Among the slides a marketplace company brings to its board, two curves are almost always placed side by side: monthly matches and gross transaction volume. Both point upward, and the discussion that follows tends to occupy itself with acquisition cost, channel mix, and category expansion. Further back, in an appendix that rarely draws a question, sits a third figure — the share of buyer-seller pairs first matched in a given month that are still transacting on the platform six months later. As a category matures and the supply side professionalizes, that figure typically trends downward, while aggregate volume continues to rise on the strength of pairs entering the system for the first time each month. The two movements are not in contradiction; they simply measure different things, and only one of them describes what the business will look like once acquisition spending stops growing.
The same pattern surfaces considerably earlier on the operating side, where it becomes visible without any cohort work at all. Support logs show attempts to exchange a telephone number or an email address inside the messaging module clustering after a particular transaction threshold; suppliers, from the second engagement onward, ask for direct contact in order to shorten the quoting cycle; buyers, for their part, keep the first order deliberately small, treating it as a test of the relationship before routing the substantial volume elsewhere. Where sales compensation is indexed to the first match, this behaviour is rendered invisible by the company's own measurement system, since the match occurred and the commission was earned. The result is an organization in which the people who understand the problem best and the people formally responsible for reporting it sit on different layers, with no instrument connecting them.
The behaviour has a name — platform disintermediation, the migration of a matched relationship off the intermediary that created it — and its mechanism is a question of when value is produced rather than a question of loyalty or commercial ethics. What the platform genuinely supplies, namely locating a counterparty, screening it, and furnishing a trust signal that substitutes for a reference the parties do not otherwise possess, is produced and largely consumed at first contact. The commission, by contrast, is levied again on every subsequent transaction for a service that has already been rendered and will not be rendered a second time. What forms between the parties at the second transaction is a bilateral trust relationship whose maintenance cost is now carried by the parties themselves rather than by the intermediary, while the pricing continues as though nothing had changed.
Seen from that angle, the decision to transact outside the platform is entirely rational under a specific set of conditions: the matching cost is sunk, trust has been established, and the commission has become a concrete saving available for division between two willing parties. The difficulty lies not in the shortcut itself but in a pricing architecture that remains fixed while the conditions underneath it change, since any structure that bills indefinitely for a service delivered once will, predictably, push the counterparty toward looking for an exit. Consistent with this, leakage tends to run highest in precisely those categories carrying the lightest service load — where documentation, escrow, dispute exposure, and compliance obligations per transaction are minimal, the post-match function of the platform becomes largely symbolic, and its claim on the transaction becomes correspondingly difficult to defend.
The institutional cost reads at first as lost commission, though the item that actually reaches the balance sheet and the valuation is a different one. Marketplace valuations are considerably more sensitive to the durability of the take rate than to its level, because the question that matters to an acquirer or an investor is not how much of today's volume converts to commission but how many transactions a pair acquired this year will generate over the following three. When second and third transaction rates decline on a cohort basis, the number of transactions across which acquisition cost is amortized declines with them, and a CAC modelled to spread over four or five transactions compresses in practice into something closer to one and a half. During a growth phase, aggregate volume can mask this compression entirely; the quarter in which it becomes visible is, more often than not, the first quarter in which the acquisition budget is trimmed.
At the diligence desk the masking does not survive long. An acquirer or a senior lender will reconstruct volume not in aggregate but by cohort, derive twelve- and twenty-four-month transaction continuity for each, and separate the share of growth attributable to existing pairs from the share attributable to new acquisition. Once that separation has been performed, the vocabulary of the negotiation changes: rather than contesting the multiple directly, the buyer proposes an earn-out tied to the recurring portion of volume, brings the definition of GMV within the scope of representations and warranties, or calibrates the escrow ratio to observed cohort performance. What these instruments share is that they transfer leakage risk back to the seller for a period extending well past closing, which means an unmeasured leak reduces the cash component of consideration directly, and leaves little room for rebuttal once the cohort table has been assembled by the other side.
