a/b testing

A/B Testing And Your Social Media

abtestingsocialSo when we traditionally think of A/B testing, we usually think of testing a variable while leaving the time period and audience the same. We test the homepage headline during a two week period in August, where half our website visitors see THIS HEADLINE and the others see THAT HEADLINE and we see who clicks through multiple pages or sticks around on the website longer or is more likely to make a purchase. (You have to pick what you care about, since caring about everything is not only exhausting but not possible in a testing sense). If you want more about the basics of A/B Testing and how A/B Testing could work with your website, check out those linked blog posts.

You may ask ‘How can I do A/B testing on social media?’

We can’t show half our Facebook audience Image 1 and half Image 2… so we have to think of social media testing a little differently.

Using Social Media to Test Timing

One thing you’ll notice, in particular with Facebook and other social media sites that work on algorithms, is that most people never see your posts… which means you can repeat stuff with a) little to no chance of the same people seeing your message and b) even if a couple people did see a repeat message, it might not register considering the volume of other messaging they see from everyone else.



Posting the same thing at different times may feel spammy but it can give you a good idea of when a given post is most likely to be seen. It’s not true ‘A/B’ testing but similar in that we’re testing only one variable (in this case, timing versus content) and using data to give us feedback.

Using Social Media to Test Messaging

You have probably noticed those sexy ‘carousel’ ads on Facebook (the ones with multiple images). Here’s kind of a dynamic view of how they work from Quero Media:

Facebook-Carousel-Ads-1

What these kind of ads allows us to do is display different content at the same time to the same users… and see what converts them.

This is a more accurate application of A/B testing since we are showing the same users at the same time multiple content choices and seeing what they react to.

Now you’ll notice with the example above, both the image and text vary in each picture. In the best test, we’d only have one variable be different (the text or the image). Being able to see, for instance, that the Premium Italian Leather image getting more clicks, the company could further test by using the same headline with different images, seeing if indeed it was the photo or text that caused more interest in the messaging.

Like this month’s A/B testing theme? Stay tuned for next week when we talk about A/B testing with email newsletters! Subscribe here.

A/B Testing And Your Website

abtestingwebsiteEvery month here at Breaking Even, we try to have a loose theme. This theme determines a series of blog posts, our monthly email newsletter subject, and gives us an opportunity to dive into a concept we think people should care about.

How many of you have said one of the following things?

“My customers would never be interested in that.”

“My website visitors can get to my shopping cart just fine.”

“I never go to our website on my phone.”

We ALL say them (I am including myself here) because, after years in our business, we have a pretty good idea of what our customers want and don’t want.

Thing is, we don’t know what anyone thinks. Ever.

We can guess. But we can’t know.

We can really only manage what we measure.

So in order to get an accurate view of what is and isn’t happening is present the same set of customers with two almost identical options with one variable change (ex: color of a button) and see how people react. This concept is called ‘A/B Testing’ and is this month’s theme! If you want to know some basics about A/B testing, I wrote a lovely blog post about it previously. Click here to read it.



OK, so what can you test on your website?

  • Color of a button
  • Position of a button
  • Size of a button
  • Writing on a button
  • Lead photo on a page
  • Headline of a page
  • Size of product photos
  • Featured product photo
  • Length of a form
  • Color of text
  • Text in your menu
  • Position of your menu
  • Size of writing in your menu
  • Number of photos in slideshow
  • Order of photos in slideshow
  • Changing a step in an order/checkout process (ex: wording on a particular page)
  • Changing position of sidebar (or adding a second sidebar)
  • Changing position of item in sidebar/footer
  • Changing color of sidebar background
  • Length of testimonials
  • Size of social media sharing buttons
  • Position of social media sharing buttons
  • Lots more!

I wanted to make that list somewhat ridiculously long but I think you get my point.

What should you test?

Deciding what you are going to test to start off with may be as simple as deciding what’s important to you. For example, if email signups are important, you may test an email signup positioned in header area versus sidebar and see how this change affects signups.

You may want to start with testing something ‘controversial’. Maybe everyone in your company argues about whether your products should have a plain wood or a plain white background.



When should you test?

You obviously need to test long enough to collect some data, so when and how long you run your test may depend on how much web traffic you get, among other factors. That said you probably don’t want to do this during a really important time, say, during the holiday rush. Think long enough to get data to make a decision and not so long that it annoys everyone.

How often should you test?

You don’t want to be in testing mode all the time, so plan your year accordingly. Quarterly tests may feel often enough, or twice yearly may be more than enough. Just put it on your calendar/agenda as something to do at a regular interval.

How do you test?

This is probably your biggest question: how do you do split testing on a live site?

You have three options:

    1. Using your current website software (ex: Wordpress) with Google Analytics goals. There is a robust discussion and how-to here: http://www.wpbeginner.com/wp-tutorials/how-to-ab-split-testing-in-wordpress-using-google-analytics/ Note: this is somewhat limited as, for example, it would be somewhat impossible to use this method to change, say, menu text as it works best on main page content.
    2. Use an add-on to your current website software. So there are tons of Wordpress plugins that allow you to, say, split test headlines “Title Split Testing For Wordpress”, or do more general changes. Some ideas here: http://wplift.com/ab-testing-in-wordpress (Note: just because a plugin works, doesn’t mean it’ll work on ALL websites so test test test! Want to know how to pick a plugin? Read this.)
    3. Use a third party A/B testing service like Optimizely or VWO. Check out their complete guides to A/B Testing, here for VRO and here for Optimizely (not affiliate links). Note: Remember if someone makes something easy for you, you’ll likely have to pay them.

If you’d like to learn more about A/B testing this month, subscribe to both our blog and monthly email newsletter.