How to Split Test Your Twitter Marketing

social media how toWould you like to improve your Twitter marketing?

Are you testing to see what gives you the best results?

Traditionally, split testing is a practice that’s applied to individual pages on your website to compare page variations and ultimately determine conclusively which version results in the highest overall conversion rate.

However, implementing split tests on your website alone is short-sighted! In fact, you can harness the power of split testing to generate valuable information about your target audience and their overall engagement with your brand on social media websites.

To see how to do this, let’s go step by step through how to carry out a split test on Twitter.

Step #1 – Create Your Split Test Tweet

The first step in the split testing process is to craft the text of the tweeted message you’ll use.

As an example, suppose you want to determine when your Twitter followers are most active. While services like Tweriod will estimate popular times, narrowing down these ranges and confirming their effectiveness using split testing will help you determine when future tweets should be published to reach the biggest, most engaged audience.

single call to action

Sample split test tweet features a single call to action prompting followers to click through to a company blog.

In this specific instance, you’ll only need a single tweet, as the test variable in this case will be the different links used to track when followers are most likely to click through to your site.

In other cases where you’re testing the content of your individual tweets against each other, you’ll need to generate more than one split test tweet.

Obviously, you’ll want to include details that are more relevant to your company’s value proposition; this example is written to be intentionally generic.

In addition, be sure to leave space at the end of your split test tweet to include the test URLs we’ll create in the next step.

Step #2 – Use Bit.ly to Create Split Test Links

Bit.ly and similar URL shortening services allow you to turn long web addresses into shorter variations that fit more easily into character-limited social media updates.

However, they also give savvy users the ability to carry out social media split tests using URLs that vary based on your chosen test protocols.

Following the example above, we need to create two (or more) different Bit.ly links that will be deployed at different times of the day. Although all of the links will redirect to the same location, their individual click-through rates will be used to determine when the sample site’s social media followers are most active.

To do this, head over to Bit.ly and log into your account (or create one if you haven’t set one up yet). Once inside, paste your destination URL into the field in the upper right-hand corner of the page and click Enter. Doing so will pull up the following screen, which includes your unique shortened link.

create shortened links

Bit.ly allows you to create shortened links that provide experimental data for your split test.

Create as many unique Bit.ly links as times during the day you plan to test, with each individual Bit.ly URL corresponding to a specific timed test tweet.

For example, if you want to test the responsiveness of your audience at 10:00am, 4:00pm and 8:00pm, you’ll need to create three separate Bit.ly links, each with a different reference code and each pointing to the same page on your website.

As you do this, be sure to make a note indicating which Bit.ly URL you plan to use during each test period (in this example, the Bit.ly URL pictured above might be associated with the tweet that will be sent out at 10:00am), or consider customizing your Bit.ly URLs to include this information.

If you hover over the highlighted area, a link labeled Customize will appear. Within the screen, you can change the character string used in your link to better reflect your test variable (for example, bit.ly/tenamtest).

Step #3 – Deploy Competing Tweets

Once you have your tweet text and your test links ready, open a Twitter scheduling tool like Buffer and set up your split test tweets to go live at your specified times (as pictured below):

preloaded tweets

Split test tweets can be preloaded into tools like Buffer to go live at specified times.

Step #4 – Analyze Your Results

As your tweets go live, you’ll be able to monitor the number of clicks each Bit.ly URL receives within the service’s Stats dashboard area. Match up the total clicks each tweet receives with the time it went live and you’ll have a rough estimate of when your followers are most engaged with your brand’s updates.

stats dashboard

The Bit.ly Stats dashboard gives you click data for the URLs you created in Step #2.

Step #5 – Expand Your Testing

Of course, keep in mind that a single test – carried out on a single day – won’t give you as much usable information as systematic, ongoing testing. As an example, consider that running the above experiment could produce completely different results on a Sunday than it would on a Friday.

A few of the different ways you can expand your testing protocols to generate more meaningful data include:

  • Repeat tests across multiple days.
  • Testing variables other than tweet times in your updates (for example, you could use this same method to test different benefit propositions, product promotions or other calls to action).
  • Carry out tests on other social networks.

