Powering teams to build better software with data
Evolving a design system and webtool interface
1
Research
We spent three weeks collaborating with the core Split team in conducting user interviews and market research to understand the unique and complex requirements of their products and customers.
Competitive Analysis
We looked at how direct competitors, Amplitude and Looker, organize the robust data, features, and visualizations of their customer-facing BI tools, dashboards, and builders.
User Interviews
To understand the opportunity areas for Split we asked current users from companies such as HP and Twilio what could be improved and what features they’d like to see.
Learnings
Declutter dashboards
Use a flexible yet strict column system to ensure that interfaces stay clean and elements are easy to scan.
Establish color hierarchy
Limit and apply the same value of colors across the platform in order to create consistency and legibility.
Organize prevalent information
Ensure that context, such as filters, tooltips, timestamps, and advanced information, is readily available to the user and allow for user-customization.
Balance functionality and simplicity
The interactions must function as intended to be useful to the user; be wary of oversimplification.
2
Approach
We spent three weeks mapping out the product roadmap by identifying the key user needs and requests to address, and creating design system complete with visual language, components, and states.
Based on our research, we synthesized a list of high-level epics that covers the common based on customer needs and requests. We used these epics to prioritize which features and flows to address going forward.
• I want to create and manage... lists of who will see a feature.
• I want to create... feature tests and define who will see them.
• I want to know... who is seeing a feature.
• I want to know... what features a particular user or group is seeing.
• I want to turn... features on/off entirely.
• I want to turn... features on/off for specific groups or users.
• I want to see... all features active/inactive for a given environment.
• I want to see... all features active at a given time for a given environment.
• I want to see... where a given feature is active/inactive.
• I want to know... how a feature is affecting a given environment.
• I want to know... if any feature is negatively affecting a user.
• I want to know... how well a feature is performing with users.
• I want to know... how well our teams are doing deploying features.
• I want to know... if any features are orphaned or can be safely removed.
• I want to let... other departments know about my feature and when it’s available.
We found that sequential color palettes are ideal for data that shows a low to high progression such as time, while a diverging color palettes are ideal for visualizations that show degrees of change or variance from the median. We ensured that hues and value are color-blind friendly.
We learned that to show overall value and composition over time, use a stacked bar or area chart. And to focus solely on composition, a 100% stacked chart will focus only on distribution of values at the expense of overall value (this is akin to showing a series of pie/donut charts over a time interval). Color-coding and range bands provide additional context and improve scanability.
3
Execution
We spent two weeks developing mockups to realize a thoughtful experience that's visually simple on the surface yet supports the complex use cases and requirements of Split's user base.
Admin Mode + User Permissions
Configuration Management
Feature Flags + Experimentation
To the core team’s satisfaction, we nimbly and successfully shipped the newest version of Split’s platform before the holidays. In the end, we delivered a robust yet simple design system that presents the highly complex functionality in an intuitive manner to the user. In the short timeline of the project, we were able to figure out and prioritize the most impactful user needs and requests to solve. The new features and workflows we added bolster Split in their mission to help engineer teams build better products faster with data and insights.
Robust design system
New core workflows
Capabilities expanded
The team was happy with the quality of the solutions that we arrived at. Though, I feel that we missed an opportunity to create tiny moments of delight in an otherwise standardized and formal experience, for example, a playful interaction when a configuration is created or killed. Such an appropriately rigid platform could handle small elements of playfulness. The big lesson I learned through this project is that simplicity isn't always the goal. Some things, like building software, are just complex and complicated. Intelligibility is the aim; simplicity is the byproduct.
