Native Ad Auction
Learn how Promoted optimizes native ad auction within search and feed results to provide the best experience for your users.
.png)
Promoted’s ads work by first maximizing user experience in search and feed...
... and then running an ad auction to promote items at the top to increase revenue.
Promoted optimizes ads for incremental sales and bills on impressions
This is called “Optimized CPM” or “oCPM.”
CPC, CPM, and mixtures of these are also supported.
Promoted runs a VCG ad auction to compute ad prices.
To optimize across repeated and variable opportunities
Within a budget, Promoted auto-bids using real-time data.
The ad auction works by
Computing optimized bids using machine learning.
Value of No Ad
First, Promoted computes the value to the advertiser if the item already appears in search or feed. If it doesn’t appear, this value is 0.
Next, Promoted computes the value to the advertiser to every higher position as the Advertiser Bid.
The difference between the [Advertiser Bid] and the [Value of No Ad] is the advertiser bid for each higher position.
The “Event Bid”
Is the value of a sale or click and is set by the advertiser.
The “Event Prediction”
Uses machine learning to predict the probability of conversion or click. It leverages maximization systems used for non-ads.
The “Pacer”
Is an auto-bidder control system across auctions over time.
The Complementary Bid controls the user experience
The Quality Score reuses search and feed optimization systems and adds controls for diversity and exploration.
The Advertiser Bid and Complementary Bid add together
A negative complementary bid makes the ad less competitive and more expensive.
If it is too negative, the ad is trimmed.
Ads targeting leverages existing marketplace relevance rules.
You can also create custom targeting
With User or Request properties that you send to Promoted, like repeat buyer retargeting and device type.
