Five reconciliation errors account for most of the gap between what an ecommerce seller believes they earned and what their books eventually say: treating the marketplace deposit as revenue, ignoring settlement periods that straddle month end, netting refunds against sales, expensing inventory at the moment of purchase, and leaving collected sales tax sitting in an income account. Any one of them is small in a single month. Repeated across a year and three marketplaces, each is worth thousands in misstated profit, overpaid tax, or a discount applied by a buyer who no longer trusts the numbers.
1. Booking the deposit as revenue
A marketplace deposit is not a sale. It is what remains after the marketplace subtracts its cut, holds a reserve, refunds a few customers, charges storage, and settles advertising. Recording the deposit as a single revenue line collapses all of that into one number and makes gross margin permanently unknowable.
The size of the distortion is not trivial. Amazon’s published referral fees run from 5 percent to 45 percent depending on category, with most categories at 15 percent and a minimum of $0.30 per item, according to Amazon’s seller pricing page as of the 2026 schedule. A seller who books net deposits is reporting revenue roughly 15 percent light and carrying no fee expense at all. Gross profit looks fine. Gross revenue does not match the 1099-K. The first person to notice is usually an accountant in March.
The fix is structural rather than clever. Revenue posts at gross, each fee category posts to its own expense account, and the deposit becomes the reconciling total at the bottom.
2. Ignoring the settlement period that straddles month end
Marketplace settlement periods do not respect calendar months. Amazon settles roughly every two weeks, which guarantees that several times a year a settlement will open in one month and close in the next. Sellers who post the whole settlement in the month it lands are moving revenue across period boundaries without meaning to.
In a flat month nobody notices. In November and December, when volume doubles, a settlement that straddles the month end can shift a meaningful share of Q4 revenue into the wrong period. That distorts every month over month comparison a seller makes, and it misstates the year end cutoff that a tax return depends on.
The remedy is to split the settlement at the period boundary and accrue the open portion. It is tedious by hand, which is why it is usually skipped, and why tooling that reconciles settlement to deposit matters more than it sounds. Platforms that connect marketplace data to a general ledger for this purpose are listed at https://www.connectbooks.com/quickbooks-integrations.
3. Netting refunds and reimbursements into one line
Refunds and reimbursements look similar on a settlement report and mean opposite things. A refund is revenue reversing because a customer sent goods back. A reimbursement is the marketplace paying a seller because it lost or damaged inventory it was holding. Netting them produces a single number that describes neither event.
Two things break. Return rate becomes invisible, so a product line with a quality problem hides behind reimbursements on unrelated SKUs. And inventory stops reconciling, because a refunded unit usually comes back into sellable stock while a reimbursed unit never does. A seller who nets the two will count units they do not physically have.
Keep contra revenue for refunds, other income for reimbursements, and an inventory adjustment for the units themselves. Three entries, three different stories.
4. Expensing inventory when you buy it
Inventory is an asset until it sells. Paying a supplier in February for goods that sell in June does not create a February expense. It converts cash into inventory, and the expense appears as cost of goods sold in June when the revenue does.
Sellers who expense at purchase produce books that swing wildly and mean nothing. The month a container lands shows a catastrophic loss. The month it sells through shows implausible profit. Neither figure reflects the business, and the pattern is immediately recognizable to anyone conducting diligence.
The IRS treats this as a method of accounting question rather than a preference, and Publication 538 sets out the rules for accounting periods and methods. There is a small business exception: under section 448(c), a taxpayer whose average annual gross receipts for the prior three years do not exceed $32,000,000 for taxable years beginning in 2026 is treated as a small business taxpayer, a threshold set in section 4.30 of Revenue Procedure 2025-32. That exception governs what is permitted on a tax return. It does not make expense at purchase a useful way to run a business, and a seller who wants to know which products make money needs inventory carried as an asset regardless of what the return allows. A seller near the threshold should raise it with their own accountant rather than a blog.
5. Leaving collected sales tax in income
Sales tax collected from a customer is money held on behalf of a state. It is a liability from the moment it is collected until it is remitted. Recording it as income inflates revenue, inflates apparent profit, and creates a tax bill on money that was never the seller’s.
Marketplace facilitator laws complicate this rather than removing it. In most states the marketplace now collects and remits on the seller’s behalf for marketplace sales, but the seller’s own direct to consumer channel is usually still their own responsibility, and the two flows land in the books looking similar. A seller running Shopify alongside Amazon frequently has one channel where the tax is handled for them and one where it is not.
Rules differ by state and change often, so the registration question belongs with a state’s department of revenue or a tax professional rather than with any software. The bookkeeping question does not change: collected tax goes to a liability account and leaves it only when it is remitted.
What these five have in common
None of these errors is a mistake of arithmetic. Each is a decision about what an event means, made once, usually early, and then repeated automatically for years. That is why they compound, and why they are expensive to unwind later.
The practical test is whether a seller can answer one question without opening a spreadsheet: what did this SKU earn last month after every cost that touched it. A seller whose books are built on the five errors above cannot answer it, and generally does not discover that until a lender, a buyer, or an accountant asks.
Fixing the five is not a weekend project once a few years of history exist. It is close to trivial in a seller’s first year. The cost of the delay is the entire argument for doing it early.
