Channel-level profit and loss is the most popular report in ecommerce accounting and the least useful one. It tells you Amazon made $340,000 and Shopify made $110,000 last quarter. It does not tell you a single thing you can act on, because no seller has ever fixed a business by deciding to have less Amazon. My position, after enough closes to have opinions about it, is that channel-level reporting should be a reconciliation artifact and nothing more. The decision-grade unit is the SKU.
The counterargument is real, and it gets a section of its own further down.
Why channel P&L feels right
It mirrors how the money arrives. Amazon deposits a settlement. Shopify deposits a payout. Walmart deposits on its own cadence. Each of those has its own fee structure, its own timing, its own reconciliation headache. Grouping the income statement the same way makes the bank reconciliation tractable, and there is nothing wrong with that. Grouping by channel is also how most sync tools ship out of the box, so it becomes the default without anybody choosing it.
It also flatters a certain kind of thinking. A channel report produces sentences like “Amazon is 74 percent of revenue” that sound strategic in a meeting. They are not strategic. They are descriptive, and they describe something you already knew.
The problem: channels are cost buckets, not profit centers
Take two sellers with identical $500,000 quarters on Amazon. Seller A sells consumer electronics accessories. Seller B sells apparel. Per Amazon’s published selling fee schedule, the referral fee on computers and consumer electronics runs 8 percent, while clothing and accessories priced over $20 carry 17 percent. Before either of them has paid for a single unit of inventory, warehousing, or advertising, there is a nine-point gap in gross margin that has nothing to do with how well they run their business.
Now roll those two sellers into one company with a channel-level report. The Amazon line shows a blended margin that describes neither product. It is an average of two unrelated economics, and averages of unrelated things are noise wearing a suit.
This gets worse as the catalog grows. A seller with 300 SKUs almost always has a distribution where a modest number of SKUs generate all the profit, a large middle band roughly breaks even, and a tail actively loses money after storage, returns and advertising. The channel report shows one number. That number is the sum of a business you should scale and a business you should kill, and it hides both.
Four decisions channel-level reporting cannot support
Reorder. Purchase orders are placed per SKU with a specific cash outlay and a specific lead time. Knowing that Amazon was profitable last quarter tells you nothing about whether to put $40,000 into a reorder of a particular unit. You need that unit’s landed cost, its true fee load, its return rate and its sell-through.
Price. A price change happens on a listing. To decide whether raising a price by $2 makes sense, you need to know the current contribution per unit and the elasticity of that specific item. A channel margin percentage cannot get you there.
Advertising. Ad spend is allocated per campaign, which usually maps to a product or a small group of them. A campaign can look fine on ACOS while destroying contribution margin, because ACOS is measured against revenue and contribution is measured after every fee, the return rate and COGS. You only catch that at the SKU.
Kill decisions. Discontinuing a product is the move that changes a catalog business fastest, and the one owners make least often, because the channel report never surfaces the candidate.
The counterargument, which is not nothing
SKU-level accounting is expensive to maintain and easy to get wrong. Marketplace fees do not all arrive attached to an order. Storage fees, long-term storage surcharges, inbound placement charges, coupon fees, subscription fees and reimbursements land at the account level and have to be allocated by some rule you invent. Every allocation rule is an assumption, and an assumption applied to 300 SKUs produces 300 slightly wrong numbers instead of one roughly right one.
There is a real school of thought that says a clean channel-level statement plus a separate operational profit model is better than a single system that pretends to know per-unit truth it cannot know. I disagree, but I understand it. If your allocation logic is arbitrary, SKU-level reporting gives you false precision, and false precision is worse than acknowledged imprecision because people act on it.
The resolution is to be explicit about which costs are directly attributable and which are allocated, and to keep them in separate columns. Referral fees, fulfillment fees, returns and COGS attach to a unit. Storage, subscription fees and overhead do not. A SKU report that shows contribution margin before allocation, then allocated overhead as a clearly labelled second layer, is honest. One that blends them into a single “net profit per SKU” figure and does not say how is not.
What a usable reporting stack looks like
Three layers, in this order.
The financial statements stay at the entity level, because that is what the tax return, the lender and the eventual buyer look at. Consistency with IRS guidance on accounting periods and methods matters more here than analytical elegance, and a seller carrying inventory generally has to use an accrual method for purchases and sales anyway.
Channel-level detail exists underneath, but its job is reconciliation. Its question is “does the deposit that hit the bank match the settlement that produced it,” not “which channel should we invest in.”
SKU-level contribution sits underneath that and drives operating decisions. Software in this category, including tools such as ConnectBooks that push per-SKU marketplace detail into QuickBooks or Xero, exists specifically because the third layer is the one spreadsheets cannot sustain past a few dozen products.
The test
Pull the last report you produced and ask what you would do differently if the number were 20 percent higher or 20 percent lower. If the honest answer for a channel report is “nothing,” you have learned what that report is worth. Run the same test on a SKU-level contribution report and you will usually find three or four items you would reorder differently, and one you would stop buying entirely.
That is the difference between a report that describes the past and a report that changes what you do on Monday. Channel-level P&L is the first kind. It belongs in the reconciliation file, not in the meeting.
None of this requires expensive software to start. Export a month of order-level data, attach landed cost per unit, subtract the fees that arrived attached to those orders, and sort ascending. The bottom of that list will surprise you, and it will not be a channel.
