I’m finding myself looking back over the experience of the 2010 Census and drawing some lessons learned. If we’re smart as an organization, we’ll build the 2020 Census planning effort with these in mind.

Here they are:

Lesson 1: The multi-decade cost increase of the decennial census must be halted. We have looked at the cost trend of the last several decades, and we know that this trend is unsustainable. One way to do this is to make the next census as convenient as possible for people to answer. Another way we can do this is to maintain our collaboration with over 257,000 partnership organizations throughout the decade using modern communication tools instead of mounting an independent partnership operation later in the decade. Indeed, if the American public is fully engaged in the Census, it becomes a relatively inexpensive enterprise. This is true because the costs of following up houses that don’t respond are very high.

Lesson 2: Traditional non-response follow-up procedures are inefficient and costly. We must make the census more convenient to diverse groups in society. Thus, the second principle is that the 2020 Census will be a multiple-mode census, using mail, telephone, internet, face-to-face, and other modes as they emerge. We have to move beyond the mailback questionnaire and the personal interview. We need to ensure that the response options for the census reflect the communication platforms that people are using. For the remainder, requiring follow up, we need to squeeze every ounce of efficiency out of the costly knocking on doors.

Lesson 3: Systems development that requires first-use perfection must be abandoned. In the 2010 Census, given some of the Census Bureau’s challenges, we were developing critical systems weeks before their use; important weaknesses were discovered in the early days of production. The third principle is that we need end-to-end tests of production systems, which use all subsystems in the integrated form needed in the production phase. In early planning, we need to mitigate the risks that led to this need for first-use perfection. New system capabilities will be developed in an incremental and modular fashion so that users have a chance to test and evaluate mission critical systems well before they are deployed to production environments.

Lesson 4: Too few of the system and procedure developments of the 2010 Census were designed to have residual benefits to other Census Bureau data collections. All of the valued demographic and economic data we produce use operations similar to those of the decennial census. There were too few plans to utilize the systems built for the decennial to enhance the efficiency of our many other survey operations. As a result, the large investment benefited only the decennial program, not the bulk of the Census Bureau. Thus, the fourth principle is to develop systems within similar survey production environments within the Census Bureau, test and enhance them repeatedly over the decade, ramp them up for use in the 2020 Census, and then continue to use and enhance them in our ongoing surveys. We plan to use the American Community Survey as the chief test-bed for 2020 Census systems development.

Lesson 5: The short form and replacement questionnaire provided cost-saving benefits to the 2010 Census. With the ongoing American Community Survey, the Census Bureau is providing the needed socio-economic, housing, occupation, and commuting data that are important to local communities. Utilizing the decennial census to mainly focus on the key reapportionment and redistricting purposes was wise. Thus, the fifth principle is to build on the success of the reduced burden of the 2010 short form.

Lesson 6: A small number of large tests creates intolerable risks for the Census Bureau. The sixth principle is to mount many, small tests throughout the decade. We are committing to a faster cycling of ideas and testing, relying on a lot of small tests versus a small number of large, expensive tests. For example, although we cannot know the full features of the Internet option for the 2020 Census, we will have repeated tests of Internet census measurement throughout the decade, using platforms that will increasingly resemble those available in 2020.

Lesson 7: In my professional judgment, the voluntary partnerships with over 250,000 local and national organizations, coupled with a paid advertising campaign successfully improved awareness of the coming Census. We have empirical evidence of the increasing awareness of the Census as the partnership and advertising campaign rolled out. Thus, the seventh principle is that we would like to keep these relationships warm, to seek input from these groups to inform the 2020 campaign, and to return to them Census Bureau data useful to their organizations.

Lesson 8: Updating of the master list of addresses using the Postal Service list produced a stronger list in 2010. We want to build upon this success this decade. One critical component is the Geographic Support System Initiative (GSS), an integrated program to improve address coverage and provide continuous spatial feature updates, as well as enhanced quality assessment and measurement. As a part of these efforts, the Census Bureau hopes to initiate programs to work with partners, both state, tribal, and local governments, as well as private entities, to receive addresses and map feature information continuously throughout the decade, and to update our systems so that we can more fully leverage GPS and GIS technologies and resources.

We know of no single method of collecting census data that is optimal for all residents of the U.S. Some residents have told us they do not want people visiting their home; some residents told us to use information they have already provided in other government forms; some residents want to use the Internet at any time of the day on any device they favor to fit their lifestyle; and some want to speak by telephone to someone guaranteed to speak their language and understand their sub-culture. By making the census more convenient, we hope to reduce the size of very expensive field follow-up activities. This is the most expensive part of the data collection, and by concentrating our efforts there, we want to achieve a quality census at a lower cost per household.