Read this first: Seller Labs makes the product reviewed here, the Seller Labs Amazon MCP, and we wrote this review. So do not take our word for the good parts. Check the limits below, and run the test near the end on your own numbers.
You already have a profit spreadsheet. It pulls the transaction report, the ad report and your landed cost into one row per SKU, and the number it gives you is right. It is right for the day you built it. It also costs you an afternoon each time, so you rebuild it monthly, for your top sellers only. The sheet works. What you pay to keep it current, and how many SKUs it leaves out, is what this review looks at.
The one decision: keep the spreadsheet or switch
Switch only if the monthly cost of rebuilding your sheet is higher than the subscription, or the sheet no longer covers the SKUs you make decisions on. Otherwise keep the sheet.
That takes two things: the true-profit method done properly, and an honest count of your hours. Both are below. Neither needs our product.
The spreadsheet method, done properly
True profit for a SKU is what is left after Amazon’s fees, refunds, storage, ad spend and your landed cost. Not sales minus fees. Six lines, three downloads and one number that only you know.
| Line | Where it comes from | What to sum |
|---|---|---|
| 1. Product sales | Seller Central → Payments → Reports Repository → Transaction report (CSV), custom date range | The Order rows for the SKU: the column that holds the item price charged, before fees |
| 2. Amazon fees on those orders | Same file | On the same Order rows: the referral (selling) fee column and the FBA fee column |
| 3. Refunds | Same file | The Refund rows for the SKU: the product sales refunded, less the fees Amazon hands back on those rows. Sum the sales and fee columns yourself |
| 4. Storage | Reports → Fulfillment → Monthly Storage Fees | Every row for the product, for the same month |
| 5. Ad spend | Amazon Ads console → Measurement and Reporting → Sponsored ads reports → Sponsored Products, Advertised product | Spend for the advertised SKU or ASIN, same dates |
| 6. Landed cost | Your own records: supplier invoice, freight, duty, prep | Cost per unit × units sold |
- Match columns by meaning, not by name. Check the header names, and whether fees show as negative numbers, in your own file before you write a formula.
- Amazon’s own Seller Forums walkthrough of date range reports (posted by Amazon staff in 2026) gives the path as Payments → Reports Repository, offers a Summary PDF or a Transaction CSV, and says reports typically generate within three hours.
- Take ad spend from the ad report, never from the transaction file. Finaloop’s 2020 guide to the transaction report notes that advertising is charged under other transaction fees. Leave those rows out of a per-SKU sheet or you count ads twice. PPC Ninja’s January 2026 guide gives the ads path as Measurement and Reporting → Sponsored ads reports.
- Storage runs a month behind. NovaData’s 2026 guide to Seller Central fees says the Monthly Storage Fees report posts after the 15th, and Seller Assistant’s storage guide says the charge is for the previous month. So an honest close for August happens in the second half of September.
A worked example: one SKU, one month
Illustrative example
One SKU at $28.00. In August it sells 400 units. Every number here is invented to show the method.
Sales: 400 × $28.00 = $11,200.00.
The Order rows show $1,680.00 in referral fees and $2,200.00 in FBA fees. $11,200.00 − $1,680.00 = $9,520.00. Then $9,520.00 − $2,200.00 = $7,320.00.
Refunds: 17 units come back. 17 × $28.00 = $476.00 in sales refunded. In this example the Refund rows hand back $3.36 in fees on each of those units: 17 × $3.36 = $57.12. $476.00 − $57.12 = $418.88. Then $7,320.00 − $418.88 = $6,901.12.
The short version stops here and subtracts product cost. Landed cost is $9.40 a unit: 400 × $9.40 = $3,760.00. $6,901.12 − $3,760.00 = $3,141.12. That is $7.85 a unit, a 28.0% margin.
Now the two lines that live in other reports. Storage for the product: $64.00. $6,901.12 − $64.00 = $6,837.12. Sponsored Products spend on the SKU: $1,344.00. $6,837.12 − $1,344.00 = $5,493.12.
True profit: $5,493.12 − $3,760.00 = $1,733.12. That is $4.33 a unit and a 15.5% margin.
Same SKU, same month. The short version said $3,141.12 and the full version says $1,733.12. The $1,408.00 between them is storage plus ads, and it is the difference between a SKU you would reorder deep and one you would look at twice.
