In the past Amazon AI was based on “collaborative filtering” where shoppers’ predicted purchases were based on similar customers’ buying habits.

Recently, though, Amazon AI has switched to “item-to-item filtering.” Analyzing purchases at the item level now gives Amazon better predictions than at the customer level.

Amazon AI has not thrown out customer similarities. They are still used to inform product relatedness. For now, though, item-to-item filtering leads the Amazon AI recommendation engine.

Amazon AI also now prioritizes recent item purchases over less recents ones.

Amazon AI’s item-to-item evaluation criteria can be broadly grouped into: (1) visual, (2) title and (3) product behavior. Amazon AI knows that shopper emphasis of these three criteria varies across categories. In fashion, Amazon AI knows the visual matters the most to shoppers; for office products, the title is most important; for jewelry, customer evaluation is largely based on product behavior. Amazon AI ultimately ranks recommended products with a compound evaluation of these three groups within the item category.

900%™ earns you money by leveraging Amazon AI in placement bidding. You can make money with any campaign at Top of search or Product Pages placement if you can keep base bidding appropriately low and maximally accelerate your bid when the Amazon AI-determined likelihood of a sale is high. Additionally, for most campaigns, with 900%™ you’ll find Rest of Search paying for the whole campaign.

4 Basic Rules of Campaign Structure to Leverage Amazon AI

Rule #1—Structure for Maximum Separation. You’ll want to structure campaigns to achieve maximum separation of placement bidding strategies. Maximum separation means that a campaign will clearly exhibit a preference for Top of search or Product pages placement. Proper structure leverages the deep learning of Amazon AI and increases your proportion of ad-based revenue while proportionately decreasing ACoS.

Rule #2—Train 900%™ With A Short History. Since Amazon AI now puts increased emphasis on recent item purchases, keep the 900%™ training history short. We recommend a six-week look-back period, but you may need to vary it according to your product. In general look for the shortest possible training history to best leverage Amazon AI.

Rule #3—Isolate Branded Keywords. Branded keywords are almost always to be placed in Top of search. Keep them in their own campaign. Adding non-branded keywords dilutes campaign impact and creates sub-optimal placement and bidding.

Rule #4—Leverage High View-to-Purchase Ratio. If a single product variation garners the majority of views, advertise only that variation. When the shopper clicks on your ad, they will be directed to the product detail page in which all the variations will be displayed. The customer can then choose from the available variations as per their interest. Amazon AI will increase your ranking based on the new, undiluted View-to-Purchase ratio. Over time 900%™ will be able to lower your base bid and increase your placement accelerators.

900%™ is the seller’s front end to Amazon’s AI back end.  There is an optimal money-making bid for every campaign placement—900%™ helps you find it.