Google’s Privacy Sandbox Update

The arrival of Google’s Privacy Sandbox brings immense confusion to the Adtech industry. As it soon replacing third-party cookies, there’s a lot of confusion among the industry on its functioning & impact on programmatic advertising.

What is Google Privacy Sandbox?

Google Privacy Sandbox is a new approach that makes a way for ad tracking, measurement, and fraud prevention without the help of third-party cookies. It mainly works with the different browser’s APIs, where advertisers can collect the aggregated data on their issues such as conversion, attribution, etc. In simple terms, Privacy Sandbox is an alternative for third-party cookies to make a profit from the user’s browsing(on chrome) habits.

Here’s an article to understand Google’s Privacy Sandbox deeply. Now, coming to the latest update,

Google Tests its Privacy Sandbox Proposal

FLoCs proposes that users can be grouped based on their interests, thus forming cohors. Cohorts can be then targeted to show interest-based ads. Well, last week Google published results from its first experiment that tested FLoCs.

The test is primarily to prove that generating cohorts based on interests (FLoCs) can work for advertising and it is better than randomly grouping users.

Datasets:

To evaluate FLoCs (clusters), Google used two publicly available datasets — Million Song Download and MovieLens. It then used its proprietary dataset (de-identified collection of URLs from Google Display Network across 7 days) to evaluate the viability of FLoCs for ads personalization.

So what was the result?

Google concluded that there’s a 350% improvement in recall and 70% improvement in precision (precision refers to the precision of the algorithm. That is, predicting whether the users are converting on the interest groups they’ve added via FLoC within the 7-day window).

“We were very pleasantly surprised. This particular kind of algorithm could be a very useful substitute for third-party cookies while still providing for increased privacy.”

Takeaway:

Two things to consider:

1. Datasets used aren’t relatable to common publishers. Both MSD and MovieLens have a dozen first-party data fields while most publishers wouldn’t have even half of it.

2. The result is comparing FLoCs with random clustering — not another clustering algorithm.

Google agrees that it is just the first step and invites the industry to run their own experiments using a proprietary dataset to evaluate the viability of FLoCs for clustering and ads personalization. We’ll keep you posted.

Though the ad tech industry is full of jargon, it’s quite easy to cope with them if you are aware of it prior. Each day the digital advertising is filling up with the latest updates, happenings, trends, etc., As a publisher, you should get ahead with this to achieve your revenue goals. So, let us help you by sending the right information at the right time with just signup.

Originally published at https://headerbidding.co on October 27, 2020.