Split Testing on Other Social Media Websites

Now that you understand the basic mechanics of split testing on Twitter, you’ll probably be able to do it on other social media websites.

The possibilities truly are endless once you grasp the concept of social media update split testing. By taking the time to test your social profile updating assumptions – rather than relying on guesses or the advice of third-party experts – you’ll ensure that each new message you post to your social profiles results in the greatest possible impact on your company’s online performance.

What do you think? Have you done split testing? What results have you had? Please share your comments in the box below.

Tags: , , , , , , , , , , ,

About the Author, AJ Kumar

AJ Kumar is co-founder of Single Grain, a digital marketing agency based in San Francisco. Single Grain specializes in helping startups and larger companies with search engine optimization, ppc, and social media strategies. Other posts by »




More Info
  • http://www.eBizROI.com Rick Noel, eBiz ROI, Inc.

    Great article/top AJ. We (@eBizROI) have been using buffer as our primary Twitter scheduling app for about 6 months now. It is a great testing tool that it keeps a history of all Tweets scheduled through it, even the non-paid version (we are users, not otherwise affiliated w/buffer).

    In addition to day of week and time of day, we test variations of,hash tags (# and placement), links (shortened/not shortened), length of tweet, placement of links within tweets. Click troughs, favorites and retweets are things to look at. Believe it or not, we find that quality, trusted, recognized sources and consistent, high content relevancy to followers are the biggest triggers for improved performance.

    Thanks for sharing.

  • http://www.facebook.com/wanna.be.with.you.forever Devesh Verma

    Yes, I used split testing process in Facebook. It’s really a
    wonderful method to find out best posting time on social media website. One
    line you said in your post that is “The possibilities truly are endless once
    you grasp the concept of social media update split testing” Very true
    possibility in social media are endless….

    Be creative & win the market is a rule of social media
    platform.

    Keep Writing AJ Kumar ……& SME Team Thumbs Up!

  • Pingback: How to Split Test Your Twitter Marketing « Social Media / SEO / Mobile / Digital Marketing News

  • http://www.EntrepreneurOnFire.com/ John Lee Dumas

    Great post AJ. Split testing seemed an interesting way to test the effectivity of our Twitter marketing and the scheduling of our posts. What I want to know is if it will also work on Hootsuite since it is what we use for scheduling our posts.  And, do we still need to use bit.ly, since Hootsuite has its own URL shortening service. Thanks in advance!
    Have a great Thursday everyone!

  • Pingback: How to Split Test Your Twitter Marketing | Social Media Examiner | Complete Twitter

  • http://dougridley.net Doug Ridley

    AJ, 

    I’d suggest readers check out Google’s URL shortener as well. Google doesn’t have all the functions that Bitly has, but for A/B testing I find their service easier to use.

    Great steps and thanks for putting this together,
    Doug

  • http://www.convinceandconvert.com/ jaybaer

    It’s a good premise AJ, but the trouble is that you can’t step into the same river twice. The fact that a portion of your followers will have seen the first tweet will have an impact on the success of the second tweet, and that impact will grow on the third tweet. It’s helpful directional data, but it’s not a true (or truly objective) split test because you can’t segment recipients. I hope Twitter (and even more so, Facebook) roll this out for real someday. 

  • Pingback: How to Split Test Your Twitter Marketing | Social Media Examiner - Bonnie Sainsbury Views the World

  • Bruce Frazer

    Split testing on Twitter is always difficult I have tried different ways. I have 2 different Twitter accounts and I send out different msgs for the same Location.I agree with @jaybaer You must be able to segment your followers to be able to test with usable results. I find that sometimes by just repeating the tweet you will get clicks. @dougridley:disqus  I looked at some research that told me that more retweets are done with bit.ly shortened links than any other (Dan Zarella). Thanks AJ this is the best article on the subject that is readable.

