We suggest to duplicate your top 1 or 2 ad sets that have gotten you Amazon purchases in the past 30 days. If you don't have any previous Amazon purchases or any previous ads, you can use any FB ad strategy of your choice that you learned from other content or our other pages.
Don’t touch the original ad set, this way you have a comparison point after duplication. For example, decide based on the Amazon purchase column (if you had ZT-Track feature) or your best guess which ad set produced Amazon sales, and change out this optimization event for the Purchase from the new pixel. That's it for now and can follow along results before doing it for more campaigns.
Optimization window select "7 day click or 1 day view" (or "7 day click" if you know what you're doing) and also use Automatic bidding. These options are good while you are testing. Normal FB pixel optimizes right away but this Amazon feature optimization is happening when Amazon sends order data to Facebook which typically happens 0-3 days because order must be "shipped" for Amazon to send data to Facebook. Sometimes it happens same day but it’s best for the delayed attribution to come in and for Facebook to optimize after ~3 days passes to make decisions.
The more data FB receives, the better it will optimize for more Amazon purchases. The above is a conservative approach to test one campaign at a time. Alternative is more aggressive. Duplicate any or all of your campaigns that are producing you results or likely Amazon sales, then add or change out the conversion event to be Purchase from this new Amazon pixel. Make decisions as you normally would turning off lower performing ads/ad sets/campaigns comparing to your previous marketing and within the new setup to see which ones to funnel your budget into. Go to the Members portal (free content too) to watch other FB strategies in detail on how to test/optimize/scale FB ads for Amazon.