A/B testing, or split testing, should be an integral part of launching any new digital ad campaign. The Brand Architect integrates testing into all our clients’ campaigns. It’s nearly impossible to be sure what will garner the most attention, whether that be the simple difference in headline, imagery, or design. Simple differences can make a significant impact, so if you’re not already a split test aficionado, read on!
What is split testing?
Split testing is a concept, but it is also a feature of many online platforms. The concept is simply to test two different things for a period, see which performs better, and then scrap the lower performing one. Facebook, Google Adwords, Google Analytics, Mailchimp, and others even have specialized features that make split testing extremely simple to do, and the results simple to understand. Even if you are positive that your ad is spot on, there may be certain visual cues that effect potential customers in subconscious ways that you’d have no way of knowing before just giving two different options a try.
What to test
Ads and landing pages should be split tested for things like:
Audience – age/gender/location etc
Call to action – naming conventions and colours
Understanding your objective
Make sure before you set off on a testing frenzy that you clearly understand your objective. Considering where in your funnel you are trying to optimise, will help you focus your testing efforts. Where is it currently broken or perhaps not performing to expectation – is it not enough clicks or traffic to the site, is it a lot of traffic but not enough conversion? Once you understand your objective then you can hone in on the right testing approach (and what it is your testing e.g. website landing page or facebook ad by way of example).
Take a measured approach
The best way to stay measured is create yourself a testing calendar hanging from your objective identified. Consider how much time you need to test each element so that it is statistically relevant – it might be a guess at first, but as you launch into you will be able to soon see if you need to cut it short or extend it pending how quickly the data is gathering. Take note of the result and set that as the new benchmark. Continue through the test program to you get to a point where you are satisfied your objective is being met.