How Data Visualization helps you understand your UX research results
How Data Visualization helps you understand your UX research results
Intimidated by user research analysis? Visualize your data first
I used to have to psyche myself up to analyze my user research.
I enjoyed talking with and interviewing users and learning how they use the website. However, the thought of staring at a spreadsheet of data after a day of interviews used to fill me with dread.
Whether it’s user interview/testing responses, survey results, or looking at analytics, there are times when UX Designers are faced with a bunch of data.
However, one of the tricks I’ve learned from the Data-Informed UX Design process is that there is a shortcut you can use around specific Data: Data Visualization.
To explain, let’s talk about what Data visualization is.
Data Visualization as a User Research tool (and design pattern)
I’ve previously discussed Data Visualization as a design pattern. However, to summarize it quickly, the primary use case of Data Visualization is to present complex information) visually to make it easier for the user to understand and interpret.
That’s the case if you have to design something complex (like a status-tracking website of dozens of products). However, doesn’t that also sound like what your user research might be: a giant spreadsheet of user responses that we need to understand and interpret?
If that’s the case, Data Visualization might be a technique you can use to make sense of data. Data Visualization often relies on many design principles to provide accurate, helpful, and scalable data. It’s part of many design systems, such as Google’s Material Design System.
However, how you use it for your user research processes will be slightly different than how you might use it in your designs: it’s there to make your data easier to read, not statistically significant or beautiful.
So here’s how you do that.
The purpose of Data Visualization with User Research
While you could spend time creating beautiful and pixel-perfect Data Visualizations, the primary purpose of visualizing user research is to make sense of otherwise complicated or complex data.
Therefore, you don’t need to go too fancy with how you present data (although you could if you want to). This starts with understanding the three things you’ll use Data Visualization for:
First, does the Data Visualization make sense?
Do basic calculations make sense with the data?
Are there any interesting patterns in the data?
These research questions ensure you can skip the tedious parts of analysis (cleaning and maintaining the data) and move more toward the parts you care about (understanding user insights).
With that in mind, let’s talk about the simplest case: when you can visualize everything immediately.
With a small dataset, visualize it right away to check for errors
If you have a small enough dataset, you can visualize everything immediately to check for things that don’t look right at first glance.
Ben Shneiderman, an Interaction designer and Data Visualization pioneer, used this technique when he got a slightly odd medical data set. When he visualized the data, he immediately found what was causing many errors.
A hospital code used to fill blank fields, 999, was in the age field of three patients. As a result, it registered that they had three 999-year-old patients, which would have thrown off any calculations they made around age.
You’re looking for these things when you visualize a small dataset. Rather than review each row or column for outliers, you’re looking for the obvious problems that might have occurred by mistyping information, missing values, or other outliers.
If you feel like you can’t make this work with your data because your data is most words and not numbers, there are two things you might do. The first is to check if you can do a bit of cleanup and turn things like 5-point Likert Scales (Strongly Disagree to Strongly Agree) into numbers to visualize it.
You can also use word clouds to get an introductory look at the data.
Remember, we’re not looking for statistical perfection at this stage. Instead, we’re checking to see if we put responses in the wrong column (such as several “Frustrating” responses for a question with answers 1–5).
Assuming these views look all right, we can move on to the next, more detailed view: Averages and Distribution.
Visualize averages and distribution for different insights
If that first glance at the data seems okay, you can do two quick calculations (and visualizations) to view your data in two insightful ways.
The first is to use the average (mean). Sum up all the responses in a column and divide by the number of values. For example, if you had 5 participants giving survey responses (from 1–5), you’d sum up all of the responses, then divide by 5.
Calculating and visualizing these averages can give you a sense of how users quickly responded to various questions. However, it’s most often helpful to visualize one other point of view as well: distribution. Distribution refers to how the data might be grouped, and it’s helpful to see if there are any patterns or insights in the data.
You probably didn’t realize that these two visualizations, Averages and Distribution, are a core part of every Amazon product review you run into.
If something has a 4.5-star rating, but the ratings are either 5* or 1*, you might look closer at the 1* to see if there is one common flaw that some customers encounter (for example, one piece always breaks).
However, just like it’s hard to trust an Amazon product with two reviews, it’s hard to trust the statistical validity of these averages. As a result, you probably cannot present these visualizations (and calculated averages) to stakeholders.
But that doesn’t mean it’s not helpful to understand your data.
Visualizations allow you to take a top-down view of your user research
These visualizations won’t work for every type of user research. If you have only open-ended questions in a user interview, you must go through it like usual.
However, this technique can be used in more places than you realize to provide quick clarity about the data. Some examples include:
5-point Likert scale questions (Strongly Disagree to Strongly agree)
Survey Responses
System Usability Score (SUS) ratings
Average Time on Task
Average errors per user
Average Task difficulty
etc.
Visualizing these, rather than going through the rows individually, helps you get a general sense of what might be happening and allows you to understand where you want to dig further.
Even if the visualizations don’t show anything out of the ordinary, it’s a quick way to check to see no outliers, errors, or anything out of the ordinary.
So the next time you feel intimidated about doing Data Analysis, see if you can use this technique to check out the data quickly. It might help you get a jumpstart on user research without being bogged down by tedious data checking first.
Want to become a UX professional? I’ve written a free e-book explaining how to make yourself an ideal UX candidate.
Kai Wong is a Senior Product Designer and a top Design Writer on Medium. His new book, Data-informed UX Design, explains small changes you can make regarding data to improve your UX Design process.