The New York Times reports that it was able to deliver highly efficient targeted advertising without third-party cookies

With the abolition of third-party cookies becoming more and more realistic, new advertising mechanisms that replace existing targeted advertising are being tested everywhere. The New York Times, which is building an advertising system based on its own platform without relying on third-party cookies, has revealed the details, saying that it was able to perform as well as third-party cookies.

To Serve Better Ads, We Built Our Own Data Program | by The NYT Open Team | Dec, 2020 | NYT Open
https://open.nytimes.com/to-serve-better-ads-we-built-our-own-data-program-c5e039bf247b

Most of the advertisements on the Internet are behavioral targeting advertisements using third-party cookies, but in recent years, the use of targeting advertisements has been reviewed. This has the advantage that targeted advertising using third-party cookies that widely tracks user behavior 'enables efficient advertisement display that suits the user's interests', while providing user's personal information. This is because it is collected excessively and problems have been pointed out from the viewpoint of privacy.

Accelerating the review of third-party cookies is the new EU data protection regulation 'GDPR' enacted in 2016. The GDPR stipulates that 'the collection and retention of personal data by the website must be based on the consent of the user', and the use of third-party cookies without the permission of the user is now strictly checked.

In 2017, Apple released a privacy protection feature called Intelligent Tracking Prevention (ITP) that blocked third-party cookies in Safari. In addition, Google has announced that it will abolish third-party cookies in Chrome in 2019, and is in the direction of reviewing the existing advertising system.

It has been pointed out that the performance of targeted advertising is so high that if targeted advertising becomes unavailable, website and store revenue may decrease. On the other hand, it has also been suggested that profits may increase, and many websites are looking to build new advertising systems.

There is a possibility that the profit on the website side will increase by stopping targeted advertising --GIGAZINE

The New York Times, a major American newspaper, also announced in May 2020 that it plans to stop using data from third-party cookies and display advertisements using its own data platform.

The New York Times plans to stop using third-party data for targeted advertising-GIGAZINE

On December 18, 2020, the New York Times made further mention of how the data platform under construction works.

Over the past few years, the New York Times has been using Google Cloud Platform to stream real-time events to the data analytics tool Google BigQuery with Google Cloud Dataflow . Event data sent to BigQuery was then batched and used to train multiple machine learning models built by New York Times data scientists for ad operations.

This machine learning model is basically based on a survey answered by tens of thousands of volunteer New York Times readers. It is a mechanism that trains the model with this questionnaire data and event data and predicts the user who should display each advertisement. Supervised learning was required in regression, ranking, classification, multitasking, etc. before statistically predictable and interpretable results could be obtained. Airflow was used to output the machine learning model.

According to the New York Times, the actual mechanism is that when a user visits the New York Times website or opens an app, a call is made to a Go language microservice using Google Cloud Memorystore, with an average delay. User data is sent to the page in less than 100 millimeters per second. JavaScript embedded in the page sends this data to the ad server, which finds the campaign that is most relevant to your user data and sends it when the campaign's targeting criteria match the data sent from the page. It will be delivered.

Since processing and data activation are done on the server side, it seems that this mechanism could be easily extended to multiple browsers.

Since launching the program in June 2020, ads based on the New York Times' own data have performed as well as ads using third-party cookies. Internet advertising systems are divided into programmatic advertising that relies on Demand Side Platform and Supply-Side Platform , and direct-selling advertising that buys and sells direct inventory. The new mechanism of the New York Times is the latter. The New York Times says the New York Times can better protect readers' privacy by eliminating the need to share their reader data with the outside world.