Website testing is critical to conversion optimizers.
Testing helps you identify the elements of your website and other marketing collaterals that are working and the ones you can improve on.
But it’s important to do testing right if you want to get insights from your testing activities that are meaningful to your organization.
If you’re new to split testing, keep these best practices in mind so you know you’re on the right track when optimizing your website for increased conversions.
1. Assess your testing readiness
Before you start any test, you should be able to clearly answer if you should be testing in the first place.
First, you can’t test if you don’t have the traffic. You need to have enough, stable, recurring controllable traffic to get valid results from your tests. A rule of thumb is for the page to have at least 10 conversions per day.
Then there’s the technology factor: can your CMS support testing? What tools do you already have to run tests? The technology aspect is important as you don’t want to risk inadvertently breaking the user experience while implementing your tests.
Which leads us to the most important consideration prior to testing: are you sure your website doesn’t have serious usability and user experience issues to begin with? Because those problems cannot be solved by split tests. As Conversion Conference Tim Ash states, “Testing inside of a pile of garbage is not a good solution.” Fixing the obvious usability and UX problems on your site that hinder visitors from converting is a better use of your time and resources at this point.
Testing is most useful when you’re at the stage of refining your website for more persuasive elements as well as usability and UX tweaks. Don’t use a/b testing as a crutch for the glaring usability and user experience issues on your site.
2. Clearly define your hypothesis
Testing is pointless if you’re doing it on a whim.
The idea behind testing itself is that you want to have concrete proof that something will work better on your site. That means that before you test, you should already have a theory about what’s wrong with your site and how it can be addressed by testing.
Testing is no other than scientific research carried out on your website. In fact, split tests utilize the experimental design where the original iteration serves as the control (A) and the element that’s changed for the study is the variable (B).
You shouldn’t take this to mean that you can’t have testing ideas arising from your or your team’s experience. You just need to make sure that these ideas are backed up by data from your business, and that testing your hypothesis will yield useful information about your customers, product, and/or market.
3. Learn what statistical significance means in testing
If you didn’t take up inferential statistics, now is the time to do so.
In website testing, it isn’t enough to look only at the descriptive statistics. You need to make sure that the results you’re getting aren’t due to random chance. Imagine you’re running a test where you changed a video with an image. When the results come in, you see that more visitors converted better on the new version. Although you’re excited with the results, you still need to check for statistical significance to confirm that the difference in conversion rates is real and didn’t happen by chance.
Also, don’t confuse the word “significant” to mean that your findings are “really important.” When your findings are statistically significant, it only means that confidence levels are high that the effect or relationship exists.
4. Start with simple tests
Optimization is exciting, and most marketers can’t help but get carried away.
They learn about the benefits of website testing and they’re already thinking of some grand tests on their web pages.
But before you start planning sophisticated experiments, you should know that like everything else, testing has a learning curve. It would be better to start with simple tests to familiarize yourself with the process. For instance, you can try changing the text on your call-to-action towards a more emotional variation to see if your visitors respond better to them. Or swap one image for another to learn if doing so would make more people convert.
There are lots of simple tests you can do that can have a big impact on your website conversion rate. Reserve those more ambitious tests for when you and your team are ready for them.
5. Segment before you test
Digital marketers often make the mistake of running their test on all of their web visitors.
But your visitors are not homogenous. They have important differences in terms of roles, culture, browser, and devices used. They also differ in terms of demographics, psychographics, and location, all of which influence how they respond to your website and marketing as well as their expectations and motivations on your web pages.
It is therefore best to target your test to a relevant audience. This way, you’ll gain a better understanding of visitor behavior based on where they’re coming from or according to their specific characteristics. For instance, you can target users who access your site on mobile to see how they respond to your changes. Later on you can do the same test on desktop website visitors. You can even compare if there are any differences in their behaviors.
6. Beware of testing biases
In website testing, as in all other research, you want nothing else than to arrive at valid and reliable results. Otherwise, all of your efforts will have been for nothing.
Validity refers to the soundness of your research design and methods. It’s a matter of how well your chosen variables, instruments, and procedures measure what you’re trying to measure in the study. Reliability, on the other hand, is concerned with the replicability of your research process. Will repeating the same study under the same conditions yield the same results?
It is your job as a researcher to control all threats to validity and reliability. These threats, or biases (since they can skew the results from your testing), can be internal or external. Some examples of internal threats are flawed sampling (often due to poor randomization), visitor bias (when a large portion of your sample are people who are familiar with your site), and shifts in your traffic mix (which commonly happens because of sequential testing).
External bias occurs when you test while there are seasonal traffic spikes, like holidays or when your testing coincides with the launching of promos. It can also result from malfunctions in your testing technology or tool.
These biases can affect the validity of your data set and skew your results. In turn, validity problems may undermine the reliability of your test. For example, if your test is affected by flawed sampling, your results will be unreliable and inconsistent with successive testing. It pays to be aware of these biases so you don’t waste your time and resources on flawed tests that could have dire effects on your website and other marketing activities.
7. Lack of success in testing is not failure
What if you keep getting statistically insignificant results?
What a lot of testing vendors and consultants won’t tell you is that a lot of your tests will fail. But this shouldn’t be a reason for you to give up.
Treat your failures as an important part of the learning process. Try to find out if you’ve been doing things wrong especially with your data collection and testing methodology. But when you’ve done enough tests, you’ll also discover that even failed tests reveal something about your customers and visitors.
Failed or not, you can use insights from your website testing to achieve your goal: improving usability and user experience so you can improve your conversion rate.
Get more tips for website testing success! Learn directly from the experts and exchange ideas with the brightest minds in conversion optimization by attending Conversion Conference Las Vegas 2017, April 19-20. The last two conferences sold out so don’t delay, register now!