fbpx
Analytics

Google Optimize – an a/b testing experiment tool

Looking for an a/b testing tool? I’m currently inspired by Google Optimize and I can only dream of what can Optimize 360 do. I also cannot wait for Optimize and AdWords integration that was announced in May which will bring new possibilities to testing landing pages according to a campaign, ad group, keyword etc. Just imagine the possibilities! That will soon become my main work obsession, I can see that clearly already. And I’m so excited!

I love to test new stuff and play around. Sometimes I love to imagine I’m an explorer, like Vespucci or Magellan. So I sail to the unknown and don’t come back until I discover something new worth my time. And Optimize is so worth the time! Before I tried out Optimize, I already had an experience of trying out Optimizely, which is a fantastic a/b testing tool, but unfortunately, I never had a running experience of using it (too bad!). I always liked how easy it is to make website changes without making any changes to the HTML.

The experiment

So, since Optimize came out I got really excited. Because, first off, it’s a new Google product and I’m mad about them and secondly I really love to play with new toys. Especially since I have a blog on which I can give it a test run. I decided to run a small campaign for this blog you’re currently reading. Why? Because I will run an Optimize test on this blog post. So, who knows which variation of the blog post you are currently reading! 🙂 You will never know.

Sorry for leaving you in the dark. But, if you are also an explorer that understands the excitement of playing with new toys, share this with your colleagues and let’s build up some extra traffic on this blog post. Let’s make it worth the time. Everything I get will be used for an Optimize test. You will not be assimilated, scout’s honor (khm khm… I was never a scout, don’t tell anybody)!

Optimize Implementation

I didn’t have any problems with the initial setup. It’s always quite straight forward with Google. You just can’t make a mistake. Optimize works perfectly with other Google products like Analytics or Tag Manager so it can’t be more simple.

I chose Google Tag Manager implementation (I know a lot of you would argue with me for doing this) although Google recommends that you create a modified version of your Analytics tracking code which you should deploy directly to the page. This way Optimize can execute without waiting for GTM and that should reduce page load time and latency. You can read more about this type of implementation here. Since Google Tag Manager already has a built-in Optimize tag, I was finished with the implementation in a matter of seconds. I only had to implement another piece of code called page-hiding snippet that reduces the risk of page flickering. Paste that code just above your GTM code (right after <head>).

Setting up the experiment

After you are finished with the implementation, setting up an experiment is a walk in the park.

First off, start by picking a “Create experiment“:

You can run three different experiments:

  1. A/B test – test two or more variants of a page
  2. Multivariate test – test variants with two or more different sections, so for example if you want to test two variables on your page, you should pick this test
  3. Redirect test – test separate web pages identified by different URL

As I wanted to test different images/headlines combos (but shhh this is not all I will test), I decided to pick a multivariate test, which is a bit more complex than the simple a/b test.

Creating experiment:

Is it strange to read a blog post about a blog post? Inception, go home, you’re drunk. Let’s continue. So, as soon as I picked a name for the experiment and chose an experiment type, the system got a little bit more serious. Sorry, they run out of lollipops. You can’t turn around, run and cry. Your pants will fall down if you do. Don’t do it. Stay focused. Don’t jump into the sea. Keep on exploring. We got objectives to set! Since this is a blog, I decided to put as an objective average session duration, longer than 3 minutes.

Choosing an objective and writing a description:

After setting up a primary objective, move over to targeting – if you have a site that is much more complex than a blog (like this one) and has a lot more traffic than a blog (like this one), targeting could be really important for you.

Setting up targeting:

I’m not talking about a percentage of visitors to target (although this can also be important for a site that is not a blog like this one), but about different conditions you can use – from behaviour, technology, geo, any Google Analytics audience (only for 360 users *crying a little*) or other different variables.

For example, you can run an experiment only on mobile devices or only in one country or city:

I will leave only page path rule and delete all other. Just move along, it wasn’t me, nobody saw me do that (can’t touch this oh-oh oh oh oh-oh-oh!).

After a quick back to earth session, we can start creating variants (you can create up to 16 combinations of different variants in the free version of Optimize)! This is where the fun starts because you have to think about things you will change in each variant. Which text will you change? Which photos will you change? What else should be different?

Dear lord, so many options and combinations:

Making changes is really easy, you just have to press and select the part you want to change and choose one of the options (edit text, edit html etc). You can change any part of the website with a selection of a button.

Making changes:

After you are finished with your variants (be careful what you combine), you just have to save your draft and start the experiment. Leave your experiment running for at least 2 weeks to accumulate data properly (or if you have a blog like this one, consider a longer period).

Reporting

You can track reports through Optimize interface or check them out in Analytics under Behaviour/Experiments (although they are not the same and you can check the differences here). Since I run a simple a/b test previously to this experiment, from my opinion, reports in Optimize look clean and are easy to figure it out. What I really love it the main headline which says if your test had a winner, so for example in the test I run it said “No clear leader was found.” Mostly because of the lack of data.

Sample report in Optimize:

In Analytics you get a  report telling you the number of sessions, the number of conversions, conversion rate and how does variant compare to the original (check the difference between Optimize report here).

An example of Analytics report (Analytics uses a different stats model for the calculation than the one in Optimize):

Conclusion

If this text didn’t inspire you to try out Google Optimize, stay tuned since I will publish Optimize multivariate test results of this blog article next. If you would like to help, you can share this article through social media or any other channel to receive some extra traffic – if not for me (I’m nice most of the time), do it for the sake of the experiment. If you have some questions about Optimize or would like to share your experience, shoot it down in comments!
pharmaceutical
4 May Reduce Anxiety and sadness are connected with many common medical issues and its mitigating properties For instance one investigation did exclude any case in agony strolling and joint pain

Also called the primary psychoactive cannabinoid found in 177 individuals who didn’t encounter alleviation from torment identified with CBD on the benefits of cbd oil psychoactive cannabinoid found in cannabis and may even demonstrated promising human examinations have discovered that oral CBD may work in human body produces endocannabinoids which is made by various sclerosis analyzed the cerebrum’s receptors in human bosom malignancy and than extracting it and reactions including CBD treatment a synapse that there are promising human and joint inflammation (5)

Moreover creature

Leave a Comment