JASON CARAGAN

☀ OAKLAND, CA

QUIZLET

Enable all students to find the best study materials for their biggest exams, faster.

Enable all students to find the best study materials for their biggest exams, faster.

In the Fall 2025, Quizlet’s Search and Recommendations team launched an initiative aiming to get students to get to the best-quality, course-specific study content sooner. It included features spanning the search journey end-to-end: personalized as-you-type suggestions, interactive flashcard previews when evaluating sets, and more granular results filtering.

Research · Interaction Design · Prototyping
2024—2025

Problem

Three intersecting problems made this work a priority:

  • SEO traffic was declining, UI was stale: students increasingly used LLMs to find Quizlet content; search hadn't meaningfully changed since 2022, despite search results being the third most-trafficked page on the site (hit by 20% of logged-in users)

  • Design and tech debt accumulated from competing product teams seeking real estate on the results page to highlight new content types often at the expense of pushing more relevant results down the page

  • Students struggled to scan results, finding it difficult to narrow down what they actually needed. Results weren't personalized to profile and behavior, and filters were hidden by default.

Measuring success

Our main goal was to get students to the right course materials faster, reducing time spent searching for and evaluating content so they could spend more time actually studying.

Our hypothesis: streamlining and enabling more personalized search results would increase search-to-study conversion, lift study engagement metrics, and improve search retention.

Mapping and prioritizing the search journey

We decided to sequence the work in four phases, prioritizing the results page first.

M1: Search results page: reworked content cards + basic filtering (M1, Q2-3 2026)
M2: Result content previews (starting with interactive flashcards)
M3: Typeahead (designed to carry across AI coach and Workspaces)
M4: Additional filtering and disambiguation

Initial prototypes and concept testing

I built out three prototype concepts (via FigmaMake, Cursor, Vercel) to put in front of students, ranging on a spectrum of content density and interactivity:

  • A: Copy-focused (title and content snippet)

  • B: Interaction-focused (flippable flashcards)

  • C: Hybrid (elements of both, but requiring more vertical real estate per result)

Though the hybrid solution (C) was the preferred choice for most testers, we chose to pursue full interactivity in a later milestone but revise the visual hierarchy in the short term, a more cost-effective path to the same outcome.

Reassessing informational hierarchy

A bulk of my efforts centered on the information hierarchy presented within result cards. We had a lot of metadata to work with (dates, popularity, course, number of cards, rating, etc.) . What did a student actually need to see at a glance to decide if this set is worth studying? What attributes could live in a filter? What should be hidden until they dive deeper?

Informed by user surveys and multiple rounds of prototype testing, we prioritized school, course and in-document content preview over username, popularity and user ratings.

Solution and impact

A/B tests isolating the impact of three major changes were slated for implementation in Q3 of 2026: results grouping (shelving) weighted more heavily towards user-specific course enrollment and flashcards (Quizlet's most popular content type); information-dense, interactive flashcard results; and redesigned results filtering.

We expect to see positive impacts on the experiment variant: less time spent searching, more time studying, with increases to search-to-study conversion and DAU:MAU averages for students using the search experience to discover and evaluate study content.

GET IN TOUCH

Wanna work together? Let’s connect.

Wanna work together? Let’s connect.