The idea behind Facebook Lookalike Audiences is simple: upload a list of customers and let Facebook find more people likely to buy.
Each Lookalike Audience is available in a range from 1% to 10%. This describes what percentage of the population is being matched. For instance, a 1% Lookalike Audience in the US targeting the 1% of the US population that looks the 'most' like the seed list. In theory, as you increase your percentage, the similarity with the seed group declines and, with it, similar activity, such as purchases. But this change in behaviour isn’t always smooth.
The intent pool
Behind this, there is another, more important layer that actually drives how well a certain Lookalike Audience will work: intent pools.
Every category and every brand has a number of people who are ready to buy at any given time. This group of people is much closer to the bottom of the purchase funnel and will need fewer touchpoints to convert than consumers further up the funnel. Facebook has likely made this intent data a key part of their Lookalike Audience algorithm. When you ask Facebook to find people who will convert, Facebook will go after this intent pool first.
Having observed the pattern across many different clients, verticals, bidding methods, and formats, we believe these intent pools are the key driver for Facebook performance:
1. Limited reach in Lookalike Audiences
In many Lookalike Audiences, reach is limited at 40-80% of the total audience size, no matter how high bids go.
We believe that the reach tops out because the intent pool is smaller than the total Lookalike Audience. When bidding for outcomes, Facebook expects that performance will drop when targeting users outside of that pool, so does not show the ads.
2. Step-function Lookalike Audience performance
Increasing Lookalike Audience percentages does not result in smooth changes in performance, but is often a step-change in performance (for example when moving from a 1% to a 2%).
When performance is driven primarily by an intent pool, not broader characteristics of the target audience, once it is exhausted, Facebook will perform like standard display media.
3. Higher retargeting CPMs
In Facebook’s ad auction, the price you pay depends on what others in the market are bidding. Regardless of how you are bidding and paying, all auctions are effectively for impressions.
We found that retargeting CPMs are frequently much higher than prospecting CPMs for the same audience (as high as 10x). Why does the price jump so dramatically once someone visits your site?
The answer is the intent pool. Once a user has visited your site, Facebook’s pixel recognizes that the user is in an active shopping mode. They are now included in the intent pools for Lookalike Audience bidders across the vertical.
Implications of Facebook intent pools
Lookalike Audiences are, and will remain, an important part of Facebook strategy, but the understanding that underlying intent pools have a bigger impact on performance has some implications for the industry:
1. Media planning
Recognising that Facebook offers two distinct media types – mid/lower-funnel Lookalike Audiences and upper-funnel awareness media – means that planning will have to adapt to balance each role.
2. Scaling performance buying
Limited intent pools mean that good performance against Lookalike Audiences can’t scale beyond demand. Like search, if advertisers want more of what works, they will need to expand to brand and awareness advertising.
3. Pixel implementation trade-offs
If visitors to advertisers’ sites are being included in competitors’ Lookalike Audiences, some brands may want to consider whether the benefits of Facebook’s pixel outweigh the potential to lose customers at the last stage of the funnel.
Regardless of your goals, when planning and buying Facebook media, you will likely improve your outcomes by considering the size of your intent pool and how that fits with your media mix.
Nelson Elliott is head of biddable media at Croud USA.