Usage Analytics Dashboard

Data visualization, visual design, UX design
Disclaimer: The product designs I describe and show are not necessarily representative of Thomson Reuters's product Westlaw. All data shown is fictional.

Duration

03.2020 — 04.2020
4 weeks

Team

Design & data visualization mentor — Corey Ouellette
Sole designer — me

Project overview

If you were putting in lots of work into providing help for a group of people, wouldn't you want to know if your hard work is actually creating positive impact?
‍
Westlaw is a database where lawyers can search through to find content like: case studies, regulations or legal journals. Think of it like the “google for lawyers”. Law librarians are the ones who manage the content on Westlaw. I designed a brand new dashboard experience using data visualizations to tell a story to law librarians of how lawyers are using the content resources they provide.

The high level problem I will be solving with this application is: Librarians don't normally get to interact with lawyers, therefore there's not much indication of if their work is creating positive impact. If not, they need to find out why and make changes to the current way they operate.

My contributions

Conducting customer conversations to identify user needs
Designing the dashboard experience from end-to-end
Prototyping and iterating on the visual designs, data visualizations and user experience

Outcome

Used by 1000+ librarians across Canada.
Connected librarians with lawyer usage data from 60+ different offices.

The Challenge

Right now, librarians are working without seeing their impact

Librarians don't normally get to interact with lawyers, therefore there's not much indication of if their work is creating positive impact. If not, they need to find out why and make changes to the current way they operate.

As administrators, law librarians are managing this database for lawyers to use. Changes must be made to the current way that librarians are almost blindly doing their work. At a high level, they need answers to questions like these:

Are the lawyers actually using the content? How often are they using it?
What type of data do lawyers find most useful?
What kind of lawyers are using which content?
Are lawyers finding the provided content useful?

Business goal

Use data about how Westlaw is being used to tell a story, in order to bridge the gap of communication between lawyers and law librarians.

Measure of success

Increase overall usage of the Westlaw by lawyers. To do this, the dashboard will help librarians understand how the Westlaw is being used.

The approach

01 Building empathy & finding the north star

The first step of the process is to figure out what exactly the librarians are looking to achieve from using this dashboard. I approached this by having conversations with librarians to build empathy. Throughout the conversations, they listed some insights that are missing that they hope the dashboard can help them gain. These insights were put into two categories to form the north star vision of the dashboard:
‍
High level view: A top level aggregated data to show insights on how frequently different demographics of lawyers are using Westlaw.
‍
Detailed level view: This level will show insights on what exactly the demographics of lawyers are looking at while using Westlaw.

02 Laying down the foundations

Based on the north star vision, we now have a better understanding of what data fields users might be looking for in the dashboard. Since a constraint of the projects is that the data will be updated manually, creating a data schema was important. This schema has to be accessible and easily understandable to the librarians who are not too data savvy.

03 Design & test

I worked on understanding the intention of why each type of chart is used and chose based on which charts will help the librarians get to the insights they need fastest.

Top level view

User activity by time

This graphs in this section will help answer: Are the lawyers actually using the content? How often are they using it?
‍
What immediately came to mind for representing this data set is to use area graphs, more specifically: stacked area graph or layered area graph. Here are some of the pros and cons of both that I took into consideration.

For the first iteration, it almost seems like I couldn’t have the best of both worlds (be able to compare total monthly values and individual divisions monthly values at the same time), so we opted for a graph that could toggle between stacked area graph and line graph. After experimenting further, I realized adding another line on the line graph (grey line on final design) representing the total value simply gives the best of both worlds and avoids the user from toggling back and forth.

Top level view

Percentage user activity by division

One thing that the line graph still is not able to achieve is compare monthly values against total values. This is why a donut chart was added. I’ve come to learn that there many designers have very strong feelings about donut vs pie charts! Here is my take on why I chose to use a donut chart in this scenario.

Since screen real estate is precious and this chart is not necessarily the most important one that a librarian needs to see right away, a donut graph is chosen instead of a pie graph.

Top level view

User activity distribution by division

This second part of the top level view goes into more detail about specific users.
‍
Before I dive into the design process for the Usage activity distribution chart, one downside of the final design is that the bar graphs don't all start at 0. Therefore it’s hard to compare total values side by side. This is why Divisions Ranked by Activity has been added to simply summarize the values.

This chart took quite a few iterations to get to the final design. Often when I’m sketching or wireframing, I don't use colours. And it's easy to forget that I can use colours to my advantage. In the end, I was inspired by heat maps in combination with bar graphs to show both how many searches lawyers are making and how many lawyers are making that amount of searches.

Detailed level view

User activity per content type

This is the detailed level view where specifics of how the lawyers are interacting with the content is represented. This section helps answer the questions: What type of data do lawyers find most useful? What kind of lawyers are using which content?

A stacked bar chart was chosen as it was the simplest way of representing the data. To save space, the bars are stacked on top of one another to save space.

Testing with real data

I have learnt from this project that my process for designing data visualizations should include a very important step of testing with real data. This prevents graphs from looking like this:

In a more realistic scenario, if a data used for a bar graph varied too much and the graph looks like the stacked bar chart below on the left, then using a percentage bar graph might be more helpful.

Other honorable mentions

Here is one honorable mention of a chart that made an short lived appearance in the very initial ideation phase. I was inspired by the 3D structure of the data we are visualizing and created this bubble graph. Although this customized visualization would've been a fun addition to the dashboard, it was simply not practical enough to fit into an enterprise software. Keeping in mind the character trait of librarians not being very data savvy, this design did not make the final cut after all.

What did I learn?

The head law librarian along with other librarians were involved in giving feedback to all the design iterations throughout the process. The dashboard is now being used by 1000+ librarians across 60+ different offices. It helps law librarians answer the question: are lawyers finding the provided content useful?
‍
Designing enterprise software is quite a good exercise for both product thinking and also user research. As someone who knew almost nothing about what law librarians did coming into the project, I've had to ask hundreds of questions in attempts to understand how librarians operate from day to day and almost every time there is always just one more question that come up as I iterate my design. "If only I knew which of these two values they would find more useful to show!" A/B testing made a regular appearance in my user feedback sessions, this allows users to see tangible visual designs and give me their radical candor on all the variations.

Thanks for reading!