IT Vulnerabilities Identified during Census’ 2018 End-to-End Test

Each major show likely follows a string of practices or dress rehearsals. Likewise, each decennial count follows a practice run of operations and in the case of the 2020 Census, it is the ongoing 2018 End-to-End Test that is helping identify the improvements needed before show time. According to the latest test update, the End-to-End Test in Providence County, RI seeks to “test and validate major operational threads, procedures, systems, and infrastructure (IT and non-IT) together to ensure proper integration and conformance with functional and nonfunctional requirements.” So far, the test has revealed to the Census Bureau and GAO that several IT-related enhancements are needed before the actual survey.

The GAO recently released a report that studies the results of the completed in-field canvassing procedure done in the 2018 test. The watchdog agency found discrepancies in both the workload and productivity levels of canvassing operations. For example, the report cites instances where address canvassers, referred to as listers, duplicated workloard in scenarios when work was reassigned. The software the listers use while canvassing is unable send real-time notification of caseload assignments. Moreover, the Listing and Mapping Application (LiMA) software that the agency planned to use to show updated address file and maps was not developed due to budgetary restrictions.

 Moreover, the test revealed that 11 out of 330 issued laptop computers to listers did not properly transmit address and map data collected for a number of locations, causing a recanvassing effort. Once data is collected into the laptops, it is encrypted and then transmitted to the Census Bureau’s data processing center upon return to the office. The GAO learned that Bureau officials were aware of problems with data transfers using the LiMA software and “told us that the software problem has been persistent across other census surveys that use LiMA and they are not certain it will be fixed.” Further, Bureau officials stated that alert reports unknowingly failed to notify staff that data was not transmitted. To resolve this, the Census Bureau plans create two new alert reports to monitor the transfer of data.

Issues were also discovered in data-driven tools used by field supervisors and managers. In one system, alerts were sent incorrectly to managers to follow-up with listers about a particular issue that did not exist (i.e. completing a timecard) or was already resolved. More often than not, managers and supervisors were dismissing alerts without properly investigating the notices due to how unreliable the alerts became. Furthermore, the Unified Tracking System (UTS) is used by the agency to combine data from a variety of systems so that users can run or create reports for monitoring of operations. Unfortunately, GAO investigators found that information in the UTS was inaccurate during canvassing operations, causing managers to rely more on the systems that feed data into UTS rather than UTS itself.

Finally, listers at several sites reported connectivity issues in transmitting their work assignments. Each laptop in the 2018 test used only two broadband internet service providers, which likely contributed to the connection issues. The Census Bureau intends to analyze network coverage for all regions of the 2020 Census to ensure proper connectivity and which carrier would be best in limited coverage regions. In fact, as reported by Federal Computer Week, Al Fontenot, associate director of the 2020 Census, stated that the bureau intends to add more carriers under its device-as-a-service contract with CDW-G and AT&T.

In all, the GAO provided seven recommendations to the Census Bureau to improve operations for the 2020 Census:

1. Improve workload estimates for in-field address canvassing to ensure staffing and laptops are procured accurately.
2. Finalize procedures for reassigning blocks to prevent duplicated work.
3. Finalize backup instructions for the secure transmission of data when mobile carriers are unavailable.
4. Finalize plans for alternate training locations in areas where internet access is low.
5. Fix the software problem that prevents the transmission of data or at the least, address why triggered alert reports were not addressed.
6. Develop a plan to make managerial alerts more useful.
7. Finalize UTS requirements for canvassing reporting to ensure data used by managers is in real-time and accurate.