Set a clear goal by deciding which conversion to improve

The first step is to decide which conversion you are going to focus on in your test. Do you want to increase newsletter signups, consultations, checkout completions or just clicks on a key CTA? Your decision will be based very much on where in your funnel you feel there is a weakness. You can determine this by looking at conversions and events you’ve got set up in analytics to get an idea of where visitors are dropping off.

Create your hypotheses

What is a hypothesis?

A hypothesis is essentially a statement that you make about a belief you have. An example would be

 “If the colour of the main CTA on x page is changed to orange the CTA will get more clicks”. 

The statement needs to be detailed, specific and actionable. The hypotheses should also relate to your conversion that you want to increase. In the above example, the conversion you might want to increase the number of clicks on CTA. Also note that it would probably not be ideal to use the above hypotheses for a goal like increased purchases of a product as this goal is too far removed from the CTA. In other words, there could be seven more steps from this CTA to get to the checkout point. Along this journey there could be a multitude of points where people can drop off and if this conversion doesn’t show improvement you might never know that clicks did actually increase on the CTA even though the conversion did not.

For a detailed guide on how to write a hypothesis, read this article

How many hypotheses should you create?

In a nutshell, as many as you can think of. Some things to keep in mind though: They should all relate to the conversion you are aiming to increase, you’re not going to test all at once and you are not going to test all of them. More on this below.

Prioritise your tests (hypotheses)

Now that you have your lists of hypotheses that you want to test you need to prioritise them. As the main metric for everyone is ROI you need to keep the same mindset. So you need to ask yourself: “Which change will have the biggest positive effect on revenue whilst involving the lowest investment”. It comes down to a balancing act between cost versus impact. The test that could have the biggest positive effect on conversion rate versus the lowest development cost should be at the top of your list and each subsequent one will rank lower according to this scale.

This article by VWO has some excellent guidelines and references to tools you can use to prioritise your tests and a lot of other information as well.

Create two versions of your website / webpage

Below are some tools you can use to make changes, most of themdon’t require any intense coding knowledge but most will require some HTML and CSS knowledge to some degree:

Google Optimize

Visual Website Optimizer

Optimizely

Unbounce

How many changes should you make?

In a very small nutshell, ONE change per test. Remember what we said about making your hypothesis very specific and choosing one at a time (based on priority) to test.

How to set up a test in Google Optimize

Google Optimize is a free tool by Google. There is a paid version as well which, amongst other things,allows you:

• to set up to 10 experiment objectives as opposed to three in the free version.
• You can run more than 100 simultaneous experiments as opposed to the 5 in the free version
• With multivariate testing you can have up to 36 combinations whilst in the free version you can have 16

For a more detailed list of differences have a look at this article. 

We’ll now explain step by step how to set up an A/B test with this tool.

The very first step you are going to take is to create an account by clicking on the blue “Create account” button on the top right:

Note that the Google Optimize account you use needs to be in the same account as where your Google Analytics properties exist.

Next you will give your account a name

Once you’ve accepted all the terms and conditions thingies you can click “next”

Now you have to “Add a container, see below

You can now create your first experience but we recommend that you link your account to Google Analytics first:

Once you clicked on “Link a property”, you will see this:

Once you selected a property from the dropdown another list will appear that shows all the views associated with the property, select the one you want to use and click “Link” in the top right

Now you will see this: 

Click “Get snippet”, it’s just better to deal with this now

Once clicked the snippet you need to add at the top of the <head> section of your page. TIP: Add this to all pages to save you from bugging your developer again.

Now you can go ahead and click that beautiful CTA that says “Let’s go” to create your first experience.

Now you will see this and now you can give your experience a name and select which type of test you want to do. As this is a post about A/B testing we will select that one in this example. VERY IMPORTANT, it is here that you also select the page that you want to run your test on so add that url in here:

Next up you can start creating your B variant by clicking, you guessed it, “Add variant”. A dialog box will pop up where you can give your variant a name. We suggest making it something descriptive and indicative of what you are testing, for example, “Orange Button”

Now you will see this:

And now the real fun starts! Click on “Edit” to start causing mayhem … erm … I mean make the changes dictated by your hypotheses.

If you hover around a bit on what you are now seeing you will see that blue blocks appear around most elements now. Once you click on an element you will see that the “edit element” button on the right will change to blue and is now clickable.

I selected the CTA because I want to make it orange (as stipulated by my CTA) In the box on the left I scroll down until I see the background colour icon for the selected element (the button) and I change it to “orange” (You can also use hex and rgb codes here to make fancier oranges)

And voila! The button is orange!

Now you can click “save” in the top right corner and “Finished” to set up the rest of your test.

What you can do now is to set the targeting of your “experience” (test), in other words, who is going to see your test (one of the two versions). Click on “Customize” next to “Audience targeting” to set this. Check out all these sexy options:

Note that you can select and customise one of the above or just leave it on the default which means that everyone that visit your site will either see the A (original) or B (variant) version of your experience (test)

The next very important step is to set up your objective (conversion or event that you are testing against)

You will see a dropdown appear, one option allows you to test against a goal or event you’ve set up in Analytics already or you can create a custom objective which can be an event you set up in Google Optimize.

Now all that is left to do it start your test which you can do by clicking “Start” in the top left of the page:

The awesome thing about using Google Optimize is that it will tell you when you should end your test to get statistically significant results but you can also end it earlier if necessary, although this is not recommended.

And there you have it, now go forth and create amazing A/B tests and make your bosses or clients lots of money!

Tags: AB testing, Google Optimize, UX