# How can I create an Amazon performance deck for a client call quickly?

A playbook for an AI assistant connected to Nova. It builds a client's weekly performance deck from live Amazon data: the week, the month and the year against their baselines, the products that moved, advertising, stock and listing risks, and search share, with three actions to close. It only reads your data.

Every step calls a Nova tool that reads your live Amazon data, so paste this into Claude, ChatGPT or any assistant connected to Nova's MCP server. Run it the day before each client call, once per client.

```text
Use Nova to build this week's performance deck for one of my Amazon clients, ready for the client call.

Scope: the client I name, every marketplace of their accounts unless I name some, and the last complete week (Monday to Sunday).

1. Find the client. Call Nova's list_accounts and keep the accounts of the client I name. Say which accounts and marketplaces the deck covers, and whether they are Seller Central or Vendor Central.

2. Week slide. Call Nova's query_metrics with groupBy "marketplace", and sales, units, orders, sessions, conversionRate, adSpend, sponsoredSales, sponsoredTacos and netProfit, for the last complete week, with the week before as comparisonFrom and comparisonTo.

3. Month and year slide. Call query_metrics with the same metrics and no groupBy three times: month to date against the same days last month, month to date against the same days last year, and year to date against the same days last year.

4. Products slide. Call query_metrics with groupBy "childAsin", sortBy "sales", limit 10, and sales, units, adSpend and netProfit for the week against the week before. Then call Nova's find_movers with metric "sales", dimension "childAsin", direction "down", limit 5, over the same two weeks.

5. Advertising slide. Call Nova's query_advertising with groupBy "adProduct", adProducts sponsored_products, sponsored_brands, sponsored_brands_video and sponsored_display, and spend, sales, acos, orders, clicks and cpc, for the week against the week before.

6. Risks slide. Call Nova's query_inventory with groupBy "asin" and oosRiskFlag true, then Nova's query_listing_health sorted by severity, limit 20. Keep the products that sold in the last 30 days.

7. Search slide. Call Nova's query_sqp with grain "week", groupBy "period", queryTypes ["core"], and search_volume, your_impression_share and your_purchase_share, over the last 8 complete weeks. Say whether the client's share of their core searches rose or fell.

8. Build the deck. One slide per step from 2 to 7. Give each slide a title that states the finding, at most three bullets, and every figure with its comparison. Add a last slide with three actions for the coming week, each tied to a figure. If you can create files, return a slide file; otherwise return the slides as an outline I can paste into our template.

Rules: write for the client, with no internal notes. Relay every notice Nova returns before the figures it qualifies. Never add Seller Central and Vendor Central figures together. Say which figures are partial, such as traffic for the last two days. Never claim an action was taken.
```

## How to read the result

| What you see | What it usually means | What to do |
| --- | --- | --- |
| Sales up, net profit flat | Growth came from ads or discounts | Show the TACOS and promotion lines next to it |
| Week down, month to date up | One weak week inside a good month | Lead with the month, explain the week |
| A top product on the risks slide | It may stock out or stop selling soon | Put it first on the actions slide |
| Search share falling on core searches | Competitors are taking the client's searches | Bring the ad and listing plan for those searches |
| Traffic marked as partial | Amazon has not published the last days yet | Compare conversion on complete days only |
| A slide with no data | The client has no ads, no Brand Analytics or nothing at risk | Drop the slide for that client |

## Limits

- Amazon publishes traffic about two days late, so the last days of a week can read low.
- Search Query Performance needs Brand Analytics, so it needs a registered brand.
- Net profit is only as complete as the product costs entered in Nova for that client.
- Nova does not send decks or reports by email. Run the playbook as a scheduled task in your assistant to have the deck ready before each call.
- Written for Seller Central clients.
- The assistant builds the deck from Nova's data. It does not change campaigns, prices or listings.