A second institutional cost accumulates inside the company's own measurement and incentive architecture. A commission indexed to the first match invests, by construction, in the beginning of a relationship rather than in its continuation; where a category manager is assessed on the number of suppliers under management, the volume those same suppliers transact off-platform falls within nobody's area of responsibility. Over time the most valuable relationships in the portfolio — high-volume, repeating, low-service-cost pairs — become the relationships that exit the system fastest, while the mix that remains consists increasingly of one-off transactions carrying heavy support burdens. This is a selection effect compressing unit economics from both ends simultaneously, and it is rarely legible in a monthly dashboard, since each individual departure resembles ordinary churn rather than an outcome the pricing structure itself produced.
The intervention that neutralizes this tendency is built into architecture rather than into appeals to loyalty, and it separates into four components. The first is service positioning: relocating payment flow, escrow and warranty, dispute resolution, insurance, and the production of tax and compliance documentation away from the moment of matching and into the recurring transaction stream, so that the saving obtained by leaving is narrowed by the operating burden the parties must then absorb themselves. The second is pricing architecture: a commission that declines on a per-pair basis, a cap per transaction, or a shift to a subscription structure beyond a volume threshold — the smaller the saving, the lower the expected return on exit behaviour. The third is measurement: tracking leakage not in aggregate volume but in the per-pair second transaction rate, in the widening interval between match and first order, and in contact signals within the messaging module. The fourth is incentive: assessing teams on a pair's cumulative on-platform volume rather than on the first match.
Among these four, the contractual clause is the weakest lever, and its weakness derives from enforcement economics rather than from legal validity. A provision prohibiting circumvention places the burden of proving breach on the platform, while the information required for that proof is frequently the very information the platform cannot observe, and pursuing a breach ordinarily means losing the pursued party as a customer. The practical function of such clauses is therefore rarely deterrence across the general population of users; it is closer to establishing a threshold with counterparties for whom reputational cost is high — corporate buyers, regulated institutions, large accounts with procurement oversight. Omitting the provision would be an oversight, but a defensive strategy resting on it, insofar as it leaves the structural cause of leakage untouched, tends to delay measurement rather than to alter behaviour.
BEIREK's intervention in structures of this kind begins not with redefining what the platform business is but with separating where value is produced from where it is collected. The first mechanism established is a leakage cohort model: matched pairs are fixed to their month of introduction, on-platform transaction continuity, transaction size, and support burden are tracked for each pair independently, and aggregate volume growth is decomposed into the portion originating from existing pairs and the portion originating from new acquisition, then carried to the board in a single table rather than across three unrelated slides. The second mechanism is a take rate architecture review layered on top of that table, in which the analysis proceeds line by line — which service the commission corresponds to in each category, beyond which threshold the available saving begins to finance exit behaviour, and which services can plausibly be moved past the match on both technical and operational grounds.
A second layer concerns the record and the cadence of decisions. Changes to pricing and to service scope are recorded at the moment of proposal rather than at the moment of approval; which cohort behaviour a change is intended to affect, which indicator is expected to move and in which direction, and in which quarter that movement will be measured are all written down in advance, so that the rationale cannot be reconstructed retrospectively once the result arrives. The same discipline ensures that the questions a diligence desk will ask ahead of a sale or a capital raise are asked internally first — cohort decomposition, the scope of the GMV definition, the share of volume that recurs, and the service on which that recurrence rests. Having those answers prepared before a process opens ranks among the more concrete levers available for narrowing the eventual scope of earn-out and escrow provisions.
The durability of a marketplace is measured less by the number of introductions it makes than by how long the parties it introduces continue to require it in order to sustain the relationship that followed. Where that interval shortens, the explanation is unlikely to lie in the conduct of the counterparties, who are responding sensibly to the price they are asked to pay for a service already delivered; it lies instead in the answer the platform is able to give to a narrower and less comfortable question — what work, precisely, does it perform after the match has occurred, and would a buyer reading the cohort table arrive at the same answer.