  • http://www.marketerproductreviews.net/ Bruce Frazer

    Split testing on Twitter is always difficult I have tried different ways. I have 2 different Twitter accounts and I send out different msgs for the same Location.I agree with @jaybaer You must be able to segment your followers to be able to test with usable results. I find that sometimes by just repeating the tweet you will get clicks. @dougridley:disqus  I looked at some research that told me that more retweets are done with bit.ly shortened links than any other (Dan Zarella). Thanks AJ this is the best article on the subject that is readable and good for ongoing dialogue.

  • Pingback: facebook | Neil deGrasse Tyson On What Cities Of The Future Will Look Like + MORE 12/13/2012 | Free SEO Training

  • http://blog.cdginteractive.com CDGInteractive

    What a great idea! If you use Google Analytics, you could also add campaign tracking parameters to the test URL before shortening it in bit.ly, using the ?utm_content variable to distinguish between the two. That would allow you to also see if each segment behaves differently once it clicks through to your site.

  • http://twitter.com/ajkumar AJ Kumar

    Great ideas, Rick – thanks for sharing!  Obviously, I’m a big fan of the Buffer App myself, and I’m glad to hear it’s working for you as well.  Couldn’t agree with you more on the importance of trusted sources and content relevancy when it comes to social media performance.

  • http://twitter.com/ajkumar AJ Kumar

    Thanks Devesh!  Glad to hear you’re seeing good results using this split testing method on Facebook, and I hope you find the tips above helpful should you decide to move your optimization efforts over to Twitter as well.

  • http://blog.cdginteractive.com CDGInteractive

    You can shorten links on bitly and then add them to your scheduled posts in Hootsuite if you don’t want to use their shortening service, ow.ly.

  • http://twitter.com/ajkumar AJ Kumar

    It’s been awhile since I’ve used Hootsuite, but I don’t see why it wouldn’t work as well as long as you’re able to schedule set times for your tweets to go live.  Same thing for Bit.ly – as long as you’re able to track click-throughs on a per-link basis, you should be fine.

  • http://twitter.com/ajkumar AJ Kumar

    Thanks for the recommendation, Doug!

  • http://twitter.com/ajkumar AJ Kumar

    Obviously, this protocol isn’t going to have the statistical significance of an on-site A/B test that’s carried out correctly, but that doesn’t mean it can’t still generate some useful information when it comes to social media marketing.  Testing in this way can still generate plenty of interesting data on the phrasing that appeals most to your audience, the types of tweets they respond best to, your most effective calls to action and the times that your followers are most responsive – even if this information is more “guideline” than “conclusive result.”

    You’re right, though – I’d love to see Twitter and Facebook roll out this capability at some point in the future!

  • http://twitter.com/ajkumar AJ Kumar

    Thanks for reading, Bruce.  That’s interesting that you’ve tried testing using different Twitter accounts, but I’d be curious to hear how similar your follower profiles are for each username.  Also interesting on the frequency of Bit.ly re-tweets – could be another fun test to carry out!

  • Chris Picanzo

    Great content AJ, much appreciated I’ve wanted to try this myself and this gives me the knowledge and push I needed to give it a whirl. Now if I can just get people to click through we’ll be on to something..lol
    Thanks again!

  • Pingback: The Web was created to knock down walls. |

  • http://www.marketerproductreviews.net/ Bruce Frazer

    Here is a link for Dan Zarella’s (Hub Spot Manager) report on ReTweets, It is very intense reading. he studied over 40 million retweets to gather this info
    I Ifeel that Twitter is underestimated in it’s power as a marketing tool.People treat it too much like a “by line ” therefore the followers don’t expect much more.The key to unlocking the power of Twitter is too start viewing your Tweets as a “Door Knob” so when it is opened, a whole new experience is introduced to your Follower . (c.f. Dorothy opening the Door to the Technicolor land of Oz.), But that is another Journey to explore.

