Facebook Ads is awesome, but you already knew that. What you probably didn’t knew is that Facebook recently developed a support for Facebook Marketing Developers (such as ourselves) to create tools for split testing – which we did. Are you amazed? You should be. 

    Elevator pitch me, Sherlock!

    Finally, you have the possibility to split test bids, conversion events and billing events. This is something that has never been possible before, making it so revolutionary. Data-driven marketers who loves A/B tests out of independent and identically distributed target groups are drooling of happiness. So we hope you’re drooling too. Not on your keyboard though, that’s just unnecessary.

    Examples, please…

    Let’s say you’re planning to run a campaign towards a new lookalike audience you’ve created. You don’t know what bid is the most optimal and you don’t know what conversion event is the most optimal for you to optimize towards. Your main goal is simply to achieve the lowest CPA on the event “Purchase”.

    An example on what you can do with our feature (“Zalster Split Tests“):

    1. Choose an ad set (which has an audience you want to perform an experiment on).
    2. Choose what you want to split test.

    Let’s say we want to split test these four combinations of bids and conversion events towards each other:

    • Optimize towards the event ‘Purchase‘ with a bid of $33 per ‘Purchase’.
    • Optimize towards the event ‘Purchase‘ with a bid of $42 per’Purchase’.
    • Optimize towards the event ‘Add to Cart‘ with a bid of $11 per ‘Add to cart’
    • Optimize towards the event ‘Add to Cart‘ with a bid of $15 per ‘Add to cart’

    zalster split test

    What happens next is that the ad set (and its ads) you’ve chosen is automatically duplicated into the number of test groups your experiment results in. In this case, four groups => four new ad sets.

    The wonderful thing that happens next is that we inform Facebook that these four ad sets are part of a split test. What Facebook does next is that they divide the target audience into four independent and identically distributed target groups – making the split test statistically valid.

    Show me some case studies!

    Sure, let’s look at a split test, similar to the example above. This is a website conversion campaign. Our goal was of course to get as cheap purchases as possible. As you can see in the screenshot below, in this case, it was optimal to optimize towards the event Add to Cart with a specific bid ($15). This is something that is really hard, and risky, to just guess. The advertiser achieved 14% lower CPA compared to the next cheapest combination of bid and conversion event – and stunningly 31% compared to the most expensive.

    case_splittesting_51249

    Now, let’s look at a split test for a Click to Website campaign. As you can see, the test was divided into three groups testing three different bids. The most optimal bid was 33% cheaper than the next optimal bid (and 42% cheaper than the least optimal bid).

    case_splittesting_91241

    Curiosity strikes!

    If you want to try out this split testing feature of ours, just drop your email at our homepage.