# Did my Amazon deal make money?

A playbook for an AI assistant connected to Nova. It reads each product's discount, deal fees, ads and profit, compares them with the weeks before and after, and tells you which deals to repeat. 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. Put in your deal dates, or leave them for last year's Black Friday week.

```text
Use Nova to find out whether my Amazon deals made money, and which ones to repeat.

Deal period: [dates]. If I name none, use last year's Black Friday week, from the Monday before Black Friday to Cyber Monday.
Baseline: the same number of days right before the deal period.
After: the same number of days right after the deal period.
Fee window: from the first day of the deal to 21 days after its last day.

1. Check the costs first. Call Nova's query_metrics with netProfit for the deal period. If the answer carries cogsNotice, say what share of units has no cost before you quote any profit. If adsHistoryNotice says ad data starts after the deal, say that ad spend is unknown there.

2. Find the products. Call Nova's query_metrics with groupBy "childAsin", the deal period against the baseline, and the metrics units, netSales, netSalesPerUnit, productPromotions, adSpend, organicSales, netProfit and netProfitMargin, sorted by productPromotions, limit 30. A product counts as on deal if productPromotions rose or netSalesPerUnit fell. Some deals lower the price, so they show in netSalesPerUnit and not in productPromotions.

3. Read the fees Amazon books later. Call Nova's query_metrics with groupBy "childAsin" over the fee window, for those products, with the metrics dealParticipationFee, dealPerformanceFeeRollup, promotionRunLightningDealFee, couponParticipationFeeBase, couponPerformanceFee and couponRedemptionFee. Amazon can book these well after the deal, so the deal period alone can miss them. If the fee window has not ended yet, say the fees may be incomplete. Keep the sign each figure comes back with.

4. Work out the extra profit, per product. Take netProfit in the deal period, minus netProfit in the baseline. Then take off the fees from step 3 that fall outside the deal period, because netProfit in the deal period does not hold them. Show the result before and after those fees.

5. Read the discount. Discount share = productPromotions / (netSales + productPromotions). Also show the change in netSalesPerUnit against the baseline.

6. Check what ads did. From step 2, compare adSpend and organicSales with the baseline. Say whether the lift came with more ad spend or came organically.

7. Check what happened after. Call Nova's query_metrics with groupBy "childAsin", the after period against the baseline, with units and netSales. Then call Nova's query_bsr with groupBy "childAsin" for those products, the after period against the baseline. Say whether sales fell below the baseline after the deal, which suggests buyers came early, or held, and whether rank moved.

8. Check what else changed. Call Nova's list_events for the baseline, deal and after periods. Name any price, content or PPC change that could explain the result.

9. Check you can run it again. Call Nova's query_inventory with groupBy "asin" for the products you would repeat. Give days of inventory and inbound units against the units sold in the deal period.

10. Report, one row per product: discount, units against the baseline, extra profit before and after fees, ad spend change, what happened after, and a verdict: repeat, change the discount, or drop. Then give the total before and after fees, and the one change you would make for the next event.

Rules: these are correlations, so never state a cause as proven. Events like Black Friday lift demand anyway, so say that the baseline can flatter a deal. Say which figures are missing and why. Relay every notice Nova returns (P&L rules, marketplaces added up, COGS gaps, currency, ad history). You cannot create deals or change prices.
```

## How to read the result

| What you see | What it usually means | What to do next time |
| --- | --- | --- |
| Extra profit stays positive after fees, sales hold after the deal | The deal paid for itself | Repeat it |
| Extra profit positive before fees, negative after | The fees took the gain | Lower the discount or pick a cheaper deal type |
| Units up, profit down, sales below the baseline after | Buyers came early | Shorten it or drop it |
| Lift came with higher `adSpend`, flat `organicSales` | You bought the lift | Judge it with the ads included |
| Rank better after the deal and sales above the baseline | The deal may have helped beyond its dates | Check again a month later before you repeat it |
| Stock ran out during the deal | The lift was capped | Fix stock before you repeat it |

## Limits

- Nova shows what moved at the same time as the deal. It does not prove what the deal caused.
- A few fees come back with no product on them. Read those at account level.
- Stock is the state today, not a history, so it cannot show a stockout during last year's deal. Units falling to zero point to one.
- A deal that overlaps another promotion cannot be separated from it.
- Ad data starts about 60 days before you connected Amazon Ads, and ad spend reads 0 before that.
- A trial reads 3 months of orders, fees and returns. Last year's Black Friday needs a paid plan's 2-year history.
- This playbook is built for Seller Central. It does not create deals or change prices.