level before the budget period opens. A decision record is then opened for every material spend item at the moment it is proposed, stating in advance the expected effect, the interval over which the effect will be read, and the threshold below which the item will be treated as unsuccessful. At period end, assessment is conducted against that record rather than against the realised table.
The second layer concerns cadence itself. Reconciliation between the attribution table and realised cash collection is operated as an independent quarterly agenda item rather than as a footnote to monthly reporting, and where the divergence between the two sources exceeds a defined threshold, the triggering of a channel-level incrementality test has been agreed in advance. The same discipline serves a second function in transaction preparation: the growth-repeatability question a buyer will ask is one the organisation has already put to itself, and recorded, several periods before the transaction table. Whether a growth narrative survives scrutiny depends far less on the force of the narrative than on when the evidentiary chain supporting it was assembled.
Attribution, in the end, is not merely a reporting preference; it operates as an implicit constitution governing which explanation of its own growth an organisation accepts, which team receives resources, and which expenditure is deemed defensible. Where that constitution is not written deliberately before the period opens, it is written at period end by whichever number can be defended most loudly. The question worth putting is not which channel performs better, but which activity the current measurement rule renders structurally invisible, and in which quarter the consequence would appear if that activity were halted today.
Answering that question falls not only to the marketing function but equally to the finance leadership preparing for a valuation negotiation, since an explanation of how growth is produced that is not constructed internally will be constructed externally, by the party on the other side of the table.