If you’re not familiar with the concepts of remarketing or retargeting, the gist is that you run advertisements to people who already know about you but presumably haven’t been willing to buy anything. Remarketing and retargeting are incredibly valuable tactics because in theory, giving someone a final nudge to purchase should be less effort and expense than convincing someone who’s never heard of you to do business with you.
In the world of paid media, one of the most powerful tools for remarketing and retargeting are social media custom audiences. Custom audiences allow you to advertise to very specific audiences, usually by groups of advertiser-provided IDs, such as mobile device IDs, social media usernames, or email addresses. Facebook rolled out Custom Audiences to all advertisers on October 23, 2013. Twitter rolled out Tailored Audiences to all advertisers on January 14, 2014.
We asked ourselves here at SHIFT, which custom audience platform performs better?
To determine this answer, we’d need to do as close to an apples-to-apples comparison of the two platforms as possible, with the same ads and audience, over the same period of time, with the same call to action. Coincidentally, one of our clients, the National Center for Families Learning, was looking for a way to promote their Families Learning Summit and drive registration.
With a budget of $100 (split evenly between Twitter and Facebook) and an email database of over 20,000 audience members, we engaged the custom audiences features on Facebook and Twitter to ascertain how effective each platform was in remarketing to NCFL’s existing audiences.
One of the challenges of this kind of test is that in terms of content format, Twitter and Facebook are very different platforms. Twitter Promoted Tweets don’t look nearly the same as a Facebook Newsfeed Ad. Despite these limitations, we created ads that were as close to each other as platforms permitted, using the same language, wording, and images. All ads ran on the same timeframe to ensure as close to an equal testing environment as possible.
What were the results?
Out of the 20,000 subscribers, Facebook was able to identify 7,400 Facebook users in common, while Twitter identified only 1,820. Facebook’s Custom Audiences reach was 36.89% while Twitter’s was 9.07%.
Out of 7,400 identified users, Facebook distributed ads to reach 3,234 unique impressions, or 44% of the custom audience. Twitter distributed ads to reach 284 unique impressions, or 16% of the custom audience.
Here’s where the rubber meets the road: how effective were the ads in driving traffic to the destination website? Despite a larger audience, more impressions, and greater reach, Twitter’s ads achieved a 5.63% clickthrough rate (CTR), while Facebook’s ads achieved only a 1.48% CTR. Twitter was 279% more effective at generating clicks.
The final perspective is on cost per click (CPC). Twitter landed a cost per click of $0.50, while Facebook reached $1.04 CPC. Not only did Twitter deliver more clicks, it did so at a lower cost per click.
It’s easy to leap to the conclusion that with a higher clickthrough rate and a lower cost per click that you should choose Twitter for your custom audiences, but that conclusion is premature. Custom audiences are by definition custom, which means they are truly unique to the organization. NCFL’s audience is likely to be significantly different from your audience. Additionally, because we didn’t have access to the back end registration system, we can’t tell which advertising system delivered more actual leads. That will need to be the subject of an upcoming test.
The experiment also does not take into account the social influence that the organization has; a company that has a highly engaged Facebook audience already should logically see Facebook ads outperform Twitter ads, particularly if a company has little or no Twitter presence.
The solution is to test using your own custom audiences and find out which works best for you. Follow the steps we’ve used in this test to create as close to an “apples-to-apples” testing environment, and re-test frequently.
Christopher S. Penn
Vice President, Marketing Technology