Benchmarking

Final version of the benchmarking widget, meant to be used in GE APM (Application Performance Management) dashboard.

The Challenge: Power plant workers and managers need to know how their assets are performing against other plants monitored by GE, while maintaining anonymity between assets not owned by the same organization. Their old solution is frustrating to use and difficult to read.

Solving for: GE Internal Customers: Sr. Risk Mitigation Manager, Engineering Manager of Performance Engineering, Engineering Manager of Fleet Management.

GE External Customers: Asset Manager, Plant Manager, Operations and Maintenance Manager, and Executive. All personas, internal and external, are power plant workers or managers who are customers of GE Power APM (Application Performance Management).

Used data from role based personas created by team UX Researcher and panel of GE Internal customers for testing. 

Role: UX Designer on a UX team. 

This widget was part of the project to migrate the complete functionality of a legacy GE web tool called MyFleet into APM, GE's primary IoT (Internet of Things) application for asset management, which provides real time data to users. Product management had made the decision to migrate the individual pieces of MyFleet as widgets to the Dashboards microapp that lives within the larger APM application. Benchmarking functionality already lived in MyFleet, but its design was frustrating for customers and its performance was poor. The UX Researchers on our team validated that customers still needed the Benchmarking functionality before we proceeded with this project.

Legacy MyFleet Benchmarking tool (not designed by me).

Some main issues I identified with the previous solution were as follows: 

  • Colors used provided no indication of what being in a certain quartile meant for the asset

  • Hidden KPIs (Key Performance Indicators) that users could gauge their asset's performance based on

  • Confusing 'Filter' and 'Benchmark' functionality

  • Data points were hard to read when many are clustered around the same value

  • Out-of-the-way, time consuming configuration

  • Unclear zoom or pan functionality

Phase 1: White board ideating of widget.

As a first pass at this redesign, I did some white board sketching with my team lead. I sketched out a possible solution that used the traditional mathematical representation of quartiles, which I thought might be a clearer alternative to the previous solution. In discussions with users, we determined that organizing assets into quartiles was important to the users' understanding of where their assets stood in relation to each other.

The next step was to wireframe additional possibilities for our initial user testing sessions. In the below options, I changed the quartile colors to match alert colors - from green representing best, signifying that they were in the top of their class, to red representing worst. I also made selected filter options visible, as well as selected options for the 'benchmarking' functionality that determined which regions of the larger GE fleet your units were compared against. Each of the below options match the newest release of the GE Predix Platform design system, which encompasses the APM app. For simplicity, they only represent the 'Output' KPI. 

User Testing Option 1: This option is most similar to the original solution in terms of how the quartiles are represented, but uses alert colors.

User Testing Option 2: Percentile mapped on the y-axis, and frame types represented by different shapes.

User Testing Option 4: Similar functionality to option 2, but with a different visual treatment.

User Testing Option 3: Using the traditional mathematical representation of quartiles, as sketched on the whiteboard in the above image. Frame types are mapped on the y-axis, and colors are used per asset to mark which quartile they are in.

Due to availability, most of our user testing was done with the GE Internal personas. Though I felt that Option 3, the box-and-whisker plot, made it more obvious to the user that the chart mapped quartiles, the users themselves were typically middle-aged men and lifelong power plant workers who had not seen a box-and-whisker plot for some decades. To them, the 'whiskers' seemed just as arbitrary as the rectangles, but were less familiar. Option 4 was favored by some as more visually pleasing but more that one Senior Risk Mitigation Manager, who would need to benchmark dozens of assets at once across many frame types, raised the question of scalability and clarity with this representation. For our next round of user testing, I modeled a much larger data set for options 1 and 4. 

Option 1 modeled with a larger data set.

Option 4 modeled with a larger data set.

In this next round of user testing, it became clear that users prefered the visual representation of quartiles to structurally remain the same as in the legacy tool. They felt that the rectangles gave them clear indications of how close an asset was to the edge of a particular quartile, which was useful for the GE Internal personas in writing reports. Also, comparisons were only truly useful within frame types, and rectangles made it easier to focus on assets of one frame type at a time. The change of colors was appreciated, as was displaying the selected filters. Now that I was settled on a visual structure, I hoped to solve the remaining problems I had previously identified. 

To solve the issue of the number of clarity when multiple data points were clustered together, I proposed a solution in which a cluster was represented by a larger circle and the user would be able to hover for more information about each asset that lied within. To give the user zooming and panning functionality, I added the GE Design components on the top right of the widget. Users still struggled with the idea of setting the geographical regions of the GE Fleet that would compose the quartiles which their units would be compared against, confusingly called 'Benchmark' in the legacy solution, so I consumed that functionality into 'Filters.' 

Example representation with clusters of data points.

Expanded filter functionality.

The configuration that I designed for this widget can be seen in the interaction guide below.

 

To learn more about the final product that went through development, please reference the following two assets that I created for the engineering team.