# Count the Amazon halo before cutting Meta

> Meta ads can look unprofitable on site sales while also driving Amazon sales. Connect Meta Ads, Shopify and Amazon Seller Central, ask Claude to find weeks when Meta spend was cut or paused, what happened to Amazon and Shopify sales after, and your Amazon share of sales. If Amazon dropped when Meta paused, count that before cutting.

Type: Play (a chance to grow)
Run it: Before any big cut
Source: https://gooseworks.ai/playbooks/check-amazon-sales-before-cutting-meta-ads

## 1. Connect

Give your AI assistant read-only access to Meta Ads, Shopify and Amazon Seller Central.

## 2. Ask

```text
Over the last 12 months, find any week where Meta spend was cut sharply or paused. What happened to Amazon sales and Shopify sales in the weeks after? What share of my total sales is on Amazon?
```

## 3. Act

If Amazon sales dropped when Meta paused, count that before cutting.

## Steps

1. Connect Meta Ads, Shopify and Amazon Seller Central read-only through a Claude connector or MCP server, or export weekly spend and sales reports and upload them.
2. Ask the playbook prompt to find past Meta cuts or pauses and compare Amazon and Shopify sales in the weeks after.
3. Check what else changed then, such as prices, Amazon ads or stock. Treat the result as a warning, not proof.
4. If you still cut, do it in steps with a review date and watch Amazon sales weekly.
5. Re-run this before any big cut, and compare with the same weeks last year.

## What it tends to find

Meta looked weak on site sales, but 35% of sales were on Amazon. (Illustrative example, not a result from a named store.)

## Watch out

Other things change at the same time (prices, Amazon ads). Treat this as a warning, not proof.

## FAQ

### What is the Amazon halo effect?

It is when ads seen on Meta lead people to buy on Amazon instead of your site. Meta cannot track those sales, so its reports undercount what the ads do.

### How can I tell if Meta drives my Amazon sales?

Look at past weeks when Meta spend dropped sharply. If Amazon sales fell soon after and nothing else changed, that is a sign. A planned holdout by region gives a clearer read.

### Is this proof that Meta works?

No. Other things often change at the same time, like prices, Amazon ads or stock levels. Use it as a warning to test carefully, not as proof to keep spending.

### Will another assistant handle this prompt?

Yes. ChatGPT, Grok, Muse or any assistant that can read your ad spend and sales from both channels can run it, connected or from weekly exports.