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

Personalization at scale: Mastering the challenges of personalization to create compelling user experiences

Rahim Daya (Pinterest)
4:20pm5:00pm Wednesday, March 7, 2018
Average rating: ***..
(3.50, 4 ratings)

Who is this presentation for?

  • Those working in product, machine learning, data science, or user experience design

What you'll learn

  • Explore Pinterest's frameworks for personalization
  • Learn techniques to identify data needed to build a personalization algorithm for a specific product vertical, methods to collect personalization data on users, and how to apply personalization to your user experience in a trustworthy way


Personalization is a powerful tool for building sticky and impactful product experiences, but what does “personalized” really mean, and how does one effectively build a personal experience into an existing product or service? Rahim Daya shares Pinterest’s frameworks for building personalized user experiences, explaining how Pinterest search departed from the legacy tradition of objective search to build a search experience that allows users to feel like Pinterest knows exactly what they were asking for.

Rahim walks you through the basic concepts of a personalization algorithm and breaks down the types of personalization and the data used in each. Since personalization can expose your product to certain risks, Rahim also discusses rules of thumb for identifying good and bad personalization. He then explores techniques to identify the right data sources needed to personalize results and details how Pinterest selects data sources for its algorithms and aligns its data models with how it conceptualizes personalization.

Often data sources needed for a personalization algorithm do not yet exist within a company’s dataset. Rahim demonstrates how to collect this data through product input, implicit discovery, and use of third-party data sources. Part of what allows Pinterest to build successful data product experiences is its design philosophy around providing value to users while collecting training data for future features. Rahim highlights how this fusion of design and data science helps Pinterest develop compelling user experiences.

Evaluation can become more nuanced once you introduce personalization into a product experience. Moving away from a unitary experience to one tailored to each individual user requires changing human evaluation techniques and success metrics. Rahim showcases tooling Pinterest built that allows the company to continue to iterate quickly. Rahim ends with a discussion of how Pinterest personalizes results while allowing users to adjust their settings in order to create a trustworthy experience. This includes ensuring that Pinterest does not overpersonalize results and restricting users to a few personalized topics.

Photo of Rahim Daya

Rahim Daya


Rahim Daya is head of search products at Pinterest. Previously, he led search and recommendation product teams at LinkedIn and Groupon.