Enhancing 1 Billion Daily Searches
Bing AI
OPPORTUNITY
Autosuggestions are triggered for every search conducted on Bing, helping users find information more efficiently. How can we enhance Bing's autosuggestions to deliver a more relevant and engaging search experience for its 100 million daily active users?
MY ROLE
Led the end-to-end conceptualization of the project, worked closely with the cross-functional product triad, & streamlined collaboration across Bing verticals
TEAM (ICs)
1 Designer, 2 Product Managers & 12 Software Engineers
Direct Initiative from Microsoft Leadership
This project was initiated based on a direct ask from the CVP of Web Services and was presented to the leadership team on multiple occasions throughout its progression.
What are Recommendation Experiences?
Before we begin, let's understand what I mean by recommendation experiences in the context of search. Also, let's look at some of my other related work in Bing!
Research
Conducted surveys, in-depth interviews, secondary research & leveraged internal resources to examine how users engage with autosuggestions and identify their needs & opportunities for improvement.
1:1 User Interviews: Conducted interviews on UserTesting.com to gather qualitative insights into the usage of the feature.
Surveys: Conducted surveys via UXLabs, a user recruitment platform, to obtain quantitative insights into the feature.
Faster Search and Low Interaction Cost: Users seek to obtain answers promptly with the least effort.
Context is Important: The value and purpose of suggestions vary throughout the search journey.
Lack of Visual Hierarchy: Users struggle to distinguish suggestions due to the lack of visual differentiation.
Zooming Out: A Fresh Perspective
Considering autosuggestions not just as an isolated feature but a crucial element of a user's search experience to create better searches.
User Intent Breakdown in the Journey
Let’s break down each step of the user journey to better understand the user’s intent at every stage.
Zero Input State
The user's intention remains unclear at this point.
Can be a gateway to the internet, where the users can see all the things happening around the world.
Query Formulation
We can partially discern the user's intent from the initial input.
By assisting users in refining their queries more effectively, we can streamline their path to results with fewer keystrokes.
Requery
Users may return to the search box to fine-tune their query or to dive deeper into a topic.
Can provide relevant related content to enhance their search experience.
Collaborative brainstorming session with senior designers, PMs, and engineers. Worked together to generate product ideas, focusing on solutions that would address our users' needs. Discussed these ideas as a group, ensuring that everyone had input and that the most impactful solutions were identified.
Key Considerations
Product Metrics
Keystrokes Saved Per Query (KSPQ)
DAU and Segment DAU
Technical Constraints
Latency
Algorithmic Precision
Data & Legal
Dependence on the type and quality of data available to Bing
Ensuring that any data we planned to fetch and display in autosuggestions was legally permissible, complying with all regulations
Anatomy of an AS Box
Query Formulation
Rich Entities - Predicting Next Steps
Displaying relevant data from Bing's SERP upfront for entity queries to save users' time and streamline their journey.
Instant Answers - Saving Keystrokes
Saving user’s time by providing answers within autosuggest.
Requery Experience
Exploration Intent
Leveraged Bing’s vertical segments (e.g., travel, health, shopping) to provide more relevant autosuggestions for the user's exploratory intent. This initiative helped in strategically driving engagement towards Bing’s verticals and supporting business growth.
Information Hierarchy
Requery Framework
Before vs After
Assisting users in trip planning by minimizing search efforts
Redirecting users to Bing’s shopping vertical
Displaying essential health information and professionals QnA
This approach successfully demonstrated how a deep understanding of user behavior could be harnessed to optimize product features, drive engagement, and ultimately contribute to Bing’s success across multiple verticals.
Inclusive Design
Focus Order
Ensured that the sequential navigation order adhered to WCAG guidelines.
Collaborated closely with the front-end engineering team to establish clear tab navigation rules for keyboard-only users.
Patent Granted
- Multi-Intent Searching (411836-US-NP): Granted on February 8, 2024, this patent focuses on enhancing the efficiency and relevance of query suggestions through a sophisticated information management system.
- Multi-Level Autosuggestions (412693-US-NP): Granted on February 17, 2023, this patent addresses advanced techniques for layering and refining autosuggestions, offering users a more intuitive and multi-faceted search experience.
Looking ahead, there is significant potential to further evolve these concepts by integrating them with co-pilot, which could provide even more personalized and context-aware suggestions to users.
