HOW TO: Use VWO or Optimizely instead of Unbounce's A/B Test


#1

Unbounce’s A/B Test reporting left me desiring more. Specifically:

  • Track Multiple Goals
  • Segment Report (i.e. desktop vs mobile visitors)
    I still want to build the variation in Unbounce, rather than using VWO’s editor. This is so I don’t have to rebuild the variation in Unbounce in the event the test is successful.

Step 1: Make your variation in Unbounce.
You’ll want the “Weight” to be 0% on the new variant, since we aren’t using Unbounces A/B test. Unfortunately, the variant won’t publish unless it has “Weight”. To get around this, temporarily set the weight of the new variant to 1%, publish the page, then revert it back to 0% and publish once more.

Step 2: Install VWO’s code
You can use the Global Scripts to make it easier, or just put the code on the page scripts. Make sure it is on all variations and confirmation pages.


Step 3: Configure Split URL
Create a “Split URL” campaign in VWO. For the control, set it to the Unbounce page URL. For Variation 1, use the page URL + the variation letter + .html ( www.example.com/page-name/b.html ).

When you launch your campaign, VWO will randomly redirect half of the traffic to Variation 1. This is why we set the Unbounce variation weight to 0%. VWO will handle splitting the traffic.

Step 4: Set-Up Your Goals
I always add as many goals as possible. It’s always best to have too much data, than not enough.

Below is a screenshot of one of my live tests. I’m testing the addition of testimonial videos at the bottom of my page. I want to track the following goals:

  1. How many people scroll to see the video thumbnails?
  2. How many people click to view the video?
  3. How many people click the Call to Action?
  4. How many people fill out the form on next page after clicking the CTA?
    Using Unbounce’s A/B test, I’ve had to pick one goal. It’s usually Goal 3, but ultimately, I want more of Goal 4. If I measure Goal 4, the test takes longer to reach statistical significance. If I choose Goal 3, there is a risk that while the CTA is getting more clicks, they may not be qualified clicks resulting in more form submissions. With multiple Goal tracking, I don’t have this problem.

Step 5: Segment Results
With VWO, you can segment your results to see if traffic source or device affects your test. Maybe your test is great for Desktop, but isn’t so awesome with Mobile. You’d never know that with Unbounce’s A/B tests.


#2

This is why I like to share my work. It get’s the ideas flowing! Tell me more.

I don’t have variation data in GA. I set it up with Google Tag Manager before Unbounce released the GA integration. I haven’t switched because of the custom configurations in my GA code (such as the Display Advertising Features).

Are you using Unbounce’s GA integration? Have you figured out a way to use custom configurations with it?


#3

Found my answer. The Display Advertising Features (Remarketing) can now be configured directly in Google Analytics. I no longer need the special tracking code.

Just switched to Unbounce’s GA integration. Excited to see how this improves my analysis!


#4

Hi Phillip - sorry for my late reply! I’m in Frankfurt speaking at Conversion Summit 2015. 

I can write a post in the community on this. But it sounds like you’ve figured it out. Let me know if you have more questions :) 

  • Michael 

#5

Guys - this thread is awesome. I’ve been tied up with work over the past few weeks and am just catching up on Unbounce posts - and this is just totally awesome. Thanks Philip and Aagaard !

I too find that GA is the best tool. I’ve been using it for years, but only got my certification early this year, and it’s opened of loads of new doors for analysis.

I’m intrigued about how you’d track Goal 1 using GA though - I’m still getting my head around the more techy side of GA - but can you track a scroll down as an event ?

Cheers

Amit


#6

Hi Phillip,

I always use GA to track test results - regardless of which testing platform I use.

I don’t trust data from tools like VWO (compare with GA numbers and you’ll see why). Moreover, you can do much more advanced segmentation, etc. with GA than VWO or any other testing tool for that matter. 

I assume you already have your GA script on your Unbounce landing page, so all you have to do is set an event to fire in each variation. Then simply build an advanced segment for each variant and go nuts with your segmentation. 

I love the fact that you are aware of the importance of digging into your test results to get the full picture! However, if you only look at data from e.g. VWO, you’re in all likelihood going to get bad data - which then defeats the entire purpose of testing. 

My philosophy: use a testing tool for running the test and an analytics tool for analyzing the test data : ) 

  • Michael