On the heels of its earnings report Thursday, Amazon Ads has developed more advanced machine-learning models and optimized campaign-control systems to enhance bidding and pacing decisions for the Amazon DSP.
Neal Richter, director of Amazon DSP Technology, believes that “every percentage point of improvement counts to advertisers, and these new upgrades have helped increase engagement and return on ad spend.”
These models better predict the likelihood of a bid request converting, enabling algorithmic changes that improve pacing-to-goal while optimizing for performance.
The models are trained to better predict the likelihood of a bid request converting and enable algorithmic changes that improve pacing-to-goal while optimizing for performance.
Early tests show that the enhanced models generate 12.6% increase in click-through rates, 34.1% increase in return on ad spend, 24.7% decrease in cost per click, and between 20% to 30% incremental addressability on inventory that was previously unaddressable.
Modeling available signals to reach desired audiences is critical, and is especially important as the advertising industry moves away from third-party cookies.
The new machine-learning models analyze a range of signals to help advertisers predict and reach highly relevant audience segments with optimal cost-efficiency.
Amazon on Thursday reported sales of $127.4 billion for Q1 2023 -- up 9% compared with the year-ago quarter.
Revenue at the company's advertising services division contributed $9.51 billion -- up from $7.87 billion from a year ago.
By the end of 2023, Amazon’s worldwide ecommerce sales are forecast to grow nearly 9% to $685.39 billion, giving it 11.6% share of the global ecommerce market.
Amazon's U.S. advertising business is estimated to grow more than 17% this year to reach $33.96 billion, giving the company a 12.9% share of the U.S. digital ad market.
This year, 51% of Amazon's worldwide ad business will come from mobile, the first time it has crossed the halfway mark, according to Insider Intelligence.