Ad network AdTheorent can now forecast if you’re likely to visit a physical store after seeing an ad
Partnership with location analytics provider Placed tracks which ad-targeted users actually visit, so its machine learning-created model can predict others.
Like anything that purports to tell the future, predictive analytics keeps discovering new uses.
In marketing, it is used to predict such things as who will become a good customer, what kinds of emails will get the best responses, and more.
This week, multi-platform ad network AdTheorent announced another use. Working with location analytics provider Placed, it is now offering predictive geo-targeting.
In other words, the joint effort predicts which recipients of mobile or desktop ads are more likely to respond by visiting the nearby physical location of a retailer.
Here’s how it works:
Let’s say that for a given retailer’s ad campaign, AdTheorent serves ads to users it can anonymously identify as being likely to visit the retailer because of user attributes, like the user’s location or interest in the products the retailer sells. It serves the ads to desktops, laptops, tablets and smartphones and identifies users anonymously through cookies and mobile IDs.
The New York City-based company has built up an ID Graph with about 140 million user profiles that contain behavioral and other attributes of the users, as well as the devices each user has. This cross-device matching has been inferred through IP addresses, geo-location of wireless signals and other clues.
Whether the ad is served to a computer or a smartphone, AdTheorent sends in real time the smartphone mobile IDs for those users to Placed. Placed then watches for those mobile IDs in the retailer’s stores, in some cases in specific sections of a large store. Placed tracks via GPS positioning (when it has been turned on by users), as well as through geo-fencing for mobile IDs and other techniques.
Predictive model for visits
During the course of the ad campaign plus a week afterwards, Placed sends back to AdTheorent the mobile IDs of the users who have been served ads and who have visited the store. As it does so, AdTheorent runs machine learning analysis on those users, looking for patterns about them. This might include, say, the kinds of apps they use, what they buy, what websites they visit on their devices, where they live and other characteristics. AdTheorent says it looks at demographic, behavioral, psychographic and geographic data points.
This creates a predictive model for what kind of users visit the retailer in response to that ad. AdTheorent then sends ads to other users with similar profiles to induce them to visit the stores, or it delivers follow-up ads to those users whose profiles resemble those who visit more than once.
This is similar to many kinds of ad targeting, where lookalike user patterns are employed to find new or repeat customers. Except that here, the desired and tracked action is the visit to a physical store.
CEO and co-founder Anthony Iacovone told me that, to his knowledge, his company is the only one currently employing predictive tech to forecast physical visits.
In a test six-week campaign recently run for an unnamed big box retailer, he said, 87 million ad impressions were delivered, and there was a lift in store visits of nearly 60 percent. The cost per visit: $0.33.
This is compared to other users who did not receive AdTheorent’s ads. They are among the 1.6 million in Placed’s panel of US users who have accepted incentives to install its tracking app.
Iacovone said his company is now working on adding the ability to track in-store purchases, so that likelihood can be added to its model and to predictive analytics’ resume.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.