Descriptive Analytics vs Diagnostic Analytics

Recently, David Attard wrote about analytics and KPIs (key performance indicators), and how they can be used to understand our website users better — and, in turn, to help us design better experiences for those users. He told us about the important metrics to analyze (time on site, bounce rate, conversions, exit rates, etc.), but also mentioned that, while these metrics help us to understand what users are doing (or not doing) on our website, the reasons why can still be a bit of a blur. This is because, while data is objective, the conclusions drawn from it are often subjective.

Even though KPIs describe our users’ behavior, more context is needed to draw solid conclusions about the state of our UX. In order for this to happen, we have to use other techniques — such as A/B testing and usability testing — to diagnose the UX flaws we identify through descriptive analytics. In this article, I’m going to explain the difference between descriptive analytics and diagnostic analytics, so that you have a realistic expectation of what descriptive analytics can do, and what you’ll need to gain from descriptive analytics before you begin A/B testing and usability testing.

Descriptive Analytics

Descriptive analytics in a nutshell: what has happened?

--ADVERTISEMENT--

When you visit a nurse or doctor, it’s because you have undesirable symptoms that indicate bad health. You have trouble doing the things you need to do because of this. You don’t know what’s going on exactly, only that you aren’t functioning at an optimal level. This is likened to analytics, where business goals can’t be met because of bad user experience. Certain KPIs might indicate this, such as high bounce rate or low Avg. Time on Site.

KPIs describe the symptoms, but they don’t actually diagnose what the underlying issue is, and this is why we call them descriptive analytics. However, we can use the symptoms to help diagnose the UX flaws.

We can use tools like Kissmetrics to track and analyze KPIs, although many companies choose to use Google Analytics because it’s rather sophisticated for a free tool. As well as the KPIs mentioned in David’s article, analytics tools like Google Analytics can reliably tell us things about our users’ demographic and interests (that is, who they are and what they like), and also other important tidbits of information such as what device they’re using and where they’re from. This kind of data, even though it can’t be used to indicate website performance, can tell us a little more about the user intent.

Consider these descriptive analytics as background information that we can use to narrow down what’s going wrong exactly (i.e. user research).

Diagnostic Analytics

Diagnostic analytics in a nutshell: what can we do to fix it?

Let’s assume that your descriptive analytics indicate low sales, even though your website is receiving traffic. After setting up some Event Actions/Goals, you can see that users are adding items to the cart, but they’re not actually checking out. The data indicates that Exit Rates are high on the web page where users are expected to input their credit card information.

You’ve determined that low sales are likely due to a flaw in the user experience of this screen, but what is it? Here are some ideas:

  • the form doesn’t work
  • the form is too long
  • the page doesn’t look trustworthy
  • there’s a lack of customizable options
  • an unexpected charge appeared.

Now, unless you’ve made a super obvious mistake (such as forgetting to serve the website over secure HTTPS), narrowing down the UX flaw(s) could to be difficult using only descriptive analytics. We have two options that can help to diagnose the issue(s): A/B testing and usability testing.

A/B testing can help you to implement a viable solution alongside the original implementation, to see which converts better. There’s also multivariate testing that can help you test more than one variation, but if you’re still relatively clueless as to where the UX is falling short, you could end up designing multiple variations and wasting time unnecessarily.

And this is where usability testing comes into the picture. Usability testing is about watching users use your website, to see where they struggle. While some flaws are hard to discover even through usability testing (since you can’t read the users’ minds), obvious flaws like form abandonment as a result of lengthy forms/broken functionality might become more apparent. At the very least, usability testing narrows down the issues, making A/B testing easier.

A/B testing tools such as Optimizely can help you run complex A/B tests, but Google Optimize (which is free and integrates directly with Google Analytics) is a decent free option. Tools like Hotjar and Fullstory can help with usability testing (feedback, surveys and heatmaps), whereas a tool like CrazyEgg combines both A/B testing and heatmaps into a single tool.

Bonus: Predictive Analytics

Predictive analytics in a nutshell: what might happen?

Predictive analytics is about analyzing what the user has done previously, in order to make informed decisions about what they’ll want next (or next time they visit). Consider it a subset of descriptive analytics that specifically focuses on customer journeys and personalized content, helping you to gain insights into how users convert or what content they’re interested in.

Predictive analytics sometimes uses machine learning as a way to deliver relevant, targeted content using data that your apps and websites have deciphered all by themselves. Since machine learning is automated, it’s recommended that you have large data sets to work with beforehand.

Fun fact: Amazon’s recommendations engine (“Customers who bought this item also bought”) is responsible for over 35% of their overall sales!

Conclusion

In short, descriptive analytics are about listening to the symptoms, and diagnostic analytics are about finding a solution. You can use what you now know about diagnostic analytics to ensure that you’re going about descriptive analytics and Google Analytics in the right way, since descriptive analytics are needed to inform your approach to A/B testing and usability testing later on.

In a future article, we’ll introduce you to Google Analytics and talk more about KPIs. You can find all the articles in this UX Analytics series here.

To learn in-depth about UX Analytics, check out SitePoint’s book Researching UX: Analytics.