Personalization in the World of Protected Audience API

We've all experienced this type of messaging failure. A brand attempts to use personalization tags, but you end up getting an email that crosses over into being creepy.

True consumer engagement doesn't just come down to using first names in a greeting or having a flashy “call to action” button. It's crucial that each consumer is served ads that are both relevant and meaningful to them.

Unveiling the power of personalized ads: A customer-centric revolution

Personalization can bring a real competitive advantage for companies. It can result in up to a 50 percent reduction in customer acquisition costs. Moreover, it has the potential to generate anything from 5 to 15 percent in additional revenue and can boost marketing return on investment (ROI) by 10 to 30 percent.

Personalization is proven to enhance performance and deliver superior customer outcomes. Research conducted by McKinsey shows that personalized experiences not only foster customer loyalty but also contribute to higher gross sales for a company. It’s also exactly what consumers want: 71% of people now expect personalization and 76% are frustrated when they don’t get it.

Personalized customer journeys in today’s environment

“Half the money I spend on advertising is wasted; the trouble is I don't know which half”
John Wanamaker

Let’s take a look at how personalization is handled in today’s customer journey in the digital environment.

1. The user enters the publisher's website.
2. They’re not logged in. The user checks some automotive news. While scrolling the cars sub-page, they pay a great deal of attention to the SUV category.
   
3. They view a recent video review of the Audi Q5.
4. After finishing their car browsing, the user opens a new tab and reads an article entitled ‘Ten pieces of advice for new-born nutrition’.

Based on these few simple touchpoints, the ad tech algorithm can potentially collect a large amount of data, including:

• URL of visited websites
• Geolocation
• Context of both sites visited
• User device
• User browser
• User operating system
• Behavioral habits of other users with similar tastes
• Previous interactions with the ads
• Topics API signals
• And much more….

Personalized advertising significantly impacts the overall user experience.

Remaining with the automotive example, on a general news website, there may be numerous users interested in automotive news. It is crucial to avoid bombarding those users with irrelevant targeted ads from other categories. As businesses collect more and more data about customers, personalization is becoming a must-have element of customer-driven marketing. Previously, data primarily served internal purposes; nowadays, each strategic decision needs to be backed up by the right numbers.

AI can be helpful for connecting the dots and deriving conclusions from overwhelming amounts of collected data that would be impossible for the human brain to examine. Machine Learning and Deep Learning algorithms are currently leading this personalization race and are able to find hidden layers in user behaviors.

For example, based on the data points mentioned above, algorithms can determine:

• The expected CTR/VCR
• What type of banner will be the most effective (video/display)
• What creative to serve
• Which product should be displayed on the banner
• What placement on the publisher's website offers the best value-to-price ratio

What changes with the deprecation of cookies and the introduction of Protected Audience API?

In the cookieless world, the algorithms should still be able to provide answers to the above questions. Users will no longer be identified as individuals across different domains, but it seems that this will not be necessary in order to deliver highly personalized advertising.

Protected Audience API

The Privacy Sandbox initiative aims to address this challenge, especially with the Protected Audience API. This concept allows the analysis of granular user behavior from a single website and the utilization of this information to group similar users into cohorts. A DSP can then target specific cohorts without the need to reveal a group.

If a number of people on the general news site are interested in electric cars—searching for rankings or reviews of specific models—Protected Audience API allows ad tech companies partnering with the news website to gather together these users and create a group such as “electric vehicle enthusiasts”. These users can later be targeted across the internet while remaining anonymous.

Protected Audience API includes two proposals from RTB House that are crucial for personalization: Product-level TURTLEDOVE and Outcome-based TURTLEDOVE. The first focuses on products displayed on ads, and the second addresses user-level bid evaluation without the necessity of cross-site user identification. For a personalization use case, let’s focus on the first one.

Product-level TURTLEDOVE allows ad tech vendors to store information about products recommended for the user on their device at the moment of adding the user to the interest group. It enables the possibility of displaying relevant product recommendations to a specific user, without the necessity to identify them across the internet.

For example, with our electric vehicle enthusiasts, it could be useful if a user read a review of a specific model and the ad tech vendor was running a campaign for the manufacturer. It would allow the ad tech to save this product as a recommendation for this user, along with accessories and other complimentary products from the same brand. Alternatively, if the ad tech vendor was running a campaign for a multi-brand car dealership e-shop, it could stow this specific vehicle as well as alternatives from other manufacturers.

How does Protected Audience API affect personalization

Protected Audience API maintains the personalization and user experience of current advertising methods while ensuring the security of user data. The ads themselves are expected to look very similar to those that users are accustomed to now.

Preliminary tests conducted by RTB House indicate that Product-level TURTLEDOVE maintains approximately 94.5% recommendation effectiveness compared to current methods.

Importantly, data will not be mixed across different websites. Particular users might be assigned to many groups in the Protected Audience API. These are set up by the ad tech company in collaboration with the sites they visited, for example: "examplesite1_automotive" and "examplesite2_sports." And these groups will not know about each other, so they will be evaluated separately from different perspectives and the most valuable bid will win out.

Ownership of the groups is also clear. For example, it is known that the group established on site XYZ won the ad, and this could be an answer to a rather gray area with third-party data usage.

If an ad tech company that is cooperating with Audi displays ads for BMW, it means that the ad tech partner is using Audi’s data to display banners for competitors. Does Audi know about this and agree to it? Right now it is hard to check, but will be less ambiguous in the future.

Another positive aspect is preventing data leakage. In today's ecosystem, there is both massive data leakage violating user privacy and data loss in the cookie matching process. Recent AdExchanger research suggests that more than 50% of users are "lost" in this process. The Protected Audience API makes it possible to reach 100% of users added to these groups because data is stored on the user's device and does not require any matching.

Ad personalization has always been crucial to achieving great marketing results—and it will stay that way. However, the methods behind displaying ads will definitely change and it is important for marketers to understand the implications of the coming changes and work with ad tech ecosystem players on implementing them most effectively.

‍

Sources:

1. RTB House blog, How Will Personalized Ads Work without 3rd Party Cookies?, 30 June 2021
2. RTB House blog, Will Your Bidding Efficiency Drop without 3rd Party Cookies?, 22 June 2021
3. McKinsey research, What is personalization?, 30 May 2023
4. Adexchanger content studio, Study: Cookies’ Low Match Rates Cost Ad Tech Millions. Moving Off Cookies May Be The Answer, 31 March 2023