In an investment committee session, while the monthly active user curve of a two-sided marketplace advances with reassuring regularity, a second set of observations tends to remain outside the discussion: the same courier carrying three competing applications open in parallel, the same restaurant listed across three delivery platforms with an identical menu at an identical price, the same subcontractor bidding on the same scope through two separate procurement platforms. The user has gone nowhere; the user has merely also remained somewhere else. The retention table is not constructed to carry that distinction, because the question it asks is whether the user returned, not how much of that user's category spend returned alongside.
At the diligence table the gap tends to surface through a single question — what share of this user's total spend in the relevant category clears through the platform. That the question frequently goes unanswered owes less to weak data infrastructure than to an instrumentation panel designed in a way that holds the competitor outside the frame; a panel that observes only what happens within its own walls leaves volume shared elsewhere untouched by any metric. The company thus learns the one thing it did not know about itself only when the counterparty asks, and it learns it at the most expensive point on the closing calendar.
The behavioural pattern carries the name multi-homing risk — the erosion of attachment to any single platform as users transact across competing platforms simultaneously — and its mechanics are not a loyalty problem. So long as the marginal cost of one more application approaches zero on the user side, search time is measured in seconds, and a continuous arbitrage remains available between price and waiting time, multi-homing is rational portfolio behaviour; by declining to make a choice, the user diversifies exposure. The conventional switching-cost discussion consequently measures the wrong variable: what governs the outcome is not the cost of leaving the platform, but the cost of being present elsewhere while remaining on it, and the two are not substitutes.
A second layer follows from the asymmetry with which the two sides bear that cost. Demand multi-homes cheaply under nearly any configuration, whereas supply — carrying integration burden, inventory commitment, certification cycles, tied working capital, or operational training — may tend toward single-homing. Value capture originates precisely in this asymmetry: to the extent that at least one side stays single-homed, the platform becomes the sole gateway to the other side and can defend its commission rate. Where both sides multi-home, the platform's economics converge on those of a traffic router rather than an intermediary, and routers have historically priced within a narrow band.
Early on, this tendency is functional for founders, which is why it goes unrecognised as a problem for a long stretch. Borrowing liquidity from a supply pool the competitor has already assembled, standing up catalogue depth without building it from zero, generating first transaction volume on the back of a rival's marketing outlay — each is a door that multi-homing opens, and each is defensible on capital-efficiency grounds. The difficulty lies not in the shortcut but in its persistence after conditions change: once the scale threshold is crossed and the platform can originate demand on its own, continuing to operate on borrowed liquidity means growing earned volume without converting it into durable economics.
The institutional cost first appears not in the revenue total but in the revenue composition. In a category where multi-homing is widespread, an attempt to lift the commission rate typically results in a portion of volume migrating to a competitor within the same week; this is the most direct available test of the absence of pricing power, and it reads on the page as gross transaction volume expanding while take rate stays flat or declines. The same mechanism leaves a second trace on the incentive side: the discounts, coupons, and delivery subsidies deployed to hold demand often settle not into marketing expense but inside gross margin as contra-revenue, so that margin erosion behaves less like a growth investment and more like a structural weakness.
The second cost sits within the unit economics assumptions themselves. Under multi-homing, a lifetime value calculation must incorporate not the duration for which a user remains on the platform but the share of category spend directed through it across that duration; where that share goes unmeasured, the retention curve can be entirely accurate while LTV is wrong by an order of magnitude. Payback on customer acquisition cost is systematically flattered in consequence, and the growth budget continues to be allocated toward a pool whose returns are shared with a competitor. The combination of these two errors ranks among the most frequent explanations for cash burning faster than the model anticipated.
On the diligence and valuation side the structure reaches its third and most expensive surface. For an acquirer or a senior lender the governing question is not the magnitude of revenue but whether that revenue can be shown to repeat independently of the founder and of ongoing incentive spend; in configurations dominated by multi-homing, revenue repeats while ownership does not. The transaction architecture that follows is predictable: a portion of the headline multiple shifts into earn-out, renewal of supply-side commitments enters the conditions precedent, representations and warranties widen to cover the continuity of transaction volume, and the escrow ratio is calibrated inversely to the measurability of category share. That the party which did not measure ends up bearing the risk of what it did not measure, through price, is the ordinary outcome of that table.
The mechanism that neutralises this tendency is not individual awareness but a rebuilt measurement and contracting architecture, and it resolves into four separable components. The first is the measurement layer: alongside monthly active users and retention, share of category spend and supply-side capacity allocation per platform are installed as permanent indicators. The second is the pricing layer: instead of demanding exclusivity, tiered economics are constructed against volume commitments, so that single-homing becomes preferable by arithmetic rather than by clause. The third is the product and operations layer, in which assets that raise the cost of multi-homing — inventory synchronisation, payments and working capital rails, compliance and reporting infrastructure, data feedback loops — are deliberately accumulated on the platform side. The fourth is the governance layer: which side is to be held single-homed becomes an explicit decision, and budget allocation is separated according to that decision.
BEIREK begins its intervention in structures of this kind with the decision architecture itself. A side-by-side multi-homing map enters the investment committee pack — for supply and demand separately, the estimation method for volume clearing outside the platform, the observation on which that estimate rests, and the line items that would move should the estimate prove wrong all appear on a single page. The growth budget is separated by side rather than by channel, the rationale for each allocation is recorded at the moment of proposal rather than at the moment of approval, and the outcome of a controlled pricing test is placed as a standing item in the quarterly review rhythm; how far the rate can be carried becomes a measurement rather than an opinion.
The second line of intervention runs through contracts and data rights. In supply-side agreements we construct economics composed of tiered commission, priority placement, and payment-term advantages in place of exclusivity clauses, tie the triggers of those tiers to auditable volume thresholds, and embed a minimum reporting right covering transactions that do not clear through the platform. The same work feeds directly into the design of the conditions precedent list during a transaction: to the degree that supply-side concentration, the commitment renewal calendar, and the normalised trajectory of incentive intensity cease to be findings the buyer discovers and become quantities the seller has already measured and disclosed, the discount negotiation proceeds on structure rather than on price.
The question a platform ultimately owes itself is not whether the user stayed, but whether the staying persists despite the competitor also being present or because the competitor is not; both answers generate the same retention curve and two entirely different company valuations, and whichever party decides when that difference gets measured has also decided who will pay for it.