In previous articles, L2 defined dark posts and made a case (using real performance results) for brands to use dark testing for optimizing posts on Facebook. This post highlights best practices for dark testing.
For review, “dark testing” is A/B testing on Facebook conducted by 1) building multiple variations of a single post by adjusting the message, thumbnail, image, etc., 2) serving these variations to different, similar audiences, and 3) measuring performance and designating a “winner.” Oftentimes, the winner is published to the advertiser’s timeline as a promoted post. Naytev highlighted the following best practices to achieve a significant performance lift on that post:
1) Go beyond A/B: Aim to test at least six post variations– generate all possible combinations then curate. For example, Funny or Die automatically uploads multiple variations of every possible creative element to the platform– image, headline, images, titles, descriptions, and video cuts. DramaFever, a Warner Bros company that distributes international televised content, looks at each combination of image and message as a different “story” and hand selects up to 16 per test. Clients see diminishing returns testing more than 16 post variations.
2) Vary Message: We recommend substantively varying the message versus changing just a few words, for example using dark testing as an opportunity to try out a range of brand voices and see which resonates best with your target audience(s). The Penny Hoarder, a personal finance publisher, tests many different tones (e.g., casual, humourous) before publishing any post, and enjoys a +251% click-through rate boost (vs. the publisher average of +104% boost)
3) Test Different Images: While this can be challenging for brands with limited creative team bandwidth or strict brand guidelines, it is key for driving meaningful results as the image is often the first element a viewer notices.
4) Pilot then Scale: While some publishers dedicate a full head count or even dedicate a team to testing 100% of their social posts, we recommend starting with at least 25% of posts to prove the benefit internally, then scaling from there.
An A/B test is only as good as its targeting strategy, and building strong targeting capabilities has been a major focus for Facebook. The primary targeting tactics Facebook offers are:
1) Target your website visitors by actions taken: This is enabled by placing the Facebook pixel on your website. From there, Facebook allows you to remarket to your website visitors on Facebook based on site behavior (e.g., abandon cart)
2) Target your customer database: With the Custom Audience feature, you are able to upload email addresses of those opted in to your CRM database to your Facebook ad platform.
3) Target specifics: Target by location, gender, age, income, page likes, interests, etc. Dark posts diverge from published targeted posts in that they also allow you to target by keyword, e.g., specific job titles.
4) Target people like these people: Lookalike Audiences, Facebook algorithms determine which audiences are “like” the audience of your choosing– termed the “Source Audience” (e.g., your website visitors, customer database, or a fan of your page) based on shared characteristics.
Tactics one through three are ideal for segmented messaging campaigns aimed at e.g., re-engaging lapsed customers, addressing cart abandoners, or product-specific messaging for customers with those identified interests. However, if the target audience is too specific, tests will take several days to produce meaningful results.
For optimizing posts that ultimately will be published to your brand page, we recommend the Lookalike Audience option with the source audience of your choosing, targeting approximately 5 – 10MM individuals (regardless of your fan base size) to build similar audience segments for testing. This will ensure adequate sample sizes so that the test can produce meaningful results in approximately 6 hours.
Beyond the Test
Whether you contract a third party partner to run a multitude of sophisticated dark post tests or are just trying out a simple A/B dark test with Power Editor, don’t forget to use this process as an opportunity above all else to learn more about your current (or potential) customer. Within the walled garden of Facebook, brands have the opportunity to understand audiences in granular detail based on their stated interests– this is radically different from the often imprecise cookie targeting on the broader web. We encourage brands to dive into the analytics and understand which interest groups respond best to which messaging and after deeming a test winner, take these learnings outside of Facebook into the entirety of your marketing communication.