Get your Google API credentials
Google application programming interface (API) allows communication between your site and Google, and you have to register your project (your project in this case is your site) with Google in order to retrieve your credentials. These credentials allow your site to authenticate to Google’s servers but won’t allow access to GA accounts. Giving uSplit permission to manage your Google Analytics experiments comes later (with just a click of a button, by the way!). Here’s how to get your credentials.
You should end up with a Client I.D and a Client Secret. If you have them the worst part is over! A developer will complete the configuration by adding these, together with the Google Analytics account details to the website configuration file. All set, start your site and check the new A/B testing node!
Before you can run your first experiment though, you need to give uSplit permission to access your Google Analytics data. Someone with access to the account needs to do that. This can be done by navigating to the new uSplit dashboard in the Content section and hitting the Authenticate button. After you do this it can be done by all other users, even if they don’t have to the Google Analytics account themselves. Now that you’ve got uSplit installed and GA authenticated, you’re ready to set up your first experiment.
Creating your experiment
- Navigate to the Content section and right click on the A/B testing node
- Select Create new experiment and choose which page you want to test
Before your experiment is officially off the ground, you need to create and set up variations. Creating a variation is a matter of either cloning the page or by adding an existing page as a variation. When you click Clone original, you’ll be prompted to tell the software the location where the copy should be saved.
Note: when you clone a page, you will be prompted to tell uSplit where you want the page to be saved. We recommend creating a separate “variations” folder to store all of them in one place. This keeps them out of your menus and away from your master content, and it also does away with certain SEO issues (like prohibiting crawlers to access them and mistaking them for actual pages)
Editing your variation
If you’ve read previous entries (which we totally recommend doing), then you know the good thing about uSplit is that variations are just ordinary pages. That means you already know how to edit them. When a variation is selected to be rendered, uSplit overrides the original’s properties with those of the variation (leaving a property empty falls back to the original by default). Since the values of the original page are used, you don’t need to duplicate all of the descendents (which solves a lot of issues).
You can also change the template by assigning a different template to the variation.
Let the experiment begin
Starting your experiment only requires you to click the Commence experiment button, and you’re ready to go. Last thing to consider before starting is the metric. Pageviews is used by default (but you can specify a custom one in GA Experiments). This metric will be the barometer for testing the performance of the changes that you’ve made (I.E, which variation results in an increase in pageviews).
Initially, each variation (including the original, which is itself a variation!) are subject to the same amount of traffic. As the data rolls in, variations will be slowly phased out as the software reaches certainty that one is outperforming another (valuable in that it funnels your existing traffic toward the variation most likely to result in success, rather than wasting valuable time and perhaps making you miss conversions). Eventually, some variations will be deactivated completely.
As you can see, aside from the initial stages where a developer will be needed to help with the technical aspects of setup, it then only needs a few clicks to set up an experiment with uSplit and get it rolling. With the most of the processes automated, you won’t have to waste time on the little nuances that make testing difficult.
That being said, there are still plenty of companies whose needs might warrant extra help. If A/B testing is important to your business and you’d like to maximize your returns, or you are in need of help immediately, Happy Porch can help.