Role: UI/UX lead | Team: Led the design for Model Builder’s development while partnering with 5 Designers and 2 Researchers from companion experiences | Timeline: January – August 2020
Model Builder allows users across Capital One’s finance department to create, edit and run financial models in repeatable production workflows that are critical for understanding raw data and making projections. Model Builder is key capability within an integrated ecosystem of data solutions called the ‘Finance Platform.’
Curiosity
Background
In order to derive insights about Capital One’s financial position that inform both long and short term strategy, associates within finance department deal with fragmented sources of information and disconnected, manual financial modeling processes. Not only does this reduce operating efficiency and cause frustration, it leaves important business processes open to human error and increases risk.
Problem Definition
How might we create, track, verify and interpret models in repeatable production workflows?
Insights
Definitions
Creativity
Low Fidelity Mocks
Early stage designs from our weeklong sprint took inspiration from Kanban, Gantt and flow charts to simulate controlled manipulation of the models inside a system with defined rules.
I helped to test these with users at the end fo the sprint. There were two big takeaways from this first round of tests:
Users needed real-time intelligent feedback from the platform to avoid preventable errors in the model’s logic and suggest appropriate process steps
Users needed to be able to access the details about each component of the model in order to create the right data transformations for their work
Mid Fidelity Prototypes
In response to this feedback, I designed the next iteration using designated swim lanes to help associates visualize the structure of their model before building it and make sure their process adhered to the appropriate logic.
As I moved into higher fidelity, I started to develop the left and right drawers which served as the integrated handoffs between other parts of the data ecosystem.
I partnering closely with the designers and partners working on these companion products in a tight and collaborative development cycle.
High Fidelity Prototype
After promising results for in user tests, I started partnering closely with a front end developer to realize the core capability of automatically linking and visualizing the appropriate data in the model’s logic model in real time.
Our partnership led to many micro iterations and refinements of the product’s UI to balance the tech constraints with our design criteria. Ultimately the solution (left) balanced both.
Impact
Launch
Model Builder launched Spring 2021 and is being scaled to a growing number of finance associates. The tech lead on the work stream celebrated an “enduring emphasis on human centered and product centered approach” as playing an outsized role in the launch’s success so far.
Model Builder significantly improves associates’ ability to generate and manipulate crucial model workflows by cutting back inefficiency, uncertainty and redundancy in how these complex data manipulations work.