[Databricks — Lakeview BI]
Designing to expand the user base
Jul 2022 - Apr 2023
It's rare to have the option to start from scratch and build a new experience with strong foundations.
The thoughtful context, information architecture and UI scalability of an experience are key to building something that can grow gracefully.
The creative team.
Myself, another designer focused on the core visualization experience, Visualization Tech lead, PM and a Principal Software Engineer
The setup.
Databricks is a technical tool geared towards technical users. The least technical users were considered the SQL analysts. However, we were given the goal to broaden their user base with a no-code BI tool, codenamed Lakeview
Complex ecosystem.
A core challenge was how to integrate our experience inside the rest of the rich yet complicated Databricks experience. Previously, users could create and save SQL queries over data stored in the Lakehouse. These queries could be used to drive visualizations, and these visualizations could be placed on dashboards. However, each of these levels were separate entities, saved separately with separate permissions. This meant the same query could be used for multiple visualizations and the same visualization could be used on multiple dashboards. However, that caused great confusion with understanding the impact of even the smallest change, and managing permissions and publishing of the artifacts was a huge challenge.
After many discussions and debates, we drove the consensus that the dashboard should be the primary artifact that owned the query and visualizations at the core of our new experience. This would allow us to create a dashboard-first experience, greatly simplifying the context switches for users, and match the expectations and mental model of our target users.
The previous dashboarding experience required you to create queries and visualizations in separate experiences, forcing users to have multiple context switches. I built diagrams and other supporting assets to help make the case for a dashboard-first experience.
A large challenge was also figuring out how to integrate what we built with the rest of the large Databricks ecosystem. I helped identify, present and evaluate options as we built the experience.
The design
From this alignment, I started to lay out the information architecture of the new experience. This involved a more interactive dashboard canvas and partnering with a colleague for a revamped visualization editor.
I started with building out the information architecture of the new experience, defining the various UI surfaces, how they relate to each other, and support the requirements of what we were building.
The previous visualization editor (left) was a complex dialog with many tabs and options. I collaborated with a team building the new visualization editor, building out prototypes and testing out options. I was then able to integrate this visualization editor into the much more contextual and smaller footprint of the sidebar.
My process
When it comes to building deep, creative, non-linear experiences, I delve into prototyping quickly from initial brainstorming. Some experiences need to be iteracted with and explored to test. This is especially true when it comes to data analytics — if you are striving to help users make sense of their data in a state of flow, you need to be able to observe them in a state of flow. The below are some examples of prototypes I built for the Lakeview project.
Using React, I developed several prototypes for testing out concepts in user studies.
Nailing the right visual and interactive fidelity is important for me to tell the right story and get the right feedback.
When it comes to domains such as data visualization, it can be important to prototype details so users can really test out how something might feel before we build it.