Custom Analytics Dashboard
Leveraging AI & machine learning to combat cognitive bias in market research
We created a market research analytics tool for a pharmaceutical company that helps researchers analyze trends and challenge their assumptions when reviewing interview transcripts through the use of machine learning and natural language processing.
Who?
Our client was a pharmaceutical company. I worked closely with a developer and two AI experts to bring my designs to life.
When?
December 2020 – February 2021
Where?
IBM Garage, collaboration and outcomes delivered remotely
How?
Sketch, Invision, Jira
Our Process
We were challenged to create a tool that leverages machine learning and AI to help market researchers analyze interview transcript files and identify recurring trends across sets of interviews.
We identified an existing IBM Watson product that would deliver the level of trend prediction and data analysis our client was searching for, but the client feared the existing interface was overwhelming to navigate and would only serve a small group of “power users.” My team proposed a solution: a custom dashboard that surfaces the most relevant data in a clean, easy-to-read format, powered by the underlying AI technology.
I began with an audit of the IBM product that would be powering our custom dashboard. I created a rough user flow, taking note of specific areas my client wanted to leverage. I also tracked the metrics we could expose based on sample data. This audit phase informed my initial sketches and conversations with the client as I showed them what was possible to surface in the custom dashboard.
Below are some initial wireframes I created to illustrate how a market researcher would navigate through the tool. These wireframes helped us ask the right questions and learn more about how our client would upload data to the system. With a defined user flow in mind, I moved forward and focused on the dashboard’s design.
My design process was iterative and collaborative; I involved the client and sponsor users to keep feedback driving iterations, Watson SMEs to better understand the technology powering the dashboard, and a developer to ensure my wireframes were feasible.
Our final dashboard (partially pictured below) allows market researchers to view analytics and trends across multiple interview audio files. Watson speech-to-text generates accurate transcripts from audio files and natural language processing algorithms can identify similarities and trends across files. The dashboard can provide at-a-glance insight into hundreds of hours of interviews, greatly improving market researchers’ synthesis efforts and reducing any cognitive bias that may sway their analysis.
Outcomes
In the span of a few months, my team was able to deliver a fully functional MVP dashboard that provided researchers with analytics on real data and allowed the company to validate their riskiest assumptions. This project was a study in the power of Agile work methods; our team had a diverse range of skills, and we worked closely with the product owner each week to prioritize and implement functionality.
This MVP will serve as a springboard for future iterations as the company tests it’s efficacy with real users.