Please rotate your phone horizontally or reopen this page on a larger viewport to view case study content.

porsche motorsports

Prometheus-AI

An interactive web app for the Porsche Formula-E team to make key strategic decisions, optimise energy usage, and predict race outcomes.
machine learning
data viz
web
product design
Jump to design solutionView design process
my role
Lead Designer – user research, UX/ UI, motion
tools
Sketch, Principle, AfterEffects, Zeplin
team
Data scientists, Data engineers, Full-stack engineers, 1 UXR, 1 motion designer
Date
Mar 19 – Jul 19
backgroundresearchPrototypingsolutionimpact
KEY OUTCOMES

Design Impact

My team and I created a novel product that helped the Porsche Formula-E racing team analyze live racing data and predict race outcomes, which resulted in an 8% improvement in their racing performance and securing a long-term contract with McKinsey.
problem statement

How do you continuously predict a race outcome?

Race activity is volatile and as the race progresses, the Porsche team of analysts gain new information that could radically change their racing strategy. Our aim with Prometheus-AI was to provide a decision-making tool that could predict race outcomes and respond to volatility.
race plot time-lapse (click to view)
background

What is Formula-E?

Formula Electric is a motorsport championship that only uses electric cars. I like to call it "Formula 1's cooler, eco-friendly younger sibling." Each race is an unpredictable, wheel-to-wheel event in which managing your driver’s limited electric energy becomes an incredibly complex challenge.

Porsche's Formula-E Debut

Over four months of working closely with our client, McKinsey / QuantumBlack designed a bespoke, web-based tool that allowed the racing team to predict competitor outcomes, manage their car’s energy, and play out “what if?” forecasting scenarios. We called this tool Prometheus AI, which ultimately allowed Porsche to make better in-race strategic decisions.

To date, the Porsche team has seen great success by leveraging Prometheus AI during their fall training and their debut race, finishing second place in Al Diriyah and they have sought out my team’s services to continue building out the tool for the 2021 season.

Porsche's Formula-E Debut

In 2019, Porsche's legendary Le Mans racing team announced they were joining Formula-E for the 2020 season. Over four months of working closely with our client, McKinsey / QuantumBlack designed a bespoke, web-based tool that allowed the racing team to predict competitor outcomes, manage their car’s energy, and play out “what if?” forecasting scenarios. We called this tool Prometheus AI, which ultimately allowed Porsche to make better in-race strategic decisions.

To date, the Porsche team has seen great success by leveraging Prometheus AI during their fall training and their debut race, finishing second place in Al Diriyah and they have sought out my team’s services to continue building out the tool for the 2021 season.
research process

My research process required a variety of qualitative methods.

01. Subject Immersion

I watched dozens Formula-E races in an attempt to become a subject matter expert. I learned about racing regulations and fun features such as "Attack Mode" (like picking up 🍄 from Mario Kart).

02. Problem Framing

I hosted an internal team-alignment workshop to get a shared understanding of the problem ahead of us. This was a first step I took to unify my team before jumping into the abyss.

03. Current & Future State Workshop

I led two workshops with 15 members of the Porsche racing team to understand how racing is done now and how they expect it to change with assistive algorithms.

04. User Interviews

I conducted twelve, hour-long semi-structured interviews with Porsche team members, including race car drivers, race engineers, data analysts, and team managers.

05. Contextual Inquiry

I observed the Porsche team in a full-day training session, which allowed me to understand the nuances of their workflow, specifically how they use live-data to draw insights about the driver's performance.

06. Thematic Analysis

I transcribed and coded audio recordings of my interviews, uncovering insights about key tasks performed during a race and attitudes towards using technology to support decision-making.

01. Subject Immersion

I watched dozens Formula-E races in an attempt to become a subject matter expert. I learned about racing regulations and fun features such as "Attack Mode" (like picking up 🍄 from Mario Kart).

02. Problem Framing

I hosted an internal team-alignment workshop to get a shared understanding of the problem ahead of us. This was a first step I took to unify my team before jumping into the abyss.

03. Current & Future State Workshop

I led two workshops with 15 members of the Porsche racing team to understand how racing is done now and how they expect it to change with assistive algorithms.

04. User Interviews

I conducted twelve, hour-long semi-structured interviews with Porsche team members, including race car drivers, race engineers, data analysts, and team managers.

05. Contextual Inquiry

I observed the Porsche team in a full-day training session, which allowed me to understand the nuances of their workflow, specifically how they use live-data to draw insights about the driver's performance.

06. Thematic Analysis

I transcribed and coded audio recordings of my interviews, uncovering insights about key tasks performed during a race and attitudes towards using technology to support decision-making.

understanding machine learning

I sketched out schematics for our machine learning algorithms

reinforcement learning and analytics modules (click to expand)
design principles

I established design principles informing my design intent

In order to meet user needs, our design must be:

  • Performant – must be responsive at all points during race time (milliseconds matter)
  • Informative – support at-a-glance data consumption and task-switching
  • Trustworthy – engineers have to immediately understand and trust the recommendations
Research outputs

Using design outputs to shape product strategy.

Insights from my research activities played a crucial role in challenging my team's initial assumptions about our users and what we thought we should build. Insights into user painpoints, specifically track parameter-setting and visualising live racing data, allowed me to influence and prioritise development of Prometheus' key features.

journey map, personas, user epics, information architecture (click to expand)
ideation

How do we forecast a constantly changing future?

Once I defined my high-level epics, I began sketching UI concepts to bring our epics to life, and identify potentially complex data visualisations. From here on out, I began soliciting regular design feedback from my data scientists and the Porsche race engineers.

UI sketches for race plot, scenarios, parameter setting (click to expand)
data visualization concept sketches (click to expand)
prototyping

I created lo-fi protoypes to validate key concepts

Our algorithm was fed by live data, and would output a new racing forecast (future) every few seconds. Because thinking in this way was a paradigm shift for our users, it required extensive testing at a low fidelity before jumping into visual design. I used cognitive walkthroughs of key user flows with an InVision prototype and animated GIFs.

wireframe prototype, race time lapse (click to expand)
visual design

Optimising layouts and crafting custom UI components.

After determining an optimal information architecture, a major challenge for me was to lay out the content such that it was both aesthetically pleasing and easily understood by our expert users.

views for race plot, setting energy budgets, creating scenario (click to expand)
views for race plot, setting energy budgets, creating scenario (click to expand)
views for race plot, setting energy budgets, creating scenario (click to expand)
motion design

I used TV racing graphics to make Prometheus familiar.

After spending a race day with the team I observed that they spent 50% of their time watching the live TV feed, thus heavily relying on the data provided by on-screen graphics.

I hypothesized that by taking the vernacular of Formula Electric OSGs for Prometheus we would reduce the cognitive strain for our race engineers, and in early user testing we found that they immediately understood the leaderboard design.

KEY OUTCOMES

Design Impact

My team and I created a novel product that helped the Porsche Formula-E racing team analyze live racing data and predict race outcomes, which resulted in an 8% improvement in their racing performance and securing a long-term contract with McKinsey.
race plot start to finish, timelapse (click to expand)
adding flag and calculating scenario (click to expand)
reflection

Final Thoughts

This project really challenged me to understand what it means to build a power-user tool. I had to throw out some of my assumptions about making a product accessible to anyone and everyone. Generally, this way of thinking is contrary to the design of standard consumer digital products, so I was challenged to design this tool to fit the specific needs of the Porsche team.

Want to see more work?