Designing for Data Scientists: Recognition over Recall
by Annie Darmofal
As a UI/UX designer at Civis Analytics, I’m sitting pretty. Literally, I sit 10 ft away from 40 applied data scientists that use Civis Platform for their work — a scenario most product teams dream about. I’ve thought a lot about designing for data scientists, and today we’re rolling out changes we made to put the information they need at their fingertips.
Being so close to users has advantages that are obvious to organizations practicing a design-thinking approach to product development. I find some of the most valuable insights come from moments outside of the design process. I get to learn about the projects data scientists are involved in every day, observe how they work naturally, and hear about what they did last Saturday. I get to know our users personally, and that’s incredibly helpful given the complexity of the data scientist persona.
Data scientists are smart, technical problem solvers that push boundaries with machine learning, algorithms, and visualizations. They love to build things themselves, and they usually do. (“For Data Scientists, By Data Scientists” isn’t just a marketing tagline, it’s how the Civis Platform came to be.) Some data scientists are creative and messy, others more structured, but they all have their own way of working that is important to preserve. They experiment and iterate through data science problems at high speed — so we turned our attention to making sure the platform’s user experience helps data scientists move fast.
Doing data science means keeping track of lots of tables, column names, model versions, scripts, and workflows. That’s a lot to remember, and doing data science is about problem solving — not memorization.
There were a number of places we could apply the design principle of “recognition over recall” — one of Nielsen’s 10 Heuristics for Interface Design. Memory retrieval is in fact hard, but it gets significantly easier with cues. This explains why sometimes you can’t come up with a person’s name, but you recognize them on the street. Or why multiple choice questions are easier than short answer or essay.
Being close to users, prototyping and testing, helps us find the balance of surfacing information without cluttering the space, and suggesting in a way that is helpful and not hindering. After all, our users don’t want to be limited.
We’ve minimized the need for recall in Civis Platform in the following areas:
- Homepage: The homepage displays a personalized list of recent activity and objects by status so you can spend less time hunting for jobs that are running, need your attention or require more configuring.
- Data Pane: An expandable view of your data structure in your Civis Platform Redshift database gives you quick access to tables, column names and other useful details, such as data type, row coverage and min and max values, from anywhere in Platform.
- Autocomplete in Query: Inline, autocomplete suggestions of your tables and columns allow you to recognize and insert what you are looking for — faster coding and fewer typos.
- Database indicator on Scripts: If you use multiple Redshift databases and are frequently switch back and forth, you’ll see an indication on the Script telling you the database in which it runs, eliminating the need to recall your settings at the time you created it.
(You can read more about these features here.)
By making information and functionality visible and easily accessible, data scientists can spend their cognitive energy on scaling models and building tools — more important tasks than remembering what they named their last job.