Team members setting up for usability testing at the Harvard Science Centre
Finding flaws in websites is commonplace -- there are entire companies focused on it. But it is harder to find a real solution that meets the needs of people.
Our Harvard student team in a class on technology and innovation has the opportunity to do just that by helping design a better data portal for the U.S. Census Bureau.
Existing Prototype
After spending the last few months observing and interviewing people that use Census data, we started testing the Census Bureau’s existing pilot website, and proposing our own paper prototype alternatives.
The landing page of the US Census Bureau’s beta website - data.census.gov
Test
Our key objectives for usability testing were to understand the following:
what users thought of the data.census.gov site in general;
how well the search functions worked;
the kind of queries people ran on the Beta website; and
whether users would prefer some of the alternatives, including as a branched, guided search.
By enticing hungry college students with cookies, we were able to observe around 20 students try using data.census.gov to search for everything from income data in Cambridge to historical demographic data in Philadelphia. These students generally had STEM backgrounds and were comfortable working with data portals. Or so they thought.
Learn
Only around 30 percent of users tested could find the data they wanted. The search tool proved most problematic. Instead of making the search more intuitive, it often retrieved results vastly different to the search terms and prevented them from applying other filters.
Results of different queries:“Poverty in Cambridge” produced the correct table, while “Poverty, Cambridge” and “Poverty, Cambridge MA” produced different results. Users were unlikely to enter the exact search query the Census was after.
In contrast to census.gov, users said they liked the feel of the beta website but did not find it functional. Users were reluctant to click webpage links to Census factsheets, even when they were relevant, because they looked like online ads. They also asked for clearer prompts or guidance when trying to navigate the site.
That’s when we tested our prototypes for search tools, asking interviewees to select between a guided search, a search bar, and a drag and drop table. Using low-fidelity paper prototypes produced more honest feedback as there was less pretence or gloss getting in the way of core functionality.
Prototype
After their experience with a broken search function, users were overwhelmingly supportive of a simpler guided search which asked "I am looking for _(topic)__, within __(geography)___ , at __(further geography filter )___ level, broken down by __(additional criteria)____." The options in the drop-down menu would narrow based on the selections made in the prior step.
Paper prototypes of alternative search functions
We also observed that there was no way for the US Census Bureau to gain feedback from interviewees on their user experience. A simple “is this what you were looking for?” button after the search paired with capturing search queries would help the Census better understand user behavior and design accordingly.
Repeat
Our digital prototype for a guided search
Armed with our latest findings, we will continue to iterate on our guided search prototype to deliver a more user-friendly and functional search feature for Census data. We’ve already built a digital version of our initial paper prototype. Check-back in a few weeks to see how we progress.
Arjun Bisen, Ayush Chakravarty, Carissa Chen, Daniel Drabnik, Tony Thumpasery