Are you stuck in your A/B testing tool when analyzing the results? You are missing out a lot of great insights if you keep your nose in just one tool.
Over the last years I have seen many companies run A/B tests via Optimizely, VWO, Test & Target or even directly in GTM.
One crucial thing that I have learned: integrate your testing tool with a web analytics tool.
Last week I read this article on the Convert blog.
Will Google Optimize be the future of A/B testing? We will see.
In this post I reveal five steps in Google Analytics that are a tremendous help in analyzing and optimizing on the outcome of your A/B tests.
For the purpose of the article, I use the integration of Google Tag Manager and Google Analytics as an example.
Step 0: Prepare Your Test
Step zero: prepare your test set up first.
At least take into account the following points:
- Find conversion-focused pages for your next A/B test.
- Create a smart hypothesis based on data and psychology.
- Determine the sample size and find out how long the test needs to run.
- Discuss about power (false negative) and significance level (false positive).
- Design one or more test variations.
- Code your test variations.
- Test your set up.
Step 1: Capture A/B Test Data with GTM
As a first step you need to make sure to integrate your testing tool and/or Google Tag Manager with Google Analytics.
A very efficient way is to use the event tracking feature in Google Analytics.
Last year I was at the conference Conversion Hotel.
Analytics & Optimization Expert Jules gave a great presentation on how to get this to work.
Check it out below:
Do you use Optimizely for running your A/B tests? Read this guide on integating Optimizely with Google Analytics via Custom Dimensions.
VWO has a similar guide on this topic you could check out as well.
Step 2: Check Your Real-Time Reports
You can see within minutes after you start your A/B test whether things work smoothly or not. Head over to real-time reports and take a look under events:
It depends on how you have named your variations, but this is where you should quickly see data showing up.
Here is a suggestion on how to name your events:
- Event category = AB-Test
- Event action = [name of variation]
- Event label = either 0 (default) or 1 (variation)
Make sure to use naming conventions when you set this up.
Step 3: Build Powerful Custom Reports
On default, events are shown within the event report. An example is given below:
The problem here is that you are not evaluating your A/B test on the user level. You are looking at “plain” events.
Building a custom report is a great solution here.
A few things to note here:
- Use goals instead of transactions if you are optimizing on a non-ecommerce site.
- In case of a non-ecommerce site forget about revenue.
- Add a title that suits your test.
- Changing the report table is optional.
- Add a filter on the event action that is connected to your test.
- Decide on adding this custom report to one or more reporting views.
An actual report based on users and transactions looks like this:
You are not there yet.
If you want to evaluate on number of buyers instead of transactions, you need to transform the transactions into a user based metric.
Add one extra segment to accomplish this: sessions with transactions.
Note: read this in-depth post on segmentation if you are not familiar with segments.
The report looks like this after you add the segment sessions with transactions:
- Default: 36.349 users and 483 transactions.
- Variation B: 35.613 users and 475 transactions.
Statistical evaluation of a test is something for a future post, but this test looks inconclusive! :-)
Step 4: Dig Deeper via Segments
It depends on the actual test and context on which segments you want to drill deeper.
Here is an example of further segmenting on device category:
Now it’s possible to judge your test performance on multiple devices.
Note: don’t draw conclusions on very low numbers. Keep 300 to 500 transactions per variation per segment as a minimum.
Other segments/dimensions that are interesting to analyze:
- Traffic source
- Landing page
- Type of user (new or returning)
- Recency (days since last visit)
- Region (cultural differences)
Tip: use sequential segments to analyze navigational behavior differences.
Step 5: Leverage Shortcuts
Let’s assume you have created a few custom reports and applied segmentation.
What if you want to review these reports (including applied segments) at a later stage?
Shortcuts are the solution here.
It is clearly explained how to set this up on the support pages of Google:
You can find your shortcuts in the reporting interface:
Well, I hope you are inspired to analyze your next A/B tests at a deeper level after reading this post!
Do you already integrate your A/B tests with a web analytics tool and how? Happy to hear your thoughts!One last thing... Make sure to get my extensive checklist for your Google Analytics setup. It contains 50+ crucial things to take into account when setting up Google Analytics.