How to A/B test social content
“Content is king.”
That’s what we hear over and over again. But what does it mean? Is that really meaningful, or is it just another throwaway line, the kind cherished by too many internet marketers, that sounds insightful but lacks substance when probed.
I suspect the latter.
Content is king, okay – but how do you know what surely bad content doesn’t rule, and how do we distinguish good content from bad content? It can’t just be gut feeling. The ethos of the growth hacker is data trumps instinct – always.
When it comes to social media, we’re often told there are “best practices” to follow. Look at the top search results in Google for a query about ‘social media content best practices’, and you’ll find a load of truisms and vaguely helpful, but impractical, tips.
It’s time for this to change. You can bring science to social media, and in doing so you can make social an acquisition channel with a respectable ROI.
One of the most important ways you can do that is by implementing A/B testing on your social channels. This is standard practice for other marketing channels, so why should social be exempt? ‘But… it’s hard!’ ‘There’s no one-click button to create a test!’ That’s right. A/B testing social takes a bit longer than A/B testing email, but it can be done. Here’s how:
Step 1: Determine a metric that matters. First thing’s first: pick a metric that matters. This should be something that’s relevant to your larger objectives. A common metric is conversion rate on a lead generation page. Let’s use that as an example.
*Step 2: Create social content that impacts that metric. *Now you need to create social content that moves the metric you’ve picked. In our example, the metric is conversion rate on a lead generation page. Now let’s add to our example and imagine we are testing Facebook content. We might then create two pieces of content for our Facebook Page, a photo with a link and a plain link, each sending people to the lead generation landing page.
*Step 3: Create UTMs for your content. *You now have 4 pieces of content all linking to the same landing page. Now you need to setup UTMs for your content so that you can tell which content drove conversions. You will change the campaign content tag to distinguish between the two types of content. Use Google’s URL Builder tool to create these links.
Step 4: *Reduce variations. *You have your content. Now you’ll want to reduce the variations between the two that are irrelevant to what you’re testing. In this case, you’re testing whether visits sent via photos or links are more valuable. So you want to isolate and eliminate variations that aren’t directly related to the differences between a photo and a link. That means you should do things like use the same copy in your status update and update your Open Graph settings to make the og:image property of your page similar to the photo you’re posting on Facebook.
Step 5: Create test segments. Now you’ll need to setup your test segments. In Facebook, you can do this without too much hassle using targeting. If your audience is large enough to produce a significant sample size, then you can push your A content (your photo) to everyone between the ages of 25 and 26, and push your B content (your link) to everyone between the ages of 26 and 27. This will give you roughly random samples of people. Repeat the test as many times as necessary using different segments until you get a valid sample size. Use this tool to determine what sample size you need to get valid results.
Step 6: Dig into Google Analytics. You’ve pushed your content, now you can dig into the Analytics. Login to your GA account and create a new dashboard to monitor results. Setup a new widget that displays in a table the metrics you’re measuring for the different types of content you’re testing. You’ll want to display the Ad Content in the first column and then the metric you’re measuring, as a Goal, in the 2nd and 3rd (if necessary) column(s). If you don’t know how to setup Goals, here’s a helpful primer. Then filter your data to only show you traffic from the correct campaign source and landing on the correct page. Voila! You now are able to learn how many visitors each piece of content is driving, conversion rate, value of a visit, and anything else you need to know.
This is a basic example, but it’s more than is usually done. You could take it even further and use the same process to test different messages or different images. Want to know when you should tweet to produce more sales? This can tell you that.
It’s time to stop relying on studies with questionable methodology to shape our social marketing programs. Likes and retweets are not real metrics. If a piece of content generates 10x the likes but -10x the sales, that content is useless.
It’s a common gripe that social isn’t taken seriously, but the blame for that rests squarely with people who put “Content is king!” at the top of a list post about ‘7 ways to make your content more engaging!’ That is meaningless. Science is meaningful. Start testing your social content and treat is a serious acquisition and sales channel. Anything less is no longer good enough.
If you liked this post, spread the love:
HackGrow is a learning resource for growth hackers, marketers, and anyone who wants to sell products or grow audiences online.
Get the best HackGrow essays delivered right to your inbox. Enter your email and signup for free today.