Skip to main content

Finding the Best Lookalikes

graphic of two colored heads looking at each other

Online marketing allows companies unprecedented precision in targeting not just individual consumers, but, increasingly, “lookalikes” who match the characteristics of those consumers. New research out of YCCI explores the most effective approach to reaching these groups.

In 2013, Facebook (now Meta) introduced a novel way for marketers to find new customers: cast a net for “lookalikes.” In short, companies and nonprofits could create a list of people with desirable behaviors — past customers, say, or current donors — and then, from these “seeds,” Facebook would identify “matching lookalikes” who possessed similar characteristics. The concept took hold; most major companies, including Google, Twitter, and LinkedIn, now offer the service.

While an intuitively appealing practice, little is known about how to make advertising to lookalikes most effective. New research from K. Sudhir and Seung Yoon Lee, both at the Yale School of Management, along with Subroto Roy at the University of New Haven, helps unpack this issue by examining both sides of the equation. First, what kind of consumers should companies use as seeds, depending on their desired outcomes? Second, what kind of lookalikes should they target?

To study this, the researchers ran an experiment in partnership with an Indian non-profit that provides charitable support for the elderly. They used the nonprofit’s Facebook page to test different types of seeds and lookalikes.

Seeds

In the first experiment, they investigated which seeds to target based on what outcome the nonprofit wanted: did they want to expand brand engagement and make people aware of the work they did? Or did they want to boost acquisitions, in this case, donations?

In this case, the researchers found that when the desired outcome was brand engagement individuals who had previously demonstrated this behavior through website and Facebook clicks served as good seeds; this finding is especially relevant, they note, for young organizations that tend to be most interested in raising their brand profile but don’t have extensive consumer history data. When, instead, the nonprofit wanted to go after donations, it was better to move further down the consumer “journey” and rely on those who had, in the past, offered donations as seeds.

“Advertisers with the goal of brand marketing need not use downstream journey data to seed lookalike targeting,” the authors note. “In terms of [the] downstream advertising goal of acquisition (donation or purchase), lookalike targeting effectiveness increases down the journey stage up to seeding on website donations.” Importantly, the researchers found that drawing seeds from the top 5% of loyal donors did not significantly improve donations among lookalikes when compared to donors generally.

Lookalikes

The researchers next examined the kind of lookalikes that were most useful. Facebook allows companies to target their audience among the top 10% of lookalikes who best match the chosen seeds, in 1% increments. Sudhir and his colleagues found that at the beginning of the consumer journey (brand engagement), this match was less important. At the end of the consumer journey (donation or purchase), however, the quality of the lookalike match became more important: those in the top 1% of matches were much more likely to donate than those in the top 1-2%.

When seeding on downstream stages of the journey for the purpose of acquisition, focusing on the highest quality lookalike match works better.

Finally, when using a lower range of lookalike matches (between 1-2%), they found that highlighting the salience of an advertisement by explicitly telling the consumer the advertisement was recommended for them increases both clicks and acquisitions; this effect is especially pronounced for donations.

“Our results suggest that marketers can enhance ad performance by making targeting salient when they use a lower match among lookalikes,” the researchers write. “And since differences in match quality brings in different lookalike customers, pursuing such a segmented messaging strategy… is entirely additive (across two separate segments) for new customer acquisition. This is a managerially useful insight.” Overall, with a goal of raising brand engagement and awareness, firms need not use consumer journey downstream data to identify effective seeds or lookalikes. However, when looking to increase acquisitions, i.e purchase or donation, researchers found that match quality between seeds and lookalikes at pivotal moments in the consumer journey were a critical component to successfully placing targeted ads.