Published on June 9, 2017
Look-alike modeling…we have all heard it, (especially in the AdTech setting) but have we swept the actual definition, value, and accuracy under the rug where we hide the rest of our unknown DMP marketing terms? Time for spring cleaning and a little back to basics, because lookalike modeling is a topic every marketer should be an expert on when looking to reach larger audiences. So what is look-alike modeling?
Look-alike modeling is a process that find groups of people (audiences) who look and act like any of your target audiences. Put simply, look-alike modeling analyzes your seed audience, identifies some key characteristics, and finds users who are similar to your target audience. The goal is for these new, larger audiences to act (or look) like your set of uniques to increase engagement. If you’re hoping to target people who are more likely to click on your ad or watch your video, look-alike modeling uses machine learning to find more users who will perform the action you are trying to make happen. It is used to scale audiences and increase the number of users your campaigns reach.
Let’s talk about “types” of people. You probably have a type that you are most interested in…right? Well, just like you have a type of person you gravitate towards, a company may have a type of person they want to advertise to; likely a person who has shown interest in their product or is more likely to purchase what they are selling. Sometimes it is hard to find that very particular person, but when you do, you want to hold on to them. In this case, not only does a business want to hold onto that person, they also want to find people who think, or act like them, so that they can reach a larger, yet more defined audience. Look-alike modeling does just that. It helps a marketer identify a larger pool of possible customers.
Not only does look-alike modeling help in prospecting, but it can improve your overall campaign reach. By finding audiences with behaviors that match up to your target audience, you have a more likely chance for them to convert in the future, which is your ideal prospecting situation; but what is also cool about look-alike modeling is that you can extend your reach of online advertising campaigns, and gain more of a following.
Let’s say you are a shoe brand looking to encourage more online purchases for your upcoming sale. The first step in lookalike modeling would be to place a pixel (a small amount of code) on your sale confirmation page and track the behaviors of purchasers (for that sale) as they move throughout the web. These demographic, behavioral, and preferential data points are then ingested into a centralized platform (such as a DMP) and analyzed via machine learning to identify which online behaviors are most common among people who buy your shoes on sale. Once these characteristics are identified, you can use those behaviors to identify more of your target audience. In short, this process uncovers people who act the same (online) as your customers, but may not know about your shoe brand yet; and so in turn, you can target them in your next marketing campaign!
Here’s the thing, though. You have to have your goals mapped out before building your campaign. You can choose between two, very different options, enrichment or extension. For example, the more defined your audience is (enrichment), the better chance you have of finding who you are targeting, although that number may be very small. Your other option, which more and more marketers have adopted throughout the years, is to consider a less strict approach (extension) by defining less attributes and behaviors for your target audience. This helps your overall reach and awareness increase instead of focusing just on higher conversion rates.
Data Management Platforms are a unifying technology that aggregates your data from any source. From these disparate data sources, you can centralize what you know of your most important target audiences. Once you have identified the seed audience, the DMP analyzes it, applies proprietary algorithms (smart words for “DMP magic”) and finds more user profiles that match the original audience. Lotame’s Audience Optimizer is the perfect representation of a tool that a DMP uses to help build intelligent look-alike campaigns.
In short, look-alike modeling takes your existing audiences and their behaviors, and expands it based on their behavioral and demographic characteristics. It offers marketers a more valuable approach to reaching new or current consumers in a cost-effective way.