TL;DR
Date & Role
May 22- Sept 22 • Product Design Intern
The problem
Weather patterns are changing fast, and reliable forecasts aren’t always easy to find. We noticed users often misinterpreted terms like “percent chance of rain,” which led to confusion and reduced trust in Google’s weather results.
Goal
Help users trust Google to give them information eather probabilities clear, intuitive, and actionable especially during daily routines and critical extreme weather events.
What was one big challenge
Turn complex forecasts into intuitive, everyday decisions for users across North America and Asia, bridging the gap between probability models and real-world action.
Sahib's Contribution
Designed interactive prototypes and led usability testing across North America and Asia to uncover how users interpret weather probabilities and proposed designs to make forecasts more intuitive, clear, and trustworthy.
Impact
The prototypes guided teams designing weather search results, emphasized the importance of probability representations to VP and Director of AIUX and helped me have my first experience designign at scale. I learnt how to bring visibiltiy to my work across multiple teams at Google Research, helping them apply findings to improve their own models and user experiences.
Math is hard. Probabilities are even harder, both don’t always make sense. Take weather forecasts, for example. Most people don’t really know what a “30% chance of rain” means. Is it going to rain 30% of the day? In 30% of the area? Or at 30% strength?
In the summer of 2022, I joined the AIUX team to make weather predictions more easily communicable to users meeting them where they already were: Google Weather search results.
The current assistant UI was confusing and I wanted the solution to be multimodal from the ground up blending voice and visual interactions. Through usability testing in North America and Asia, I uncovered patterns that helped simplify the UI and make the assistant feel more intuitive and reliable across regions.
The end result was a simplified UI that guided users’ attention to the most important weather notifications. In user testing, over 85% of participants rated it as reliable, and more than 50% reported increased trust in the forecast. These insights helped inform the next set of updates for Google Assistant’s weather experience.
What made this even more rewarding was how cross-functional the work became. Some of the most valuable input came from outside our immediate team, partner orgs, researchers, even folks experimenting on side projects. Good ideas deserve a path forward, no matter where they come from.
Extending our teams partnership with external orgs , I created a detailed user study plan to learn more about the needs of users from weather search results and how we can help them. It led to a series of prototypes which integrated google weather search into their everyday life. I wanted to bring functionality over everything and implemented Material 3 guidelines.
I designed for common moments where upcoming weather affects what users want to do next. By integrating contextual weather snippets across key touchpoints, we created a seamless and dynamic experience. Each user’s weather updates were uniquely relevant—surfaced exactly when and where they mattered.
After exploring further, I realized that scale was essential to solve for when it came to weather search. People all over the globe trust Google to tell them the weather and if its wrong, Google is trusted less. I pushed for immersive international research, first with vendor partners, and later with our expanding in-house UXR team.
These research insights grounded our roadmap and helped us advocate for everything from new features to broader org investment. Along the way, I learned that the Floods team at Google was working on an adjacent project where users needed to interpret probabilistic forecasts in high-stakes situations. Our work helped shape how Google could present this information in a way that felt trustworthy and in many cases will be life-saving.
We designed and tested prototypes that visualized probabilities and weather events in a way users could instantly grasp, no accurate mental model of forecasting required. Our goal was to reduce cognitive load and make weather search feel actionable, not abstract. In user testing, people reported higher trust and quicker decision-making, even in high-stakes scenarios.
