What is lookalike modeling, anyway?

Who?

Nevermind. So what is lookalike modeling?

Here are the essentials: These are statistical models that can determine if prospects are like your best, most lucrative customers across a number of behaviors or other attributes, from demographics to psychographics.

Your top customers become your “seed pool,” the group of consumers whose various common attributes determine what’s key in identifying more high-value prospects like them. You then use the model to check if a prospect fits that profile. If they do, you target them with ads.

Do I use online cookie data to find them?

You can, but when you do, it’s better to think of them as “act-alike” models. They’re usually focused on the bottom of the sales funnel since they’re targeting consumers who have shown some prior interest through their online behavior and actions. You’re often trying to turn clicks and traffic into sales.

But isn’t this just retargeting?

Not really. Retargeting exclusively involves serving ad messages to consumers based on their previous contact with your touchpoints (e.g. if they’ve already visited your site). Lookalike modeling is designed to target new high-value customers based on many different attributes.

OK, so how do I do it?

You start by pulling online profile and behavioral data on your seed pool from your own customer database. You plug this into your model and compare it against third-party consumer data that contains these same attributes. When you get a match, target this new prospect. Prospects found in this way are much more likely to purchase your product or service than the average consumer.

That sounds simple.

Not so fast. Lookalike modelling can vary based on your goals, and the data you use to create your model can vary. Using cookie data is fine, but it’s limited to finding lower-funnel prospects, those customers ready to buy.

What about finding customers further up the funnel?

That’s where offline data comes in handy. Unlike the “act-alikes,” which are based on online behavioral data that’s only really useful for about 7 to 14 days (online intent is pretty ephemeral), here you’re looking at a purchase process that can take anywhere from weeks to month. Consumers might not have even thought about purchasing just yet.

But the offline data…?

I’m getting to it. You’re hitting consumers mid-funnel, so you want data that’s more permanent, or that has a longer shelf life.

Offline profile data, like household, financial and other information, will do the trick. And agencies can get access to a lot of this through third-party services. By matching the high-value customers in the seed pool to their offline data and looking for patterns, agencies can determine what to look for with their models.

Using offline data also allows you to reach prospects who might not have a heavy web presence. While some people might make their online behavior untraceable by routinely wiping their cookie cache, everybody has financial and location data.

So you use this offline data to find new prospects and get your messages in front of them, but how do you know if it’s working?

You should be looking for an increase in both online and offline sales, but pay specific attention to any increase in revenue from each new customer. Remember, you’re looking for high-value consumers, so each individual that you prospect and target using lookalike modeling should result in a higher ROI than an average customer.

This is a lot to take in. There’s really no perfect option?

Not really. Like I said, it depends on the objective and the available data and technology. Sometimes you’ll want to use behavioral data, sometimes you’ll want to use psychographic data and sometimes you’ll want to use a combination of these and more.

The key is to know what you’re looking for in a high-value consumer, what part of the funnel they’re in and what lookalike tactics will point them out.