Meta implemented the new system in the United States aimed at the company’s own ads, with the aim of making their distribution more equitable, following the agreement reached with the country’s Department of Justice.
The variance reduction system is a machine learning tool that Meta developed in conjunction with the Department of Justice so that “the actual audience that sees an ad more closely reflects the eligible target audience for that ad”, as stated in a post on its official blog.
To reduce differences between audiences, during the course of a campaign the algorithm will analyze the aggregated demographic distribution of the people who have seen the ad in the Meta services to later compare it with the target audience chosen by the advertiser.
The company stressed the privacy of this tool, which is based on improved Bayesian surname geocoding, information from the country’s census, to guarantee the privacy of people, to which “additional privacy improvements” have been added that prevent, for example, the reidentification of an individual within an aggregated data set.
(Reference image source: file)