Look-Alike Models help marketers increase reach by creating larger audiences that “look like” a smaller audience. Typically the smaller audience has known, deterministic attributes based on 1st party data or trusted 3rd party data. The larger look-alike audience has similar attributes but the known confidence level of those attributes is typically less than the smaller audience. For example, a marketer may want to target people who use the McDonalds mobile app or who have visited a McDonalds in the last 30 days. That audience in Skydeo yields a finite number of approximately 2.3 million people. We can expand reach by creating a look-alike model based on apps and places with high affinity to the original McDonald’s audience to 7-8 million people. This would offer a higher degree of accurate targeting than just targeting all fast food buyers. By loosening the targeting criteria a marketer can expand the scale and reach of their campaigns intelligently.
Best practices for creating look-alike models start with a trusted data set, typically first party CRM data a marketer knows. Targeting people who look like your best customers can yield the highest ROI. For game and mobile app companies, a common look-alike model starts with taking a list of users who made an in app purchase. Skydeo uses a variety of deterministic attributes to create high value seed audiences which can then be expanded by look-alike modeling. Attributes include: apps owned, locations history, age, gender, income, purchase history and device information.
The issue with look-alike modeling is that the desire for increased reach sacrifices targeting. Look-alike modeling, specifically in Facebook, can be over used to the extent the campaign isn’t really targeted at all. In the case of Facebook, marketers don’t really know which attributes are being used to expand the look-alike model. Skydeo enables marketers to have similar functionality to Facebook look-alike modeling but for use anywhere they want to advertise.
Skydeo’s use of deterministic attributes enable us to create exact-alike audiences based on similar apps installed, competitive apps or places visited as well as the standard age, gender and income metrics.
