hawk

How to Use A/B Testing on Twitter to Be Better at Social Media

Snaptactix Staff Author

You know how important it is to test everything. Whether it is your Adwords copy, A/B testing different landing pages, or subject lines of emails – consistently testing will ultimately lead to better results no matter what your initiative is.

There have been countless blog posts on how to test just about everything.

But not too many on how to use social media for A/B testing.

Why You Need To Be Testing

Before we jump into testing, let’s first understand why it is important to test.

Testing allows you to take a control (A) and put it up against a challenger (B). You then split your traffic between each of the variations. In most cases you are looking to increase conversion rates. This can be an increase in sales, form submissions, or another action.

Typically, A/B testing can be handled inside Google Adwords if you are sending paid traffic to the page, or you can use a service like Optimizely or Unbounce to split up the traffic for you.

The advantage of using a service to manage the testing is that it will centrally house your results and even determine a winner once you reach statistical significance.

Once you have deemed a winner from your testing, you can then display the winner 100% of the time to your visitors.

You will be showing the higher converting page to all visitors and as a result increase your sales, leads, etc. as compared to using the lower performing page before.

As you may have noticed, you pretty much have to spend money one way or another to run these tests.

How To A/B Test For Free

If you are bootstrapping, or just like to do stuff for free the options a gave above cost money to use.

But check this out:

You can run A/B tests for free using just your Twitter account and Microsoft Excel.

I won’t go into all the details here, but this is how it works.

Know What You Want to Test

First, you’ll decide what you need to test. Since most tests focus around conversion rates, for Twitter I like to consider a conversion as a “engagements” as defined by Twitter. In order to increase engagement you can take one of these approaches:

  • Test two different titles
  • See what call to action works better
  • Which image creates more engagement

Use your imagination, there are quite a few things you can test if you get creative. The goal is to increase the number of clicks you generate with each tweet.

Don’t Forget To Set up Twitter Analytics

Before you start firing off tweets, it’s important to note that you need to setup your Twitter Analytics account if you have not already. Don’t worry it’s free, but it won’t start tracking any data until you set it up. It’s just a click or two to activate.

Even if you don’t plan on A/B testing, it’s worth having setup just to collect the data in case you ever need it. There’s a lot of great data in there.

Schedule Your Tweets

Next, you will need to schedule some tweets. Unless you have a huge audience or highly engaged audience you will need to tweet multiple times with each variation.

To schedule your tweets, you can use your favorite scheduling tool like Buffer or Hootsuite. If you’re not familiar how to do this, learn more about the details here.

Although it would be nice to achieve statistical significance, chances are that you won’t. And that’s ok.

A number of variables will go into how many times you should send these tweets. But I would try to send them at least 10 times or so over the course of a week. This will give your tweet exposure to more of your Twitter audience since everyone is not on Twitter at the same time.

Analyzing Your Twitter Testing Data

After you have tweeted out your two tweet variations a number of times, it is time to get your Twitter Analytics data into Excel.

You’ll need to go to your Twitter Analytics account and click on the “Tweets” tab. From here, you will need to set your date range to include the time frame of when you were sending your testing tweets.

Then click on the “export” button to download your data (remember where you saved it).

test1

In order to clean up your Twitter data for interpretation, you can use this spreadsheet that allows you to see your testing results.

Here are the key features:

  • Unshortens all URLs
  • Removes any query parameters from URLs
  • Correcting the date formatting from Twitter

Without these, it would be very difficult to get your Twitter data formatted to view your testing results.

If you want to see all the formulas and the reasons, they are listed here.

Now that you have your exported data and my Excel sheet setup, you will need to go to copy and paste your Twitter data to overwrite my placeholder data in the “Twitter Raw Data” tab.

Depending on how many tweets you have it can take a while to unshorten all the URLs. By a while, I mean that I have some reports that take 2-3 hours to go through all the tweets to extract the full URLs.

So if you have a large amount of tweets, you can probably step away from the computer for a while and let Excel works its magic.

Once all your URLs are un-shortened, you will want to highlight the “Unshort URL” from the top row to your last tweet and paste the values back in place. So here are the steps to do that:

  • Highlight the “Unshort URL” from the top row to your last tweet and hit “control + c” on your keyboard
  • Right-click with your mouse and click the “paste values” option. This will look like a clipboard with the numbers 123 on it

What this will do is remove the formula that was there and overwrite it with the data the formula created. This will prevent you from having to sit through and wait for URLs to get unshortened again if you modify the table accidentally.

Next, go to the “Analysis” tab. Click on the Data tab in the top ribbon and then click on “Refresh.” This will show all the data you imported from Twitter.

You’ll probably see a bunch of your data from Twitter, but we’ll adjust your pivot table to only show the test tweets.

To do this, click on the filter of the column for “Row Labels” and select the URL of the page your tweet was linking to and click “ok.”

Your filtered results will look like this.

^D83CA712B1BA39A04CF4C443CA66DA40CC2399731BF23D4E89^pimgpsh_fullsize_distr

From there, you can see what the text was you used for the tweet and these points of data:

  • Number of times this text was tweeted
  • Sum (total) of impressions for this text
  • Sum (total) of retweets for this text
  • Sum (total) of replies for this text
  • Sum (total) of likes for this text
  • Sum (total) of URL clicks for this text
  • Average engagement rate for this text

Depending on what you feel is the most important metric to determine a winner, you can also modify the pivot table to show other data points.

For example, you can show averages instead of sums for the columns. Or you can add in new columns like “engagements” and have that column show an average for the engagements for each version of text.

There are quite a few ways you can tweak the data to show exactly what you need for your goals.

Use Your Testing Data To Improve Social Media

Just because this data came from Twitter doesn’t mean you can’t use it to modify your approach on other social media channels. Unfortunately, the other channels don’t allow for this type of export.

Once you see what your data is telling you, you will have a better idea of what works well and what doesn’t so you can focus on approaches that drive results.

By having better insights in how your audience engages with different messaging, moving forward you can craft your tweets to utilize this tone. Additionally, you can use it to guide your blog post titles as well since many social updates utilize blog post titles in them.

How do you plan on using this approach to improve your social media? Be sure to let me know in the comments below.

Jordan Silver

Jordan Silver

Affiliate Relations Manager at Snaptactix
Jordan has been involved in the digital marketing world since 2005. Specializing in social media, affiliate relations, and copywriting. Jordan is also a grizzled veteran in start-ups.
Jordan Silver

Comments

Leave Comment Without Facebook

Leave a Reply

Your email address will not be published. Required fields are marked *