Now the cost of knowing that. Say the three downloads take 30 minutes and each SKU takes 15 minutes to filter, sum and check. For 12 SKUs: 12 × 15 = 180 minutes, and 30 + 180 = 210 minutes, or 3.5 hours. At $60 an hour for your own time, 3.5 × $60 = $210.00 a month. Over a year, 12 × $210.00 = $2,520.00. The rest of the catalog never gets checked.
Put your own hours and rate into that last paragraph. That is your side of the comparison. The other side is the figure on our pricing page for your revenue bracket, plus the AI subscription you would need.
What the Seller Labs Amazon MCP does with the same question
The Seller Labs Amazon MCP connects your Amazon data, as Seller Labs stores it, to Claude or ChatGPT. You ask the profit question in a sentence, and the AI queries the tables instead of you filtering a CSV. Here is what we think it does well.
- The joins are already done. The data includes sales and orders, advertising campaigns and keywords, profitability by SKU and ASIN, inventory and product reviews. Tables whose names start with
sl_hold Seller Labs calculations on top of the Amazon data, including daily profit per SKU. - Your cost lives in one place. You can update COGS by asking, one marketplace at a time, and you see a preview before anything is saved. That change recalculates past profit up to 3 years back.
- Order history comes with it. At connection, order and sales history reaches back about 2 years.
- It covers more than the US. 10 marketplaces: US, Canada, Mexico, UK, Germany, France, Italy, Spain, Japan and Australia.
- It is not read-only. When you ask, it can create and manage Sponsored Products campaigns, change bids and budgets, and add or remove negative keywords. Claude acts only when you tell it to. Nothing changes on its own.
If you have it connected, these two prompts ask for the six-line sheet. The first only reads. The second changes your account.
For August, show true profit per SKU for my top 12 SKUs by sales: sales, refunds, Amazon fees, storage, ad spend, COGS and profit. Report each marketplace separately and never sum across marketplaces. Before the numbers, tell me which tables you used, the latest date each one has data for, and which SKUs have no COGS on file.
Update COGS for SKU [your SKU] to [your landed cost] in [marketplace name]. Show me the preview and wait for my yes before you save.
The second prompt writes to your Seller Labs account. Read the preview before you confirm.
What it cannot do
Every line below comes from our own documentation or from support cases.
- It is not real time. From support cases, advertising data has run roughly 48 hours behind and traffic data about 3 days behind.
- Ad history is short at the start. When you connect, advertising history is capped at 95 days. Orders go back about 2 years, so a year-over-year profit view will have sales with no ad spend for the older months. Your old spreadsheet tabs may be the only per-SKU record of that ad spend. Keep them.
- Your landed cost is still your job. Through the AI you can set COGS, lead time and product group. There is no separate freight or duty field there, so those go into one COGS number per SKU, which you work out. And until you enter your own cost, the SKU Economics view in Profit Genius assumes COGS is 25% of the sale price. A profit number built on that default is a guess.
- A COGS update through the AI always restates history. There is no “from today forward” option in chat. If your cost rose in August, updating it this way reprices earlier months too. The Bulk Update Products page in Seller Labs has an Update Historical COGS checkbox for that case.
- The profit tables have had gaps. In 2026 we handled support cases where the profit tables stopped filling for a date range while orders kept arriving. Nothing alerts you when that happens. Our Dev team investigates and backfills where Amazon allows it.
- It cannot change a price or a listing. When a SKU turns out thin, the fix on price, title or images still happens in Seller Central or your repricer. Sponsored Brands and Sponsored Display campaigns can be viewed, not managed.
- It does not run your ads for you. We do not offer fully automated PPC management, and nothing here promises more sales.
- Some data is not there. From support cases: Subscribe and Save data, Amazon Marketing Cloud and keyword rank tracking were not available (category sales rank is in the data), and a Vendor Central account could not be connected. Brazil is not one of the 10 marketplaces.
- The connection can drop. A small number of customers have reported the connector disconnecting from Claude. It is uncommon, and we have not found the root cause.
