I have recently stumbled upon a video explaining about the ingredients in McDonald’s fries. Turns out that everyone’s favorite fast food includes 19 different ingredients. Putting aside the criticism and health-related concerns that spawned after the release of this video, it is clear that in order to create a consistent product with the same texture and taste all year long, McDonalds needed to include many additional ingredients on top of the simple potato. I think most customers understood that these fries have more than what meets the eye.
Okay, so you are thinking, what do french fries have to do with native advertising? Well, by looking at a publisher’s page with native ads and content recommendations you may not think much of these native ads just like you would not think much of the fries. However, the decision to deliver these native ads and recommendations is not coincidental. In a world where performance, engagement, and user experience are of upmost importance, the technology behind this decision can be the difference between success and failure. In this post, I will take a behind the scenes look at the various native advertising recommendation and matching technologies, as these are the secret ingredients behind effective native advertising.
Basic Campaign Parameters
Traditional advertising systems allow the advertiser (demand-side) to define various targeting rules for each campaign. These rules can be basic, such as publisher sites, audience and GEO, to more elaborate rules such as timing, location, demographic, and more. This type of targeting is, of course, very basic, as it makes broad assumptions regarding the relevancy of ads. On the other hand, they allow some degree of control as to where the ad will appear.
Contextual matching technologies analyze the content of the designated article or web page (on the publisher’s website) in order to match relevant ads. For example, a page that discusses a health-related topic may be relevant to health-related ads. Most native advertising platforms and networks use contextual matching as the means to deliver relevant native ads to users.
Of course, some technologies and algorithms may vary in their sophistication levels. Advanced technologies may make an inferred connection between the content of the page and the topic of the ad. For example, they may find a connection between the mentioned health-related page to a broader topic such as fitness. Broadening the scope of related topics is essential for the user experience (to avoid monolithic recommendations) and the utilization of the ads inventory.
Although this form of matching is effective to some extent, it is still restricted to the content of the page, which may not represent the entire spectrum of tastes and interests of the user.
User Behavioral Matching
This technology focuses on the user itself. It analyzes the users‘ behavior, from the feedback that users implicitly provide during content consumption throughout time to create a high-dimensional preference model for each individual user. The model is then used for real-time content and ad selection. This type of targeting is more effective, as it captures the user’s tastes and preferences, their moods in various times during the day, and the specific content consumption habits. The targeting is not only effective, it can utilize more campaigns and match them to the existing content. For example, a female user visiting a sports site may receive an ad about baby formula. Using only contextual matching, this ad will never appear on the sports site. However, through the use of user behavioral matching technology this is certainly possible.
Collaborative filtering and trend analysis
Collaborative filtering is a targeting method that is based on the collective preferences or taste information from many users. The underlying assumption of the collaborative filtering approach is that if a person A has the same opinion as a person B on an issue, A is more likely to have B’s opinion on a different issue x than to have the opinion on x of a person chosen randomly. This type of technology is commonly used in e-commerce sites, by displaying related items that interested other users. This method is of course also relevant for the matching of native advertising. In this case, items are replaced by content recommendations and native ads.
Collaborative data is also used to perform engagement trend analysis. For example, the technology understands which articles are “hot” or trending at a particular publisher’s site and then to factor-in this data in the targeting optimization.
External data signals
Advanced matching technologies also factor-in external signals from the publisher’s DMP, social networks, SSPs, etc. These signals can include demographic data, gender, commerce profiles, and more. These signals enrich the set of parameters used in the matching optimization. For example, data from the publisher’s DMP can distinguish between paying subscribers and non-paying users. By using this information, the technology can target specific native ads/content to the paying subscribers versus the non-paying ones, based on its optimization algorithms.
Mixing all the ingredients at once
As we can see, all the technologies (ingredients) mentioned above can contribute to the native ads matching optimization. However combining them all at once is extremely challenging. It’s like having multiple opinions to make a single decision, which ad to serve. my6sense’s native advertising platform uniquely combines all these ingredients in real-time. Every time an ad/content recommendation is served, a sophisticated algorithm uses all these ingredients to help make this decision, while applying different “weights” in order to maximize the monetization for publishers, performance for advertisers, and the best user experience for readers.
To learn more about my6sense’s technology – click here