Overview

IBM Watson Studio provides tools for data scientists, application developers and subject matter experts to collaboratively and easily work with data to build and train models at scale. Watson Studio, ever evolving to solve needs for our users, implemented Anaconda-based environments for both R and Python languages. Environments are beneficial because they can contain Python packages other than ones currently used by any one specific user. For a team to have environments within Watson Studio meant they could have consistent space, any team member could then be confident that the code, and subsequently their machine models would run and function the same for anyone.


Opportunity

Environments have limitless configurations and are dependent on the capabilities of the servers their on. We had to ensure that the UI of Environment set up could be as simple as select and go and robust enough to satisfy complex engineering requirements.

Exploration for environment designs was started before I joined the Watson Studio design team. It was quickly handed over as my main focal. My first steps were to get in front of my Product Manager (PM) to understand the current state of the issue, review the designs with him and uncover the issues the team was facing at that time. Much of the churn over the design was uncovering backend engineering blockers and capabilities. As I was unfamiliar with the technicalities of environments at the time I was equally spending time researching and learning the benefits, functionalities and market examples of Anaconda-based environments in other products. Together, my PM and I spent time pouring over new wire flows I designed and agreed on a timeline to include user research to validate some assumptions we had.


Research

Data professionals work with multiple, non-collaborative tools. Many teams are newly forming where legacy data professionals often worked alone and in silos. Many open source platforms are popping up but getting them to work together isn’t always easy and the team finds they are moving between tools to complete one task. Thousands of lines of data are still often analyzed in spreadsheets then connected to git repositories and analyzed by programs written in different languages using different products.

 
Screen Shot 2018-10-24 at 7.36.43 PM.png

Method

I conducted the planning and execution of research for environments with the support of my teammate. The process consisted of taking mid-fi wires, and after iteration hi-fi designs, to a handful of users who use Watson Studio in their daily workflow. This two-step research strategy helped me identify pain points for our users early on and focus on areas where we could make environment set up easy to understand and use. Bringing the user in early and often is a methodology consistent with IBM Design Thinking practices known as the ‘Loop.’

 

Findings

Pain Point 1: too much UI

Watson Studio is based on open source platforms. Providing additional UI was inhibiting natural use of open source products for some power users.

Pain Point 2: Cognitive overload

Environment configuration details was too complex. The users were getting lost in definitions.

Outcome

We took our feedback from our sponsor users to reduce the UI overhead as much as possible, provide clear language through out set up, and provide documentation for more in depth research and learning about Watson Studio environments. Environments with connected runtimes are a direct source of revenue for Watson Studio and a source of cost for our users. We provided a clear usage table to allow users to easily stop an environment from running when not required. The runtime improvements and product releases ultimately ranked Watson Studio on Forresters’ top machine learning solutions list of 2016.

It is a perfectly balanced PAML solution for enterprise data science teams that want the productivity of visual tools and access to the latest open source via a notebook-based coding interface.
— Forrester Wave: Multimodal Predictive Analytics and Machine Learning Solutions, Q3 2018