Look-alike modeling has been an important part of the media toolkit over the past decade, allowing brands to grow their audience pool by taking a core group of high-performing individuals, grouping them together and using data and technology to find others like them.
Over the past few years, data management platforms (DMPs), third-party cookies and their associated data have become obsolete due to self-regulation by technology vendors and legislation like CCPA and GDPR.
Moving away from third-party cookies and third-party data overlays on cookies results in a decrease in the size of total audience groups, as individuals have fewer associated identifiers (cookies to connect to).
However, lookalike modeling can also help companies leverage their first-party data to create robust segments at scale for marketing and advertising purposes.
Tealium Regional Vice President of Strategic Partnerships for the Americas, Travis Cameronexplained that the value of being able to expand target populations based on data associated with a high-value segment will take on a different dimension.
These changes range from the identifiers used (PII data hashed to match and expand) to the use of different types of data (contextual, interest-based, path) versus inferred demographic or psychographic data previously used.
“The value remains – marketers need eyeballs, and the quest to find individuals like those who just converted to optimize spend will always have high associated value as a tactic,” Cameron said. “It’s just going to get harder to model and grow your audience.”
Related article: Now you see me, now you don’t: navigate the world without cookies
Understand the machine learning model
“Given the complexity of using large amounts of first-party data to build segments, it is important for companies to understand how the modeling reaches its conclusions by understanding what features are used to build the model and their importance relative,” explained Alex Holub, CEO of Vidora. “Furthermore, it’s important to understand the ultimate goal of the look-alike segment.”
For example, if the goal is to maximize clicks on a marketing campaign, using a machine learning model that directly optimizes for clicks will perform much better than a semi-supervised look-alike model.
Holub explained that machine learning offers a few key benefits when creating similar segments:
First, machine learning continuously adapts segments based on the latest information available. By learning continuously, machine learning can incorporate new user activities or actions, in addition to broader secular changes that impact the behaviors of all users.
Second, machine learning implicitly learns the importance of user behaviors and attributes, allowing machine learning to take advantage of all available first-party data and implicitly increase or decrease the importance of each data point.
“In other words, machine learning can incorporate all available first-party data and learn what information is most important to the model,” Holub said.
Cameron noted that it’s important to have a clean, correlated data set of your customers that contains similar data on each consumer before you begin the exercise. “Know the result you are looking for. Any model you use should be developed and applied with an outcome in mind, so figure out what you’re trying to optimize with that audience,” he said.
He recommended using an agile approach to testing, learning quickly, and making sure you have a testing plan in place when you start acting on audience response. “Know where you plan to activate the audience and make sure you have the right identity points and integrations to activate it.”
Deployment of data science to target groups
Data science platforms can also offer a fast and reliable solution for companies to model their first-party data. For example, Vidora’s Cortex product helps companies create, understand, and integrate similar patterns into their business in days.
“For small teams and teams that need to move quickly, a data science platform can increase productivity and be of great value,” Holub said. He added that he currently sees many companies leveraging lookalike modeling as a black box – usually as a component of their DMP.
However, given organizations’ growing reliance on look-alike modeling using first-party data as a key revenue stream, he predicts that companies will begin to take advantage of more sophisticated data science techniques.
“Using data science techniques should result in both higher quality segments for brands to market, but also segments that align more directly with brand goals,” he said. -he declares.
For example, data science techniques outside of likeness modeling can create segments that directly optimize user engagement (e.g. clicks), down conversions, and increased brand sentiment (e.g. example, elevation modeling).
“I think we’ll see an increasing reliance on real-time data for similar patterns as the number of new users and anonymous users increases,” Holub added.
Cameron said companies with smaller adtech and data teams need to focus their efforts on developing and optimizing models and understanding their customers and the key data points that drive them, not engineering and cleaning their data.
“Reaching a state of data automation where they can work with all their audience data, leverage a few key partners to better deploy their audiences, and have tight measurement theses that allow them to run at the same pace as ‘a much larger organization,’ he explained.
Related Article: 4 Steps to Identifying Your Best Leads Using Tech Data