Data Science and Design: A Tale of Two Tribes
Summary
In this talk, Jon explores the often misunderstood relationship between data scientists and designers and how their collaboration can create significant customer value. Drawing on his experience at Intuit, he discusses developing an enterprise-wide A/B testing platform that initially struggled with adoption until working closely with designers to identify key leading metrics like platform ease of use and time to deployment, rather than just lagging outcomes. Jon also recounts a retail case where overwhelming daily data was distilled by studying top-performing store managers in context, enabling the team to design a focused dashboard that increased sales conversion by 10%. Lastly, he highlights eBay's approach to extracting meaningful signals from vast transaction data by partnering designers with 300 data analysts to streamline workflows and reduce insight time from weeks to a day, democratizing access across the company. He concludes that integrating data science fully into product teams alongside design and development creates more delightful, insight-driven products.
Key Insights
-
•
Less than 1% of global data is used to generate human insights, showing the value in efficient data utilization.
-
•
A data scientist combines statistical skills, hacking mindset, and entrepreneurial thinking to solve complex problems.
-
•
At Intuit, pairing designers with data scientists uncovered that ease of use and awareness were key to increasing adoption of an A/B testing platform.
-
•
Lagging indicators like number of experiments performed or features deployed are insufficient alone; leading indicators such as time to consumption and user awareness better drive adoption.
-
•
Ethnographic studies can reveal user behaviors and misalignments in what enterprise teams measure versus what actually drives business outcomes.
-
•
Studying the best-performing store leaders on the floor uncovered critical behaviors that informed a focused dashboard replacing overwhelming 424-page daily reports.
-
•
A simple data product tracking live customer-to-employee ratios, transaction progress, staff availability, and service queue length helped retail store managers improve performance and customer experience.
-
•
eBay’s approach to ‘signal extraction’ from massive data involves time series analysis, decision trees, and close designer-analyst collaboration to cut insight delivery from weeks to a day.
-
•
Democratizing data insights across an enterprise empowers broader decision-making beyond traditional analyst-only access.
-
•
Integrating data science as a core pillar alongside design, development, and product management can transform teams into more effective problem solvers and innovators.
Notable Quotes
"Since we started this presentation, enough data has been collected to fill everyone’s hard drive about 500,000 times, and fewer than 1% is ever used for insights."
"Data scientists are part stats geek, part hacker, and part entrepreneur."
"Develop deep customer empathy — really understand who’s using a product and why — walk a day in their shoes."
"Rapid experimentation with customers is key. Iterate, iterate, iterate."
"It doesn’t matter what the insight is if no one does anything with it."
"We realized we were measuring the wrong things — we focused on lagging indicators rather than the leading ones we could influence."
"Surprisingly, many developers didn’t even know our testing platform existed."
"By studying the best store leaders, we learned patterns that informed a dashboard reducing 424 pages of data to a few key, actionable metrics."
"eBay cut the time it took to get answers from weeks to one day by integrating design with data analysis workflows."
"Bringing data science into the core product process is like adding a fourth leg to the stool — design, development, product management, and data together create delight."
Or choose a question:
More Videos
"Inviting other teams into research sessions lets us triangulate findings by combining interviews, observations, and focus groups."
Joanna Vodopivec Prabhas PokharelOne Research Team for All - Influence Without Authority
March 9, 2022
"Careers here are jungle gyms, not ladders; you can move laterally and explore many paths."
Louis Rosenfeld Lashanda Hodge Senongo Akpem Chris HodowanecBecoming a Civic Designer: Making the Move from Private to Public Sector
November 17, 2022
"We are closing out design ops 2024 with innovation curated by the incomparable John Kuda."
Bria AlexanderDay 3 Welcome
September 25, 2024
"Am I getting what I wanna get out of this conference? What’s not coming and what would I still like to learn?"
Uday Gajendar Louis RosenfeldDay 2 Welcome
June 5, 2024
"Frequent, bite-sized training is crucial so staff actually remember how to support customers with disabilities."
Sam ProulxOnline Shopping: Designing an Accessible Experience
June 7, 2023
"Unpaid design exercises are unfair because they don’t take into account people's real-life constraints."
Russ UngerOnboarding: The Ecosystem, not the Afterthought
November 7, 2017
"Physical prototyping is a tool to explore interaction modalities and physically connected environments beyond all things digital."
Catherine DubutBridging Physical and Digital Spaces: Approaches to Retail Service Design
March 18, 2021
"Characters live at the extreme of radically adaptive AI—they react in the moment and follow the user’s lead."
Josh Clark Veronika KindredSentient Design: New Postures for AI-Mediated Experiences (2nd of 3 seminars)
January 29, 2025
"Ownership needs to be shared. The craft of making is similar to raising a child — it takes a village."
Dantley DavisLeadership & Diversity—A Fireside Chat with Dantley Davis
September 17, 2020
Latest Books All books
Dig deeper with the Rosenbot
How can organizations manage the noise created by non-expert researchers conducting user research?
Is it possible to bypass HIPAA by initially targeting veterinary markets for clinical app development?
How can research repositories score or show the impact of research recommendations to enhance organizational learning?