Combine Amazon Ads data with your own data in one prompt

How do you combine Amazon Ads data with your own data in one prompt?

You connect Xnurta MCP, put your own numbers in the same chat, and ask one question that needs both. Xnurta MCP connects Claude, ChatGPT and other MCP clients to your Xnurta data. The assistant pulls performance and settings from Xnurta, reads the rows you attached, does the math and answers in one table.

Three moves:

1. Connect once. Connect Claude or ChatGPT to Xnurta MCP walks through it, and the endpoint is https://mcp.xnurta.com/mcp.
2. Put your numbers in the chat: margin by ASIN, a goal such as a TACOS ceiling, or the date your next shipment lands.
3. Ask for the join. Name the ad account, the window and the output you want back.

Why bother? Because ACOS alone can mislead. A campaign at 35 percent ACOS on a product with a 50 percent margin makes money. A campaign at 20 percent ACOS on a 15 percent margin loses it. ACOS cannot tell these apart, because it ignores your COGS.

What does Xnurta MCP supply, and what do you bring?

Xnurta MCP supplies the advertising side, and you bring the business side. Reading is all it does here: queries on performance, settings and the operation log never modify your account. Everything in the right column is yours to paste or attach.

The history reaches back about 15 months across 22 marketplaces. Data follows the Xnurta platform's update cadence, so it is current to the last refresh, not to the second.

How do you get your own data into the chat?

You attach a file or paste the rows, then tell the assistant what each column means. A CSV or spreadsheet export works. A few lines of pasted text work too, if you only have ten ASINs.

Four habits save most of the cleanup:

• Put an ASIN or SKU column in every file. Xnurta data is keyed by ASIN, parent ASIN and SKU, so that column is the join.
• Add a date column when the data is weekly, such as promo dates or store sales.
• Send totals by SKU or week, not shopper-level records. The math does not need individual customers.
• Ask the assistant to repeat the first five rows it read before it calculates anything. A shifted column shows up in ten seconds.

Agencies should keep one chat per client, so one client's margins never sit next to another's.

Which campaigns lose money once margin is counted?

Hand over margin before ad spend and let the assistant pull spend and ad sales. Break-even ACOS equals margin before ads, so a product with a 38 percent margin breaks even at 38 percent ACOS. Any campaign above that line loses money on each order.

Here is the unit margin before ad spend for each ASIN in product line [product line name]: [paste or attach]. For ad account [account name], pull the last [30] days of spend and ad sales by campaign. Based on actual margin, which campaigns are losing money? For each one, say whether I should cut it, fix it or increase its budget, and show the calculation. Do not change anything.

Flag every campaign running above its ASIN's break-even ACOS. Show the gap in percentage points and list the ten widest gaps first.

Use a complete period for this one. Data is not real time, and a short window that ends today can flatter or punish a campaign for the wrong reasons.

Are you on track for your TACOS goal?

You tell the assistant the goal, and it checks where your product lines stand against it. The goal is your number. Xnurta holds total sales and TACOS by ASIN, and product line information, but it cannot know what ceiling your business can live with.

My goal is to keep TACOS for ad account [account name] below [12]%. Pull the last [30] days, find which product lines are pushing TACOS up, and rank the adjustments by how much each would help. Show current TACOS next to the goal for every product line.

If you want this as a routine, the Monthly Ads Report skill includes goal attainment. Name the skill you want and the assistant guides you through installing it.

How do you line up ad spend with stock that is running low?

You combine the inventory fields Xnurta already holds with a fact you supply: when your next shipment arrives. Xnurta's ASIN settings include FBA inventory and inventory status, so the assistant starts from those. You add the purchase order dates.

Picture an ASIN with no orders in 60 days, a margin of about 21 percent and 50 units landing next week. Cutting budget might be wrong. Raising it might burn money on a listing about to run dry. The assistant can weigh both once it has the dates.

ASIN [ASIN] has had no orders in the last [60] days. Its margin is about [21]% before ads, and the next [50] units arrive on [date]. Using its inventory status and the ad data in ad account [account name], tell me how to adjust ad spend to get orders without running out before the shipment lands.

Here is my open purchase order sheet with arrival dates: [attach]. For ad account [account name], list ASINs whose FBA inventory looks thin against the last [30] days of sales. For each one with no shipment arriving in time, list its active campaigns and suggest a budget cut.

Here is my promo calendar with deal dates and ASINs: [attach]. For ad account [account name], show how those ASINs performed in earlier windows with similar dates. Suggest a budget per campaign for each deal window. Do not change anything yet.

The Product Diagnosis skill also checks ad eligibility and inventory, if you would rather start from a ranked list of products than from a question.

How do you turn the answer into changes?

You ask for a plan first, then decide. Reading never touches your account. Changing it needs write access, and write access is on only for accounts your CSM has enabled.

Writing comes in three kinds, each with its own permission scope: managed group management, ad-entity editing and Sponsored Products campaign creation. Budget cuts on campaigns fall under ad-entity editing. A tighter target ACOS on an AI managed group falls under managed group management. Money, bids and archiving never run straight away. The assistant shows a preview of the objects and the old value against the new value, and submits only after you confirm.

You stay in control after the change, too. Ask the assistant for the operation log, which records human and AI actions on budgets, bids and managed groups, as Who changed my Amazon campaign? shows. And if a campaign just needs to stop spending, pause it. Archiving is irreversible. Read-only vs read-write MCP for Amazon advertisers explains which access to start with.

How do you keep a joined answer honest?

You check the inputs before you trust the output. A joined answer is only as good as the rows that went in, and the assistant cannot see mistakes in a file it was handed.

Treat a pattern across channels as a lead, not proof. If the assistant sees store sales rise in weeks when ad spend rose, ask it for other explanations, then test with a planned change before you move a large budget.