Distribution: compare channels on contribution and control.
Answer:
A channel is not merely access to demand. It changes net price, assortment, data, service, returns, working capital and conflict risk. Comparison starts with an incremental order and reconstructs its full contribution, then values sales transferred from existing channels. The channel producing the most revenue may destroy value when it mainly captures customers already served at higher margin.
Method and decision:
Use the same order, customer, unit or account and the same period for every channel. Reconstruct net price, product, fulfilment, delivery, return, payment, service and acquisition costs. Estimate new customers, conquest and transfers using geographic tests, access codes, surveys, history and cohorts, and publish the interval. Include integration, team, content, inventory, systems, contractual minimums and opportunity cost; model capacity steps rather than a constant average cost. Solve the required order count, new-share or basket threshold under returns, commission, cannibalisation and competitor response. Extend a channel only when incremental contribution after transfers, fixed costs and constraints remains positive over a realistic demand-and-returns range.
Worked example:
Out of 10,000 potential orders, direct wins 8,000 at €100 net, €42 variable costs and €180k fixed costs. A marketplace wins 10,240 at €82 net, €50 variable costs and €90k fixed costs. Of marketplace orders, 45% are genuinely new and 35% replace direct orders. Direct contribution is 8,000 × (100 − 42) − 180,000 = €284,000. Marketplace own contribution is 10,240 × (82 − 50) − 90,000 = €237,680. Direct margin lost on 3,584 transferred orders is 3,584 × 58 = €207,872. Added to direct, marketplace incremental contribution is 237,680 − 207,872 = €29,808 before competitive effects. It sells 28% more volume yet creates €46k less value. Open only as a pilot if the genuinely new share exceeds 51%, assortment limits transfers or commission is renegotiated.
Limits:
Order origin is difficult to identify without a test. Platform terms can change quickly. Visibility and learning effects may emerge after the initial period.