Look-a-like modeling always intrigued me. Given the data and methods that have fueled look-a-like modeling to date, the whole premise has seemed kind of far-reaching:
“Jackie is a high-value customer. She has a HHI of $150K+, lives in a major city suburb, has children in the household, shops big-box retailers 2x+/month, and likes bargains. We need to convert more people like Jackie; let’s target a Jackie look-a-like segment based on these attributes and expect the same outcomes!”
Well, let me tell you something. I’d fit nicely into the “Jackie” segment, but my sister (Jackie) and I are very different people (evidence, right), and our shopping habits couldn’t be less alike (what mom doesn’t love Target and bargains?).
Luckily for the retailer above and all other advertisers out there, AI has hit the marketing scene. Jackie and I will never be deemed “look-a-likes” again. The game is changing… in a really big way.
Sans AI, look-a-likes are users who “look like someone who may take/has taken a desired action.” Best practice (i.e. best chances) require identifying people with known characteristics from 1st party data or past campaign data. But as many marketers and most agency folk know, the vast majority of look-a-like models rely heavily (if not wholly) on 3rd party data. Segments are built from 3rd party data that “mimic” favorable attributes, and those segments are targeted with ads in market. This modeling is static; void of intent signals, time and recency, contextual considerations (e.g. channel and publisher impact), etc. No rocket science, or even much data science; pretty basic stuff.
But with AI… oh, with AI… it’s a new world. Look-a-like modeling will never look the same again.
AI brings unprecedented machine power, like the ability to proactively identify and analyze over 1 million marketing opportunities per second, and appraise likely outcomes with precision that mere mortals can never hope to achieve. Each impression up for auction can be interpreted based on every historical interaction the AI has encountered, and hundreds of thousands of variables can be considered at each decision point to identify the best possible outcome and make it a reality (mind…blown…).
Bring that brain power to look-a-like modeling, and what do you have?
“Look-a-likes” based on (1) innumerous data variables from a variety of 1st, 2nd, and 3rd party data sources, (2) real-time, predictive propensity scoring (3) precise, calculated decisioning at the user-level, and (4) continuous AI learning and optimization – not simplistic behavioral attributes that may or may not be indicative of a desired outcome.
Basically, AI just took look-a-like modeling from “these people might engage” to “this individual is high-value because they are [x] likely to take a desired action. Oh, and they can be reached right [here], and the lowest possible cost and highest return to you, so I went head and did that for you.” Thanks AI.
Sure, there’s much more to it than that. And yes, each Marketing AI will have its own variety of look-a-like flavors.
At Magnetic, our AI powers quite a few look-a-like models. Most recently, we’ve rolled-out our Look-a-Like Conversion Model.
The look-a-like conversion model values and targets users based on their likelihood to visit an advertiser’s site and/or convert within 24 hours of the impression being served.
Yup, unlike look-a-like models to date, it uses two of the most powerful predictors, time and intent, as well as context (how more/less likely someone is to take action based on where and how they are messaged). Magnetic’s AI can even account for recency/frequency, depth of prior engagement, prior interest, etc. to very accurately predict where each person is in the decision cycle. This all means… our AI knows “this person is x% likely to buy tomorrow.”
A lot. Magnetic’s AI applies that knowledge to its bid optimization process to deliver a slew of value to advertisers, such as:
What marketer wouldn’t want that type of efficacy and efficiency? I know the guy building “Jackie segments” would be interested, that’s for sure.
Want to learn more? Contact us to find out how Magnetic’s AI can transform your media strategy and make sure your look-a-like model looks nothing like the directional-at-best models of yore.
Magnetic is a digital marketing and artificial intelligence company. We use machine learning and AI to deliver smarter, faster, and more effective advertising. Our powerful AI platform continuously analyzes the attributes of 320 million live user profiles alongside real-time inventory supply and bid opportunities to deliver highly performant and profitable campaigns for our clients.