Platform fee schedules are public. Read them as a system.

Most channel comparisons we are shown open with one number: the commission rate. It is the number platforms lead with, and it is the least informative line on the page. A platform's published pricing is not a rate at all. It is a cost structure, and the structure is what decides whether a category is viable on that channel.

So stop comparing rates, and read every schedule in the same fixed order instead. Below is the order we use, and what each layer tends to hide.

Read the schedule in four layers

1. The entry cost. Subscription tiers, listing fees, and any minimum that applies whether or not you sell. This layer is fixed cost, so it behaves completely differently from the rest: it punishes low volume and disappears at scale. A seller doing a pilot and a seller doing steady volume are, in effect, quoted different prices by the same published schedule.

2. The per-transaction cost. Commission, payment processing and any per-order handling. This is the layer everyone compares, and the only one that scales cleanly with revenue. Commission and payment processing are usually disclosed on different pages by different entities — the marketplace takes one, the payment provider takes another, and adding them up is the reader's job.

3. The fulfilment and storage cost. Where the platform also handles logistics, this layer is priced by weight, dimension and time held, which makes it the layer most sensitive to what you actually sell. Two sellers on identical commission terms can land far apart here purely on product dimensions, and a slow-moving SKU can accumulate storage cost indefinitely while contributing nothing.

4. The discretionary cost. Advertising, placement, promotional participation and anything else nominally optional — and nominally is the operative word. On a platform where organic discovery is thin, the ad spend needed to stay visible is a real cost of selling there. Treating it as marketing budget rather than channel cost is the most common way a channel model comes out wrong.

Two things the layers reveal that a rate cannot

Where the break-even sits. Layers 1 and 3 carry fixed and unit-linked costs; layer 2 is proportional. That mix determines the volume at which a channel turns profitable — a different question entirely from which channel is cheapest. A high-subscription, low-commission schedule and a no-subscription, high-commission schedule cross at some volume, and that crossing point can sit on either side of the volume you actually do.

Which categories the platform is built for. Weight-and-dimension pricing is a statement about what a platform wants to carry. A schedule that prices storage aggressively is telling you it is built for fast turns. Read that way, a fee page is a positioning document.

A fee schedule is a platform describing the seller it wants. Read it as positioning, not just as arithmetic.

What to actually do

Pull the current pricing page for each channel under consideration, on the same day, and record the four layers separately in one sheet — not one blended percentage. Then run your own top three SKUs by dimension and turn rate through each schedule, rather than an average basket. Averages hide exactly the dimension and velocity effects that layer 3 is built around.

Two cautions on sourcing. These schedules change, and they change by region — the page a platform serves you is often not the page it serves a seller in another market, so record the date and the regional page you read. And no secondary summary, this one included, substitutes for the primary page at the moment you are deciding. Comparison articles age badly, and fee pages are precisely the kind of content that gets updated without announcement.

So what

The practical output of this exercise is not a ranking of platforms. It is a break-even volume per channel per SKU class, which is a number you can test against your actual pipeline. If you cannot state that number for a channel you are already selling on, the channel is being run on the headline rate — and the headline rate was never the cost.

Our companion note on where automation belongs in this kind of recurring data pull is at AI in the research workflow.