A question an Amazon analytics AI can answer names a product or product group, a period, a metric, and a comparison. "Which of my top 20 SKUs by revenue had a contribution margin under 10 percent in August" has all four. "How are my products doing" has none, and the answer you get back will be a summary of whatever the model thought you meant.
The second requirement sits underneath the first. The data the AI reads has to contain the metric you asked for, at the grain you asked for it. Ask about margin by SKU and the books need landed COGS by SKU, fees by SKU, ad spend by SKU, and refunds by SKU. Ask about any of those in a ledger that holds Amazon as one deposit line and COGS as one monthly entry, and the AI will either decline or compute a number from averages that is wrong for every individual product.
The generic version of this advice, for any AI financial assistant on any set of books, is in how to write questions for an AI financial assistant. This article is the Amazon-specific version, because Amazon data has its own traps.
Product. Name the SKU, the ASIN, the parent listing, or a group you have defined ("the 12-ounce variations," "everything in the pet category"). Amazon reports by ASIN and child ASIN; your accounting system reports by SKU. If the two are not mapped, the AI is guessing which is which.
Period. Name the dates. "Last month" is fine if the AI knows today's date. "Q2" is fine if your fiscal quarter matches the calendar. "Since the price change" is only usable if the price change date is in the data. Settlement periods do not align to calendar months; a question about "August" needs data reconciled by order date, not by settlement deposit date.
Metric. Say which number. Revenue, units, contribution margin, gross margin before ads, net after ads, return rate, ACoS, TACoS, cash contribution. Sellers who ask about "profit" without saying which one get whichever definition the tool defaults to.
Comparison. Against what. The prior period, the same period last year, a target, or another product. A margin of 14 percent means nothing alone. Against 22 percent last quarter, it is a problem.
Before asking anything about profitability by product, check that these exist in the books the AI reads:
ConnectBooks builds this layer by reconciling each Amazon settlement into QuickBooks or Xero at the transaction level and running FIFO costing per unit. The profit reports it produces are the data an AI needs to see.
Illustrative numbers throughout.
Question: "Which of my top 20 SKUs by revenue had a contribution margin under 10 percent in August?"
What the AI does: ranks SKUs by August revenue, takes the top 20, computes contribution for each (revenue minus referral fees, fulfillment fees, ad spend, refunds, storage, and FIFO COGS), and filters.
Answer: one SKU, a 4-pack of glass jars. August revenue $9,450.00. Total variable costs $8,760.00: referral fee $1,417.50, fulfillment $2,130.00, ads $1,640.00, refunds $520.00, storage $90.00, COGS $2,962.50. Contribution $690.00. Margin 7.3 percent.
Why it holds: every cost line came from a source document. The fees came from settlement lines, the ads from the advertising invoice, the COGS from PO layers. Change the COGS to a purchase-price-only figure of $2,400.00 and the margin jumps to 13.3 percent, and the SKU drops off the list. That is how an unreconciled ledger hides a problem.
Question: "Compare the contribution margin on the ceramic planter in Q2 against Q1, and tell me what drove the change."
What the AI does: pulls both quarters for the SKU, computes contribution for each, and decomposes the difference by cost line.
Answer:
Driver: landed cost per unit rose from $6.10 to $7.05 under FIFO when the Q1 inventory layer ran out in April and a PO that shipped with a freight surcharge began to sell through. Q2 units were 1,300. The cost increase alone is $0.95 times 1,300, or $1,235.00, which explains 98 percent of the decline. The remaining $21.00 is fee mix.
What the AI cannot add: whether the freight surcharge is permanent. It knows the PO landed at $7.05. It does not know your forwarder's next quote.
Question: "How much did returns on the fleece throw cost me last month, all in?"
What the AI does: counts refunded units, sums refund amounts, adds fulfillment fees not recovered, adds COGS on units returned in unsellable condition.
Answer: 1,000 units sold at $30.00. 60 refunded (6.0 percent). Refunds $1,800.00. Fulfillment fees on the 60 returned orders at an illustrative $3.50 per unit, assumed not recovered: $210.00. 18 units graded unsellable and written off at $7.00 landed cost: $126.00. Total cost of returns: $2,136.00, or 7.1 percent of revenue on the SKU.
Why the grain matters: a ledger that books refunds as one monthly total cannot answer this for the fleece throw alone. It can only say returns cost the account $X. The seller's decision (is the throw's packaging causing damage in transit?) needs the SKU-level figure.
"Why is revenue down" with no period. The AI picks one.
"What should I reorder" with no forecast data and no lead times in the system. The AI can rank sell-through; it cannot invent your supplier's schedule.
"Is this ad campaign working" where the campaign has been running nine days and the attribution window is fourteen. The data is not in yet.
"How much tax do I owe" of any kind. Ask a CPA.
Crunch is the analytics AI inside ConnectBooks. It reads reconciled Amazon, Shopify, Walmart, TikTok Shop, and eBay data and runs each question through five steps: find what changed, compare the right periods, find the products responsible, explain why, and recommend what to do next. The three walkthroughs above are the shape of question it is built for.
It does not know what it has not been given. A SKU with no landed cost on file gets flagged, not estimated. It does not take actions in Seller Central. And it does not know why a supplier's price moved or whether a return spike is a packaging defect; it points at the number, and you make the call.
Write the question with a product, a period, a metric, and a comparison. Confirm the data underneath holds the cost lines. Then ask.
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
