Amazon seller analytics comes in five layers: Seller Central Business Reports, Brand Analytics, third-party dashboards, reconciled accounting data, and an AI that answers questions across all of it. Each layer sees something the one before it could not, and each has a blind spot the next one exists to cover. Business Reports do not know your cost of goods. Brand Analytics does not know your fees. Dashboards know both only if you typed them in. Reconciled books know everything that touched a transaction and nothing about search demand. The AI on top is only as good as the layers it reads.
Sellers who stop at layer one or two run a business on traffic and conversion numbers and find out about margin at tax time.
Seller Central's Business Reports, described on Amazon's own selling partner blog, cover sales and traffic by date and by ASIN, including parent and child item views. The metrics are ordered product sales, units ordered, sessions, page views, unit session percentage (units divided by sessions), and Featured Offer percentage. Data is retained for up to two years, and the Sales dashboard now includes AI-generated commentary on year-over-year changes in sales, units, average selling price, page views, and Featured Offer share.
What it sees: demand and conversion at the ASIN level. What it cannot see: what any unit cost you to buy, ship, or fulfill. A 9 percent unit session percentage on a SKU that loses $2 per unit is a report of an efficient loss.
Brand Analytics, per its page on sell.amazon.com, requires a Professional selling plan and a Brand Representative role for a brand enrolled in Brand Registry. The dashboards fall into four groups: one covering the stages of the shopping funnel, Customer Loyalty Analytics, Search Analytics (Search Catalog Performance, Search Query Performance, Top Search Terms), and Consumer Behavior Analytics (Repeat Purchase Behavior, Demographics, Market Basket Analysis). Search Query Performance tracks query volume, impressions, clicks, cart adds, and purchases for the terms customers use to find your brand.
What it sees: how shoppers search, which terms convert, what they buy together, and how often they come back. What it cannot see: your fees, your ad invoices, your COGS, or your refunds. It tells you the search term "insulated lunch bag" drove 1,200 purchases last month. It cannot tell you whether those purchases made money.
Categories only. Profit dashboards pull orders and fees from Amazon's API and let you enter a cost per SKU. Keyword suites track rank and search volume. PPC dashboards read the ad console.
What they see: whatever Amazon's API exposes plus whatever you typed. What they cannot see: the ledger. A profit dashboard with a single cost per SKU does not run FIFO, so when a new PO lands at a higher unit cost, the margin it shows is stale until someone updates the field. It does not reconcile to the bank, so a settlement discrepancy (a missing reimbursement, a fee charged twice) never surfaces. And it does not carry Shopify, Walmart, or TikTok Shop on the same basis, so channel comparison is a spreadsheet exercise.
Reconciled data means each Amazon settlement is matched to the bank deposit and split into its component transactions: ordered product sales, referral fees, fulfillment fees, refunds, reimbursements, storage charges, ad invoices deducted. Each posts to its own account, against the SKU and the order date. COGS runs on FIFO layers with freight and duty allocated per unit. The two-week statement itself is walked through in Amazon settlement reports explained.
What it sees: every dollar that touched the business, at the grain of a single order, on the same basis across channels. What it cannot see: search demand, competitor prices, or anything that happened before the order. Layer four knows what you earned. It does not know what you could have earned.
ConnectBooks builds this layer, reconciling settlements into QuickBooks or Xero at the transaction level and producing SKU-level profit reports across Amazon, Shopify, Walmart, eBay, TikTok Shop, and Temu.
Illustrative numbers. A stainless lunch box at $27.99, Home and Kitchen, where Amazon's published pricing page lists the referral fee at 15 percent. Fulfillment fee is illustrative; Amazon sets it by size tier and weight.
Layer one says: 4,200 sessions, 380 units, 9.0 percent unit session percentage. Healthy conversion.
Layer two says: the term "stainless lunch box adult" drove 210 of those purchases, with a cart-add rate above category average. Strong search position.
Layer three says: with a cost field of $8.40 entered in March, margin after fees and refunds is 25 percent. Keep advertising.
Layer four says:
Contribution: $10,636.20 minus $1,595.43 minus $2,356.00 minus $2,900.00 minus $3,724.00 minus $839.70 equals negative $778.93.
The SKU converts at 9 percent, ranks well on its main search term, shows 25 percent margin on the dashboard, and lost $779 last month. Three layers said healthy. The layer that reconciled the settlement and ran FIFO said otherwise. The difference between the dashboard's cost field and the FIFO cost, $1.40 per unit on 380 units, is $532. The rest is ad spend the dashboard was not netting against this SKU.
The fifth layer is an analytics AI that reads layer four and answers plain-language questions about it. Crunch, inside ConnectBooks, takes a seller's question about profit, margins, advertising, inventory, fees, or cash flow and runs it through five steps: find what changed, compare the right periods, find the products responsible, explain why, and recommend what to do next.
For the lunch box, the question "which products had a negative contribution margin last month" returns the SKU, the $779 loss, the FIFO cost change as the largest driver, and the ad spend as the second. A follow-up, "did this SKU make money in Q1," gets a second answer on the same data. That is the difference between a dashboard, which shows tiles someone designed, and an AI, which answers the question you had.
It reads layer four. It does not read layers one and two. Crunch cannot tell you that "stainless lunch box adult" is losing rank to a competitor, because search data is not in the books. Pull that from Brand Analytics.
It cannot fill in a cost that was never recorded. If the June PO has no duty allocated, the COGS is understated and the AI's margin is wrong in your favor. A well-built tool flags the gap. A badly built one estimates.
It does not know why. It knows the return count on the lunch box rose. Whether the lid seal failed or the listing overpromised is a phone call and a sample inspection.
It takes no actions in Seller Central. It answers and recommends. Pausing the campaign, changing the price, or filing a removal order is yours.
If you run layer one only, add layer four before adding anything else. The dashboards in layer three are cheaper and faster to set up, and they will tell you a SKU is profitable for months after it stopped being so. Reconciled data is slower to set up and does not lie about cost. Put it in, then put the AI on top of it, and ask it the question you have been avoiding.
Running an e-commerce business comes with plenty of challenges, but ConnectBooks is here to make your life easier. With real-time insights, seamless integrations, and detailed tracking of your profitability and inventory, you can stay ahead of the game. Whether you’re selling on Amazon, Shopify, Walmart, TikTok or eBay, ConnectBooks helps you manage your finances with 100% accuracy and confidence, so you can focus on growing your business.
Ready to level up? Start making smarter, data-driven decisions every step of the way. Try ConnectBooks Free Today or Schedule a Demo
