Facebook Ad Testing, Revisited

Briefly, A/B testing is the technique of running similar ads side-by-side in order to test individual variables like targeting and creative. Any advertisers who are delivering consistently outstanding results do this kind of testing constantly. It’s the backbone of any competent digital advertising program.

A lot of A/B testing effort is concentrated on Google Adwords and on Facebook advertising. These are the two platforms that provide advertisers the most flexibility and sophistication. Of the two, I’ve generally found Facebook more difficult to optimize. Facebook, in contrast to Adwords, relies heavily on an opaque ad serving algorithm to generate most of its success, and therefore allows fewer tweaking options to the advertiser. You can be a competent Facebook advertiser and still be completely confused as to why certain campaigns work better than others.

“Okay, so I A/B tested my Facebook ads, and still nothing’s working. Now what?!”

There are about 50 good articles out there on how to set up 101-level Facebook ad A/B testing. Here’s a good article from Adespresso. Here’s another one from Wishpond. The real challenge of Facebook ad A/B testing is not really in setting up the initial tests; it’s what to do in the frustrating situation when all your tests are inconclusive and none of your ads are converting. I’d like to talk a little bit about going past the entry level of testing. Here are some hard-learned Facebook optimization lessons that you might want to try next:

  1. Did you let the experiment run long enough before shutting it down?

    I’m very sorry for bringing statistics into the conversation. Most people took marketing classes in order to avoid statistics in the first place. There’s a concept called “statistical significance,” which very basically tells you whether or not you have enough data to know whether the effect you’re seeing is real or could be caused by random chance. Most people look at a few days worth of data, and mistake the initial differences they see for a real effect. Here is a simple calculator that can help you know if you’ve collected enough ad data to form a conclusion.

  2. Did you remember to throw out the first three days of data?

    Any ad set that you create will be horribly inefficient at first, depending on the pace at which your goal metric comes in. If you’ve set up an email conversion campaign, and you’re only receiving 2-3 emails per day, the optimization process will take longer than a campaign receiving 20-25 emails per day. In general, the first few days of data in any ad set is unreliable for testing purposes.

  3. Did you use the “Breakdown” reports to gather insights?

    Many people overlook the Facebook breakdown reports in the Ad Manager. Viewing the ad breakdown report can show you data on your ads split up by age, gender and ad placement without having to create separate ad groups to tests these variables explicitly. That means you can concentrate your split testing on creative and audience targeting.

  4. Did you try a lookalike audience based on conversions?

    In an ideal world, your visitors would first find you through organic (i.e. free) means. You would then spend your Facebook ad budget on retargeting these visitors for some kind of conversion, knowing that targeting recent web visitors is more efficient than targeting people who’ve never interacted with you before. This is one of the more effective Facebook funnel structures.

    This is the real world, however, and sometimes you’re going to have to advertise to people who’ve never come to your site before. You might be running a lead generation campaign to strangers, for example, directing them to some kind of lead magnet. Be sure, if at all possible, you try using a 1% lookalike audience for this, based on a custom audience of people who have already converted. I’ve had more immediate conversion success with lookalikes than any other form of primary targeting (i.e. non-retargeting).

  5. Did you experiment with manual bidding?

    This tactic is tricky, but has come in handy for me…particularly in circumstances where I have to hit a particular CPA metric. In your conversion campaign, start a test ad group and select “Maximum” manual bidding. Then enter a conversion figure three or four times the amount of your target CPA (this might seem weird, but trust me). If the ad group doesn’t deliver after a few hours, increase the bid until it does. Once the group is underway, start gradually lowering the bid by 10% or so each time. Wait about six hours, then do it again. See if you can fix a bid at a point where the ad set is spending most of its budget and the CPA is acceptable. Using this technique, your ad groups might not spend down their budgets completely, but it should help you control your conversion costs.

    In the background, Facebook will use your bid information to refrain from bidding on impressions it deems to be too expensive. This is why your budget might not always spend down. I’ve used the same technique successfully for click-optimized campaigns as well.

  6. Did you try calling Social Media Beast?

    Social Media Beast specializes in social media content marketing and advertising. Our team can move Facebook ad campaigns past these frustrating hurdles, and set up true multi-stage marketing funnels to generate returns on your marketing.

    Give us a call at 312-521-5200, or contact us through socialmediabeast.com today!