# What to reorder on Amazon before you run out

A playbook for an AI assistant connected to Nova. It ranks what runs out first, checks what is inbound, reads the margin behind each product, and tells you how many units to order. 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 weekly, or on a schedule if your assistant supports scheduled tasks.

```text
Use Nova to tell me what to reorder on Amazon, how much, and what not to reorder.

1. Read what runs out first. Call Nova's query_inventory with groupBy "asin", includeReorder true, oosRiskFlag true, sorted by daysOfInventory, sortOrder "asc", limit 30. Give days of inventory, days of supply after inbound, sales velocity, and units on hand. Never add the non-Pool quantities across SKUs.

2. Check the parameters. Each row gives lead times and targets, with a source for each. If a source says fleet_default, or paramsTyped is false, say the recommendation uses Nova's defaults and not my own lead times. If recoStatus is params_missing, say nothing is assumed and ask me for lead times.

3. Check what is on its way. Call Nova's query_inbound_shipments with category "in_progress", then "planned". Planned shipments are still with me, so they are not inbound yet. Then call it with category "closed_incomplete" to see shipments that came up short.

4. Read the worth of each product. Call Nova's query_metrics with groupBy "childAsin", the last 90 days, with units, netSales, netProfit and netProfitMargin. If cogsNotice appears, say what share of units has no cost before you quote profit.

5. Check that velocity is real. Call Nova's query_metrics with groupBy "week" over the last 8 complete weeks for the top 5 products, with units. A week at or near zero points to a stockout, so velocity may understate demand. If the next weeks hold a peak, compare the same weeks last year.

6. Check the listing. Call Nova's query_listing_health for those products. Do not recommend stock for a listing that cannot sell.

7. Read the order. From step 1, give recommendedOrderQty, daysUntilReorder (negative means late), coverageTotal, recommendedOrderWeightKg and recommendedOrderVolumeCbm. Subscribe & Save deliveries are in snsForecastUnits30d, so read them next to days of inventory.

8. Report, one row per product: days left, inbound, margin, units to order, and a verdict: order now, order this month, hold, or do not reorder yet. Say why for any product that loses money. Then the total units and freight weight.

Rules: stock is the state today. Do not recommend reordering a product only because it is low: read its margin first. Say which figures are missing and why, and relay every notice Nova returns. You cannot place orders or create shipments.
```

## How to read the result

| What you see | What it usually means | What to do |
| --- | --- | --- |
| Few days left, nothing inbound, good margin | Lost sales ahead | Order now, and consider air for part of it |
| `daysUntilReorder` negative | The order is already late | Order now, and plan the gap |
| Few days left, negative margin | Stock would keep a loss going | Fix price or cost first |
| Many days of stock, low velocity | Cash parked in the warehouse | Hold, and check storage fees |
| Weeks at zero units in the series | A past stockout | Read velocity from the weeks before it |
| `recoStatus` is `params_missing` | Nova has no lead times for it | Enter them, because nothing is assumed |

## Limits

- Stock is the state today, not a history. Weeks of units are the way to date an old stockout.
- Velocity reads the last 30 days, so a recent stockout or peak skews it.
- Inbound shipments carry no dates, because Amazon's inbound API gives none.
- Stock outside Amazon and supplier orders count only if you entered them in Nova.
- The recommendation uses your lead times and targets, or Nova's defaults where you set none, and each row says which.
- Nova does not place orders or create shipments.
- This playbook is built for Seller Central.