In the programmatic era, one might say to online advertisers: This is not your father’s contextual targeting.
Historically, contextual targeting has been relegated to a minor role in the digital advertising playbook. Earlier versions of the strategy had obvious limitations. The contextual hierarchies provided by IAB, for example, often proved to be too broad for niche targeting. To get that kind of specificity, an advertiser would have to rely on labor-intensive methods such as keyword search, which suffered from the opposite problem of a lack of scale. Behavioral targeting, meanwhile, has become the dominant approach in a data-rich environment that allows brands to accumulate information about users from a variety of different sources, accurately identify their interests and effectively target ads.
Today, more brands are giving contextual targeting a second look. Privacy concerns are a major catalyst, as growing restrictions on third-party data collection are putting a lid on the amount of data that advertisers can use for behavioral targeting techniques. The gradual removal of cookies and other identifiers from the digital advertising ecosystem means that brands increasingly have an incentive to turn to established, privacy-safe methods like contextual targeting. Contextual tools are also much more powerful than they were five or ten years ago. By leveraging machine learning algorithms and the faster computation methods that artificial intelligence provides, programmatic advertisers can quickly adapt to the changing language of the internet, refine their search parameters and identify more users that are likely to be interested in their ads.
Overcoming the transparency issue
There are several benefits to contextual targeting that go beyond its ability to protect user data privacy. The main strategic advantage is that it enables an advertiser to deliver messages to consumers when they are in a receptive frame of mind. When a user is browsing content about a specific topic, it signals their intent at that moment—often a more reliable indicator of purchase behavior than targeting based on previous browsing habits. This is borne out by data that shows contextual targeting can help lower cost-per-acquisition metrics for direct marketers. It can also be effective in driving brand awareness. Based on our historical analysis at StackAdapt, placing ads in the right context can increase user engagement by a factor of 4 to 10.
But challenges remain around transparency. Most automated contextual programs, including those operated by Google and Oracle’s Grapeshot, offer very little insight into how their algorithms work. An advertiser may provide a list of key phrases, but without URL-level reporting and real-time feedback on the contexts that the ads are appearing in, it is difficult to measure if their strategy is working. The advertiser is unable to know if they are actually reaching the people in those brand-specific contexts that work well with the messages.
Until now, many demand-side platforms (DSPs) have had to rely on third-party solutions that struggle to target niche contexts or provide limited transparency into how the targeting performs inside the proverbial “black box.” This was the rationale at StackAdapt to introduce our new Page Context AI tool, which uses a machine learning algorithm to expand targeting and include relevant phrases related to the context. The ability to target customizable contexts is essential to all marketers, but especially to those that require highly specific or niche targeting.
Let’s say an automotive equipment manufacturer wants to advertise its new air compressor brand. A keyword search with “air compressor” might only capture a handful of relevant websites, or those that only tangentially mention air compressors. That’s where the benefit of an AI tool comes in—to automatically generalize the input. For example, from the input of “air compressor,” natural generalizations include hundreds of other related terms such as the names of competitors (e.g., "Ingersoll air compressors,” “Champion air compressors”) or the product’s use in vehicles such as Jeep Wrangler and Chevy Silverado. An AI tool considers all of these related terms from a very limited input of “air compressor” to find the right context for the brand.
Page Context AI further allows the brand to verify that ads are served on relevant domains pre- and in-flight—a constant feedback loop that enables a brand to revise its input phrases in real-time to improve the success of campaigns.
Here’s how it works: The user inputs their in-context phrases and the tool then returns possible URLs where the ads may appear. That input can then be leveraged to further refine the input terms, either by adding more in-context terms, or adding out-of-context terms to disambiguate or remove content that they do not want included. In this way, brands can find the specific niche context they are interested in using. The tool also provides URL-level reporting of ad delivery. Brands can follow the execution of a campaign to understand the contexts where their advertisements appear in practice, instead of in the hypothetical possibilities offered pre-campaign. Through these feedback loops, brand gains transparency into their contextual campaigns and can adjust their messaging accordingly.
Spotlight on ethical targeting practices
The recent push by governments to expand privacy regulations means that marketers will be looking for more tools like Page Context AI that do not depend on cookies or other identifiers and can reach relevant consumers in a safe and effective way. This is an especially critical capability for marketers in sensitive categories, such as health and pharmaceuticals, where behavioral targeting is often considered too invasive and may be prohibited by law.
As an example, StackAdapt recently worked on a contextual campaign to drive awareness for a drug rehabilitation center. Using AI search engine technology to define in-context and out-of-context phrases, we were able to identify more appropriate targets with continually updated keywords and phrases such as “addiction,” “counseling,” “mental support” and “recovery,” while eliminating extraneous placements such as stories about entertainers and celebrity gossip (a typical outcome for a traditional contextual campaign). Ads were served on highly targeted websites such as blogs that offered tips on “how to talk with your spouse about rehabilitation” and “an open letter to a family member in need of rehab.” The campaign’s CPA performance was equal or better than the performance of an equivalent CRM/retargeting campaign, while its scale was larger than one from that kind of campaign. With the right context, this campaign successfully targeted users with intent, and produced the conversion numbers the client was seeking. And, this was done without the use of any private user information.
As marketers continue to navigate this new era of data privacy, some industry leaders have begun to raise even bigger questions about ethical targeting. In introducing their new contextual ad tool earlier this month, Vice Media executives spoke out against the practice of brands targeting audiences based on things like age and gender, calling it unethical and discriminatory. To be sure, that debate won’t be resolved anytime soon. In the meantime, brands have every reason—and now, exciting new capabilities in the field of DSP technology—to give contextual targeting a long-overdue refresh in the programmatic toolkit.