    I hope you enjoy Dan’s Data.
    http://www.marketerproductreviews.net/wp-content/uploads/2012/12/science-of-retweets.pdf

  • Kamilla

    Hello, yes split testing is really good for investigating website pages but really are you trying to convice us that sending the same three times a day every day over a period of a week is good (at least that’s how much you would need to to collect objective data or more). This is repetition, no value for followers that will just get annoyed by reading the same tweets ( imagine on Facebook – even worse). I don’t know how you are doing split testing but for it to make sense and have objective results you need to have enough data. With website testing around 1000 visits. This just cant be applied to social media properly as you risk too much repetition.

  • Pingback: How to Split Test Your Twitter Marketing « The Marketing Automation Alert

  • AA CASINO SOLUTIONS

    A great read and insight AJ and valuable comments too…  really enjoyed this thread.

  • http://www.ExpandTheBrand.net Tia Dobi

    Natch. Also…. and this may prove to be a precise hinderance to the content of this article… Twitter doesn’t always accept similar Tweets. I like that AJ has siphoned the test into ONE element: time of day. Yet, those varying bit.ly links may not post when loaded into auto-application programmes.

    (I’m open to feedback on that if your findings differ.If it’s an iffy thing… it may or may not post… please disclose that.)

    Here’s an Intelligent Creative* tip for time of day Twitter posting:

    Tweet in the middle of the night.

    Ever hear of infomercials?

    Peace and profits,

    Tia D.

    *Meaning it’s meant to put money directly into your pocket.

  • Pingback: “Split Testing Your Social Media Campaign” « The Profitable Business Entrepreneur

  • http://about.me/sookieshuen Sookie Shuen

    Thanks AJ for sharing this article. Do you have a spreadsheet that you can share with us to monitor the effectiveness of A/B tests and revise the tests on a continuous improvement basis?

  • https://www.servertastic.com/ Andy Gambles

    Why the need to create three different links? You can view your bitly stats by adding a + to the end of the URL. That will then tell you what time of day got the most clicks.

  • Pingback: Top Online Marketing News December 14, 2012

  • Pingback: Online Marketing News: Sins of Site Design, Google+ Communities, Twitter Photo Filter, Holiday Spending | Tips for the Unready

  • Pingback: Social Media For Business Guide

  • http://www.mktdojo.com/ Tommy Tan

    I’m setting up my split test at this moment! =] Thank you!

    Edit: I tried doing the split test but I think I found a hole in this strategy. If we use bit.ly to create the shorten links, bit.ly will provide the same shorten link if we enter the same destination URL to shorten link tool. For example, I want to shorten example.com, bitly will provide a shorten version, if i try to create another shorten link to example.com, bitly will give the same shorten version as the previous. Therefore, how are we able to continue to step 3 and 4?

    This might only be my problem or I missed something. Please advise.

  • Pingback: Online Marketing News: Sins of Site Design, Google+ Communities, Twitter Photo Filter, Holiday Spending

  • Pingback: Online Marketing News: Sins of Site Design, Google+ Communities, Twitter Photo Filter, Holiday Spending - mySMP Web Design Rochdale

  • Pingback: Online Marketing News: Sins of Site Design, Google+ Communities, Twitter Photo Filter, Holiday Spending - mySMP Web Design Rochdale

  • Pingback: Online Marketing News: Sins of Site Design, Google+ Communities, Twitter Photo Filter, Holiday Spending | My Blog

  • http://twitter.com/ajkumar AJ Kumar

    Very interesting resource – thanks for sharing!

  • http://twitter.com/ajkumar AJ Kumar

    Great tip – thanks for sharing ways to this experiment further.

  • http://twitter.com/ajkumar AJ Kumar

    Haha, yes – getting the click-through is the first step :)  Best of luck with your experiments.

  • http://twitter.com/ajkumar AJ Kumar

    As far as the repetition issue goes, it’s highly unlikely that your followers will be active at all three times you tweet goes live (or, even if they are, that they’d notice the repetitive tweet amidst the clutter of most peoples’ Twitter feeds).  To be clear, though, I wouldn’t recommend running the same tweet three times a day for a week – the same tweet repeated three times during a single day should be enough to generate some interesting information.

