Presented By O’Reilly and Cloudera
Make Data Work
March 5–6, 2018: Training
March 6–8, 2018: Tutorials & Conference
San Jose, CA

Breaking up the block: Using heterogenous population modeling to drive growth

Daniel Lurie (1989)
11:00am11:40am Wednesday, March 7, 2018

Who is this presentation for?

  • Analysts, data scientists, product managers, and business leaders

What you'll learn

  • Learn how Pinterest leverages external data to measure racial and income diversity in its user base and changed user modeling to drive growth


For consumer software products that grow through word of mouth, the initial user base can be relatively homogenous. Users invite the friends and family they know in real life, meaning they likely came from similar backgrounds. This similarity can appear in tangible items like device preference, in softer choices like personal interests and aspirations, or in demographic factors like race and income. As your user base grows, you will inevitably have to adapt your product to meet new needs. Blindly optimizing for key performance indicators will continue to improve your product for your core (homogeneous) users at the expense of future users.

At Pinterest, the initial user population was overwhelmingly US-based female and white, leading to algorithmic artifacts like the word “shoes” meaning high heels or “black hair” resulting in page after page of white women with black hair. But to grow internationally or even just further expand in the US, Pinterest had to undertake large efforts to debias its systems, leading to a product today that is majority international with a rapidly diversifying user base. Daniel Lurie explains how Pinterest leverages external data to measure racial and income diversity in its user base and changed user modeling to drive growth.

The first step to unlocking this growth is opportunity sizing. Dan shares a novel approach that uses census data to understand (in aggregate) how Pinterest’s historical growth and current usage varied based on demographics like race and income. The results confirmed that the early Pinterest user base was homogenous and that there was a large opportunity for growth, as more recent cohorts were much more diverse but less well served—particularly regarding new user experiences and content engagement. Dan also discusses the limitations of this technique and explains how it could be applied to other consumer applications to drive growth.

Photo of Daniel Lurie

Daniel Lurie


Daniel Lurie leads the product analytics and science team at Pinterest, a group that mixes deep data skills with strategic thinking to help Pinterest’s product team grow the company’s user base, develop new features, and increase engagement. The team’s work ranges from understanding product performance via A/B experiment analysis to identifying and sizing market opportunities to defining and tracking success through metrics. Previously, Dan led analytics for a sales-focused business line at LinkedIn and worked in consulting.