The document discusses the development of a user interface for the 2012 Economic Census, which collects data from over 4 million U.S. businesses across different industries and modes of collection. It details the challenges of designing a multi-mode interface that is consistent across paper, software, and web-based surveys. Usability testing uncovered issues with navigation, scrolling long questions, and response quality controls that could be improved for future censuses by better utilizing automated features of the web interface.
Application UI Design with Large Data Sets (Cathy Lu)
Developing User Interface Census Data Collection
1. Developing a User Interface for Large-
scale, Multi-mode, Survey Data Collection
Jennifer Beck & Elizabeth Sinclair
U.S. Census Bureau
User Focus 2012
October 19, 2012
Saturday, January 5, 2013
2. Economic Census
• A measure of the health of U.S. businesses and the American economy
every five years
• Collects detailed information on revenue and business performance
• Goes out to more than 4 million businesses across industries:
1. Multi-unit businesses = businesses with multiple locations
2. Single-unit businesses = businesses with a single location
• Businesses receive a combination of forms tailored to their specific business
activities
– more than 800 different survey forms
• The number of form combinations creates unique multi-mode data collection
challenges
Saturday, January 5, 2013
3. Economic Census reporting options
• The 2012 Economic Census has three different
modes for data collection:
1. Paper forms
2. Surveyor – a software application
3. Internet
Saturday, January 5, 2013
7. 2012 Economic Census
Internet option
• Only available to single-unit businesses:
– A pilot for future Economic Censuses
– Alternative to Surveyor software
• Respondents log into a Web interface through
Census website
• Respondents fill out and submit all information
online
Saturday, January 5, 2013
8. Multi-mode data collection
• Offering multiple modes of data collection has
benefits:
– Can save costs
– Good customer service
• Designing for multiple modes is challenging:
– Mode consistency
– Mode capabilities and constraints
Saturday, January 5, 2013
9. Multi-mode challenges in the
Economic Census
• The paper forms, Surveyor (software program) and the Web
interface “share” a creation database
• The database contains “images” that will make up the form
pages for all three modes
• Because the systems “share” pages, the pages have to work
for all three modes:
– The “shared” design creates limitations and usability issues,
especially for Surveyor and the Web
– Changes made for the Web also make the same changes to
Surveyor
Saturday, January 5, 2013
10. 2012 Economic Census Internet
option: Usability testing goals
1. Evaluate the overall function and performance
of the online interface
2. Assess respondents’ reactions to the interface
Saturday, January 5, 2013
11. Participants
• 17 respondents from small businesses
• Diverse job titles: accountants, office managers,
business owners
• From three different industries:
– Retail/Service companies
– Manufacturing companies
– Construction companies
Saturday, January 5, 2013
12. Methodology
Where:
• Met at respondents’ place of business
What:
• Respondents used their own computers (or iPad)
How:
• Respondents filled out the census forms
• Completed tasks to test features of the interface
• Used standard “think-aloud” protocol
Saturday, January 5, 2013
13. Usability findings
• The Web interface performed well during the
testing
• We uncovered some key findings about the
interface:
1. Questionnaire navigation
2. Scrolling
3. Response quality control
Saturday, January 5, 2013
14. Questionnaire navigation
• Survey navigation:
– Forward and backward paging
– Questionnaire navigation – response path through
a survey
• Skip patterns:
– Questionnaire navigation path instructions for
respondents
– Help respondents locate and navigate to the
relevant form sections
Saturday, January 5, 2013
17. Using skip patterns
• Skip patterns can have a mode-specific advantage:
– Paper:
• Require reading, understanding, and following skip instruction
navigation
• Respondents do not always follow skip patterns on paper forms
well
• Can create response errors
– On the Web:
• Skip patterns can be automated
• Directly navigate user down the correct questionnaire navigation
path
• Can reduce response error
Saturday, January 5, 2013
18. Problems with skip patterns:
Usability findings
• The Web Interface (and Surveyor) had no automated skip patterns
– Questions still contained skip instructions
– Respondents had to page through irrelevant sections
