Seattle Restaurant Data Visualization
My team created an information visualization in Tableau with the Seattle restaurant data. We hope this interactive webpage can help newcomers explore restaurants more engagingly in the new city.
Course: Human Centered Design and Engineering (Information Visualization)
Duration: 3 Months
Team: 4 Members (3 designers + 1 data scientist)
My Role: User research, Prototyping, Usability testing, Visual design
Tools: Tableau, Illustrator, Photoshop
Seattle has become the nation’s fastest-growing big city this decade. According to The Seattle Times, Seattle has gained 18.7 percent total newcomer population since 2010. Our team was interested in how newcomers explore new restaurants.
How might we help Seattle newcomers explore restaurants more engagingly?
We interviewed 3 Seattle newcomers to understand how they go about exploring new restaurants in Seattle. All of them mentioned Yelp, so we asked them to walk us through their restaurant finding process on Yelp app. We decomposed their process into 3 stages based on The Visual Information-seeking Manta:
Zoom & filter
Research Findings + Yelp Competitive Analysis
From our interview observations, we found that there were frictions in newcomers restaurant finding process with Yelp. The pain points are around the overview and zoom & filter stages.
Neighborhood, cuisine category, and restaurant quality are the most important factors to newcomers.
Business hours, address, and price range are detailed information that becomes important after user selects a restaurant.
Our design focus on providing newcomers overview and
zoom & filter features that are missing from Yelp.
We got our Seattle restaurant dataset from our instructor Ray Hong, that consisted listings scraped off from Yelp and Google. The large dataset contained roughly 6000 records spread over 41 attributes. From previous research, we learned which information is important to newcomers, and removed irrelevant data. During the data cleaning process, we filtered out irrelevant cities, pre-defined restaurant categories, edited open and close time formats.
Brainstorming - Braiding Method
Explored various types of visualizations
Tested different encoding methods for restaurant category, rating, number of rating, and opening time
Through down-selection we narrowed down our concepts and created paper prototypes for testing. Our goal is to find the most effective combination and arrangement. By testing with our classmates, we learned that:
Information architecture (IA) is important. A clear vision IA can help us organize and structure our graphs and charts so viewers can follow our flow.
Keep encoding consistent for all graphs to avoid confusion. For example, users were confused that we used color encoding for both ratings and cuisine categories.
Mid-fidelity Prototyping & Usability Testing
We used the RITE Method (Rapid Iterative Testing and Evaluation) for usability testing with 5 participants, and did quick iterations after each round. All participants found the first section of our design the most engaging, because they could compare the number of a certain type of restaurants across neighborhoods and learn about their location at the same time.
We identified and fixed 3 major usability issues:
Refine titles, headers, and labels.
Incorporate the number of rating data to make the rating data more useful and convincing. Participant 2: “When I sort by ratings, I look at the reviews, and the number of reviews validates the rating.”
Rearranging the overall user flow. In Iteration 1, the cuisine filter controls the entire visualization. Therefore, when looking at sections below the first fold, users would need to scroll back and forth to adjust the filter and view results. This prolonged the response time in the information seeking process. Hence, we decided to compact filters and all related charts into one screen, so users can get immediate feedback when adjusting the filters.
Designers are no copywriters, but from this project I learned that proper labeling is important in guiding users. I hope to explore better use of language, and more effective visual treatments.
Unlike our normal design process, our dataset and Tableau added limitations to our design. The restaurant data in Tableau is static, but I hope to find ways to make our visualization live.
In the next version, I would like to incorporate direct links to Yelp in the last stage of user’s restaurant exploration process. Hopefully newcomers can to get enough details and head straight for the food!
<Please refresh this page if Explore Restaurants in Seattle doesn’t show up!>