One of the things that the current focus on the imminent death of third-party cookies has brought to life is the set of misperceptions upon which the reliance on third-party cookies was formed. The data, while easy to come by and filled with intriguing assumptions, was often not what it seemed. Based, for example, on whose device a search occurred, the cookie collector could assume that the target was the device’s 38-year-old male owner, when the user conducting the search could just as easily be his 35-year-old wife or his 10-year-old daughter.

More critical, perhaps, was that third-party data lacked the wherewithal to foster immediately actionable insights. “The premise of third-party data containing accurate signals has always been kind of incorrect,” says Marco Godina, SVP of Product at Silverbullet, a data and digital transformation company. “Faced with technical limitations, the industry has never been able to pull it off successfully.”

One of the key directions in which brands have headed in the search for post-third-party cookie replacements is toward contextual targeting, where, as Godina puts it, “we’re moving away from making spurious assumptions based on past browsing behavior to instead take a deterministic approach and focus on content that is consumed at the time of the actual impression.”

As this has become more sophisticated, it’s allowed marketers to not only look at the obvious contexts, but to broaden their vision to include related topics. So, for example, an automotive marketer promoting SUVs might start by targeting content covering vehicles, but then realize that people buying SUVs are also interested in reading about families and lifecycle events—and the focus on that content may also influence the outcome.

Back to Basics

In looking for outcomes, notes Godina, “different brands will care about different things. Certain brands will worry about brand safety but not as much about how efficient or effective the media is. Other customers will care much more about lead submissions or soft conversion events. And then here are those that want someone to actually purchase something.”

“We’re moving away from making spurious assumptions based on past browsing behavior to focus on content consumed at the time of the actual impression.”

Godina says that “the promise of digital advertising, compared to other forms of media, is that you can actually track and know if an impression was viewed and if someone really bought something. You can loosely connect the dots.”

Dot connection was clearly a goal with third-party cookies, Godina says. “The industry has built out a massive infrastructure to do this, using third-party cookies and offline onboarding solutions where more or less every activity you do online or offline could be attributed to something”—regardless of whether the data you’re basing this attribution on is correct.

The shift away from third-party cookies is ushering in, as Godina puts it, a move “back to basics, with a focus on targeting people in the right environment, at the right time, and adding to that giving customers intelligence and using things like machine learning to influence that performance.” With contextual targeting, he notes, it’s possible to accurately match an ad to the desired content. But you can take this further. “Technology has advanced enough,” he says, “to where you can start to look at what other types of content are on the page that will help inform the outcome.” So if, going back to the earlier example, the content is about SUVs and it overlaps with keywords related to family events, “you might get a much higher click through rate than you normally would or a much higher conversion rate.” And if, he adds, “you can also ingest information from the programmatic supply chain, you can focus on things like impressions, clicks, viewable impressions, and other metrics to deterministically influence a campaign in mid-flight.” All of which adds much more value than could have been available with third-party cookies alone.

Applying the Intelligence

This approach can help the brand in two different ways. First, says Godina, “say you have a brand that’s already running contextual targeting but needs to make it more efficient and effective. If we place a tag on their media, we can monitor how impressions are served when a click happens. Then we can go back, retroactively, and see what intelligence we can apply. When we see what keywords, themes, or dimensions were present on the pages, we can add more or remove some and alter the composition of what’s being targeted, affecting—positively—the outcome of the ad.”

“An auto marketer promoting SUVs might realize that people buying SUVs are not only interested in reading about SUVs but also about road trips and lifecycle events; that focus may influence the outcome.”

Alternatively, Godina notes, a brand may be doing non-contextual targeting. In this case, the campaign might “be based on things such as people who have visited a specific site or CRM data of high-value customers. Then the focus is more on ‘Help me find people who are my best retargeted customers or my highest lifetime value customers—and find the types of contextual environments they’re engaging with.’”

In either case, with this information in hand, Godina says, it’s possible to generate a report that “ingests the brand’s media data, segments it by all the different metadata that can be seen from the DSP, and runs it through a contextual engine to gain a nuanced understanding of what the content is truly about, as well as how customer engagement changes based on the makeup of that content.” Continuing the example, Godina explains, “Say you had 2 million impressions. Those impressions are served against a variety of content. Some percentage is served against standard contextual theme like automotive SUV or lifecycle or family events. And another percentage is served against more nuanced themes such as positive page sentiment, pop-up free environments, and stable domains. These themes intermingle across impressions, but you can see the performance of each theme and generate recommendations generated by machine learning and AI to help you improve your current campaigns, make decisions about such other strategies as retargeting, or even plan future campaigns.”

What’s critical here, Godina stresses, is that the insights gathered through this process are accurate, actionable, and don’t kill the ability to scale. There’s limited value, he says, to a data vault with “cool numbers and cool graphics” if you can’t do anything with those numbers or graphics or if you do something and only reach 10 users. Signals in the report should “be immediately usable and scalable; otherwise they don’t provide benefit.” With this information, and with the ability to act immediately and impact outcomes, says Godina, “I think everyone will get into a much more accurate place.”