A US consumer-electronics launch, modelled end to end. The marketplace model everyone assumes is cheaper loses money on every order.
I built a market-entry proposal for an early e-commerce operator: what to stock, who they are really competing with, what a single order actually earns, and whether to sell first-party or take a commission from third-party sellers. Seven retailers benchmarked, a full unit-economic model, and a recommendation that contradicted the brief.
Follow the money on one orderAn early marketplace operator wanted to enter US consumer electronics. The plan was sensible in outline — start from an existing catalogue, add portable essentials, build toward higher-value items — and entirely unpriced. No one had worked out what a single dispatched order earns after the card fee, the box, the returns and the cost of finding the customer.
So that is what I built: a basket, a landed cost, and every deduction between them and the bank. The strategy argument follows from the arithmetic rather than the other way round.
I also went and looked. Three stores in one Sunday — two Best Buy branches and a P.C. Richard & Son — to see what the incumbents actually put in front of a buyer: where the staff stand, how the warranty is pitched, what the shelf edge says about returns.
Thirty per cent of a $90 basket is $27 of gross profit. By the time the card network, the carrier, the returns provision and the advertising have taken their share, $7.68 is left — and $8 of that journey was the advertising alone.
Acquisition is the largest single deduction, larger than the shipping and the card fee together. That is the fact the rest of the analysis turns on: at this basket size the business is not a product-margin problem, it is a customer-acquisition problem.
The same $8 of acquisition applied to a third-party order, where the platform earns a commission rather than a product spread. A 6% fee on $90 is $5.40 — it cannot carry an $8 acquisition cost and a $2.91 card fee, so the order loses $6.02 before anything goes wrong.
Marketplace is normally the asset-light option: no stock, no working capital, someone else ships. But the platform still pays the card fee on the whole order and still pays to acquire the customer, while earning only the commission.
Sell first-party for the pilot, and add third-party when sellers bring their own demand — or when the take rate improves, or acquisition gets cheaper. Not before. This is a fixed scenario on one basket, not a claim that marketplaces lose money in general.
A 14-day sealed-return window costs about $0.60 an order. Against a $7.68 baseline that needs 8.48% more orders to pay for itself. A broad opened-returns policy costs $2.50 and needs 48.3%. On a 2.00% conversion rate, the sealed option has to reach 2.17%.
There is no reliable electronics-specific estimate of how much a return policy lifts conversion. So rather than assume an uplift, the model states the uplift each option would need and leaves the judgement to the owner. A break-even hurdle is checkable; a predicted uplift would not have been.
On a $90 order a lost card dispute costs about $87 — product, card fee, shipping and the dispute fee — against roughly $15 for an ordinary return. Card networks decide “not as described” disputes whatever the store policy says, so funding a clear defect remedy is cheaper than defending against chargebacks.
The $63 landed cost, the 30% margin, the $8 acquisition cost and the 6% marketplace fee are all analyst assumptions. Sensitivity is stated rather than hidden: a $68 landed cost takes contribution to $2.92, $10 outbound takes it to $3.67, and a $16 acquisition cost takes it below zero. The conclusion survives the first two and not the third.
Everything here models a single $90 multi-item basket. The 391-order break-even is a requirement to cover $3,000 of assumed monthly fixed cost, not a sales projection, and no claim is made that the volume is achievable.
This was unsolicited work for an early-stage operator I was in conversation with. The analysis uses public retailer data and my own assumptions throughout; nothing in it came from inside the business. The name is withheld because the engagement never happened, and the argument does not need it.
Product images come from Samsung, Apple and Logitech press rooms and are used to identify products under discussion. They establish no stock, authorisation or endorsement, and the proposal says so on its face. Commercial reuse rights would have to be cleared before any of it reached a live catalogue.
The written proposal walks the same case in order — starting point, the US retail landscape, the proposed store, the economics, operations, and the decision. A second view is a working storefront prototype, so the recommendation about what a first-time buyer needs to see before paying is demonstrated rather than described.
The slide deck is the version that would have been presented. All three are built from the same model, and every number on all three traces back to the assumptions page.
It is a design demonstration with real manufacturer product photography and no commerce behind it. Nothing is purchasable, no stock is held, and no retailer relationship exists. It is labelled as an independent case study on every screen.
Start first-party at workable landed terms. Add the marketplace when sellers bring demand with them — not because it looks asset-light.
The work I would keep is the discipline of the returns section: rather than claiming a generous policy lifts conversion, it states the uplift each option would need to break even and hands the decision back with the number attached. That is the difference between an opinion and an analysis.
Get one real supplier quote before modelling anything. Every number here is defensible and none is verified, and a single landed cost from an actual distributor would have anchored the whole model to something outside my own assumptions.
An independent, self-directed case study. The company is not named and supplied no data; the analysis draws on published retailer policies and market coverage from Best Buy, Amazon, Walmart, Costco and P.C. Richard & Son, the National Retail Federation, Counterpoint Research and Ingram Micro, each cited in the proposal. Product photography is from manufacturer press rooms and is used for identification only, establishing no stock, authorisation or endorsement. Every financial figure is a planning assumption rather than a supplier quote or a forecast, and the complete arithmetic is published alongside the proposal.