    Or, if you’re uncomfortable with the repetitive nature of the tweets, there are always other variables that can be tested using the method described above.And again, as I mentioned in a comment above, this isn’t about statistical significance.  Social media split testing isn’t ever going to be as rigorous as website split testing, but that doesn’t mean that you can’t still uncover some interesting insight into your audience’s behavior on these platforms.

    Hope this helps!

  • http://twitter.com/ajkumar AJ Kumar

    Glad you liked it.  Thanks for reading!

  • http://twitter.com/ajkumar AJ Kumar

    Sorry – I don’t have a template spreadsheet for that purpose, but it should be pretty easy to make your own using Excel or Google Drive.  Good luck with your experiments!

  • http://twitter.com/ajkumar AJ Kumar

    Great tip – thanks for sharing!

  • http://twitter.com/ajkumar AJ Kumar

    Hmmm – that’s how I’ve set up experiments in the past, but if it’s not working for you, you could always try the “adding a +” to the end of the URL trick mentioned by Andy Gambles in a comment above.  

    Hope that helps!

  • http://twitter.com/SMARTSocMed SMART Soc Med Mgmnt

    I’m using my social media management software to help me test times and verbiage so that I can maximize my ads…so far it’s interesting and I am going to have to use split testing on a couple of ads. It’s totally cool; thanks

  • http://www.web-media.co.uk/ Rob Willox | Inbound SEO

    Agree with jaybaer that testing as outlined is not 100% accurate but that said any attempt to better understand what motivates followers and engages with them and when is no bad thing. Admittedly, it would be better being able to post to segmented groups, but beggars can’t be choosers. 

    The criticism only applies to testing ‘best time to post’ type analysis and as Rick Noel points out there are other and additional aspects of a post that can be tested in addition to timing. And, there are other applications eg SocialBro which can give a level of ‘best timing’ based on the behaviour of your followers even in the free option (although limited to 100 users).

    Would also agree that using Google URl builder to initially generate page links would contribute to overcoming the issue outlined by Tommy Tan by providing different url structure/content and allow, as another contributor has commented, the ability to further analyse onsite visitor behaviour and activity on the path to your desired conversion.

  • http://www.bizsugar.com/ Heather Stone

    Hi A.J.,
    Great post. If nothing else, you should test the effectiveness of all your social media to be sure it’s doing what you need it to do. Thanks to D. Dixon for sharing this with the BizSugar community.

  • http://tiroberts.com/ Ti Roberts

    This is a great post. I had no idea you could split test your twitter marketing in this way. I appreciate D. Dixon for sharing this post with our BizSugar community.

    Ti

  • Pingback: Do Likes really matter? | Smooth Promotions

  • Pingback: How to Split Test Your Twitter Marketing «

  • Pingback: How to Split Test Your Twitter Marketing | Elite

  • Pingback: Make The Most Of Twitter By Split Testing | EnterpriseMessagingNews

  • Julie-Soleil

    You are right Tommy. I’m still looking for a way to create multiple shorten URL from one same long URL. Bit.la does not sema to allow this practice…

  • Mac

    AJ – great writeup. How about a different type of split testing, whereby you want one shortened URL to rotate through multiple variations of URL’s. For instance if you have two different versions of a landing page, each residing on a unique URL. Do you have any recommendations on how to leverage a URL shortening service that will rotate clickthroughs to each unique URL variation?

  • Mac

    AJ – great writeup. How about a different type of split testing, whereby you want one shortened URL to rotate through multiple variations of URL’s. For instance if you have two different versions of a landing page, each residing on a unique URL. Do you have any recommendations on how to leverage a URL shortening service that will rotate clickthroughs to each unique URL variation?







Check out the Social Media Marketing Podcast!
Join our Social Media Marketing Networking Club
Download the free Social Media Marketing Industry Report
Get Your FREE Copy of the 2014 Social Media Marketing Industry Report
Wondering how your peers are using social media? Get this free report (50 pages, 80+ charts) and never miss another great article from Social Media Examiner.