Keep the spreadsheet or switch: the table
| Your situation | Verdict | Why |
|---|---|---|
| Under about 10 SKUs, monthly close, the sheet takes an afternoon or less | Keep the sheet | A few hours of your own time is hard to beat, and you know every row |
| You only want the AI connection, not the tools | Check the plans first | Seller Labs plans differ in what comes with the connection. Check the pricing page for the smallest plan that includes it before you pay for tools you will not open |
| You sell on Walmart, Shopify or TikTok Shop and want one P&L | Keep the sheet | Seller Labs is Amazon only |
| You sell through Vendor Central | Do not buy | In a support case, a Vendor Central account could not be connected |
| You need same-day numbers to make intraday calls | Do not buy for that | In support cases, advertising data has run about 48 hours behind and traffic about 3 days |
| You want AI to reprice or rewrite listings | Do not buy for that | No price or listing changes |
| Dozens of SKUs, and the sheet covers the top 12 | Switch is worth testing | The unchecked SKUs are where the surprises are |
| Your rebuild cost is above the pricing-page figure plus your AI plan | Switch is worth testing | The tool pays for itself on hours alone, before any better decision |
| You already pay for the Genius Bundle | Use it | It is included while your subscription is active. You are paying for it now |
One number turns the rebuild-cost row into arithmetic: take the monthly subscription total and divide it by your hourly rate. The answer is the hours the tool has to save each month. If your sheet takes fewer hours than that, keep the sheet.
Which SKUs to check first
Do not audit the catalog. Whichever method you use, start with the SKUs where the short version of profit is most likely to be wrong.
- Highest ad spend. Sort the Advertised product report by spend and take the top 5 SKUs. Ads were most of the gap in the example.
- Oldest stock. Slow sellers carry storage that a sales-minus-fees view never shows.
- Most refunds. Count Refund rows per SKU in the transaction file. A SKU that loses 1 unit in 10 to refunds needs its own line.
- Anything you are about to reorder. A purchase order is the most expensive decision the number feeds.
If you do try the AI route, test it against the sheet before you trust it. Pick 3 SKUs you have already closed by hand for one month and ask for the same six lines. Where the two disagree, ask which dates and which tables the answer used. Agreement on those 3 SKUs is your evidence. Our word is not.
When this breaks
- Dates do not line up. Amazon’s forum walkthrough describes a transaction status column (deferred or released) and a release date that you may see in your Transaction report. A sale in the last days of the month can be paid out in the next one. Pick one date basis and hold it across all six lines. With the AI, ask which basis a number uses before you compare.
- Returned units. The example charges landed cost on all 400 units and treats refunded units as lost. If most of your returns go back to sellable stock, that overstates cost. Adjust units, and say so on the sheet.
- Brand and display ads. The Advertised product report covers Sponsored Products. Sponsored Display has its own Advertised product report; pull it too. Sponsored Brands spend has no SKU on it, so you allocate it by your own rule or hold it at brand level.
- Old stock. Aged-inventory surcharges sit in a separate report, and the six lines leave them out. Add them for anything that has sat for months.
- Cost changes mid-month. One COGS figure per SKU cannot describe two shipments at two prices. Use a weighted cost for the month.
- More than one marketplace. Currencies differ. Keep one tab, or one answer, per marketplace.
- Overheads. Software, payroll and your accountant are not in either version. This is SKU profit, not your books.
Frequently Asked Questions
Does the Seller Labs Amazon MCP work with ChatGPT or only Claude?
Both. It connects to Claude or ChatGPT. For Claude you add it as a custom connector and sign in to Seller Labs; there is no API key. Claude needs a paid plan from Anthropic, which is a separate bill from ours. The steps are in our connect guide, linked under Keep Reading.
Can two people on my team use it?
Yes. Two users on the same Seller Labs account can each connect their own Claude account. Each separate Seller Labs account needs its own connection.
What happens to my data if I cancel?
Access stops when the subscription is no longer active, and syncing stops with it. Stored history is not deleted. If you come back, the gap can refill, but advertising only reaches back 95 days, so the earliest days of a longer gap cannot be recovered.
Running the check across every SKU
The six-line method above works tonight with three downloads and no subscription. The Seller Labs Amazon MCP is the same check asked as one question across every SKU and every marketplace you sell in, within the limits listed above.
Related Reading
- How to Connect Claude to Your Amazon Seller Account in 10 Minutes (No Code)
The setup steps, if the table above said the switch is worth testing. - Is Your Amazon PPC Keyword Profitable? The 60-Second Test
The same true-profit idea applied to one keyword’s bid. - How Do I Know Which SKUs Are Hurting My Profit Margins?
What to do with the SKUs that come out thin.