– Due to timing and programming complexity
• Respondents were confused as to how to navigate to the appropriate
section of the questionnaire
– Expected forward navigation to take them to the appropriate screen
– Expected to be able to click “Remarks” to navigate to that section
• Lack of automation was especially problematic for the Web interface
– Web interface had no navigation tree
– Surveyor software has a navigation tree
Saturday, January 5, 2013
20. Fixing problems with skip patterns
The recommendations:
1. Program automated skip patterns to navigate the
respondent to the next questions (and remove skip instructions)
OR
2. Include a navigation tree
The resolutions:
• Skip instruction stayed with the question
– Shared image database made changing question layout and
presentation not possible
• Programming complexities and time constraints prevented
adding a navigation tree
Saturday, January 5, 2013
21. Scrolling
• Economic Census has a number of “long”
questions
– Tend to be detailed business activities and cumulative
calculations
– Do not fit on a single “page”
• Long questions require mode-specific solutions:
– Paper:
• Pagination
– Web interface (and Surveyor):
• Paging
• Scrolling
Saturday, January 5, 2013
63. Choosing between paging and scrolling
• Paging creates mode-consistency problems:
– Paper:
• Easy to “flip” back and forth between the question sections
– Web interface (and Surveyor):
• Requires forward and backward navigation to go back and
forth between the question sections
• Breaking up the item across pages can create response
errors
• Scrolling “preserves” mode consistency
– The 2012 Economic Census Web interface employed scrolling
Saturday, January 5, 2013
64. Problems Scrolling: Usability findings
• Scrolling created mode-specific problems:
1. Long items were “intimidating” to respondents
2. Length attenuated the benefits of scrolling over
paging
• Scrolling was especially problematic for the
Web interface
– Timing out:
• Long items take longer to fill out
• Data save linked to navigating to the next page
• Respondents could lose data
Saturday, January 5, 2013
65. Fixing problems with scrolling
The recommendations:
1. Break up the items over multiple pages
2. (Make changes to the question to maintain mode consistency)
3. Include a navigation tree
The resolutions:
• Questions maintained scrolling design:
– Shared image database made changing question layout and
presentation not possible
• Not enough time to re-program and add a navigation tree
• Time out increased to 50 minutes
• Data saved on time out
Saturday, January 5, 2013
66. Response Quality Control
• Data quality is a big concern:
– Follow-up is expensive
– Post-collection editing is expensive and time consuming
• Respondents find data edits helpful
– Provides a “check” against errors
– Communicates the importance of the information
• Controlling for data quality is mode-specific advantage:
– Paper forms offer no “built-in” data quality control
– The Web interface allows for “real time data editing”
– Can link responding to satisfying quality requirements
• The web interface had edit triggers for individual items and for of the
Saturday, January 5, 2013
67. Item-specific data edits: Errors
Errors
= items respondents
must fill out in order to
submit their data to
Census
Saturday, January 5, 2013
68. Item-specific data edits: Warnings
Warnings
= items we would like
respondents to fill out *
* Respondents can still
submit their data to
Census with out filling
them out
Saturday, January 5, 2013
69. Overall data edits: Form Review
Form Review
= indicated problem
sections with color
coding
• One-click navigation to
problem section
Saturday, January 5, 2013
70. Response quality control:
Usability findings
• Respondents liked the quality control features:
– Item-specific data edits:
• The edit messages were clear and helpful
• Liked having the “extra check”
– Form Review:
• Really liked the layout of the form review
• Liked the ease of navigating to problem sections
• Capabilities were examples of successful design features
• Provide in quality-control measure not possible in a paper
mode
Saturday, January 5, 2013
71. Challenges to designing the 2012
Economic Census Web interface
• The Economic Census is a large data-collection
undertaking
• Presents challenges for how to maintain consistency
between modes and optimize mode advantages
• More work needs to be done to refine web data
collection for future Economic Censuses
Saturday, January 5, 2013
72. Contact Information
Jen Beck:
jennifer.l.beck@census.gov
Elizabeth Sinclair:
elizabeth.sinclair@census.gov
72
Saturday, January 5, 2013