Welcome to the Listening Post!

The Case For and Against a Census instead of a Survey

Typically, when consulting with large organizations about enterprise feedback management, one of the big issues I encounter is that they have too many surveys done with too many of their customers. For such organizations, using a centralized panel system with random sampling dramatically reduces the frequency with which any individual customer is surveyed.

I want to be clear, though, that when I am criticizing a census I am really criticizing needless attempts at a census. For me, the canonical example will always be the automobile manufacturer that started subjecting their 8,000 dealers to web surveys about all manner of issues, from the strategic to the tactical, simply because they could. Just like the tragedy of the commons, though, this wanton surveying ruined a good thing, as dealers stopped participating.

Although I describe this approach as a "census", it is really an attempt at a census, selecting everyone in the target population to participate but not reaching many of them. The 2000 United States Census surveyed 99.6% of the population, while census attempts by a corporation typically get much lower participation-as low as 20% for firms that overuse the technique.

Advantages of a Census

• Every customer or employee is invited to participate. This is important for annual measures of customer satisfaction and employee well-being, which are opportunities for everyone to provide their feedback.
• Maximizes exposure to negative feedback. By contacting everyone, and by setting up email triggers to alert you to negative responses, you will have the opportunity to intervene with more individuals to attempt to improve their satisfaction.
• Greater confidence levels in the results. For target populations with fewer than 1,000 individuals, the recommended sample size is so large as a proportion that you must typically attempt a census to gain enough responses to have a high degree of statistical confidence that your results represent the target population.
• Random sampling insufficient when assessing employees. To be able to cross-tabulate results by employee with a high degree of confidence typically means that you must attempt a census rather than randomly sample.

Disadvantages of a Census

• Limits opportunities for other surveys. If everyone in the target population has just been invited to participate in a survey, they are unlikely to respond to an immediate subsequent request to take an additional survey. We once had a client that realized it needed to do some additional unrelated research, right after it had emailed its house list of 60,000 for a tactical survey.
• Leads to declining response rates. A survey important enough to a large organization for it to be sent as a census is typically longer than average. Inviting everyone to take a long survey makes them less likely to participate in the next survey.

Sometimes you simply must do a census instead of a survey. Just make sure you are doing so because you have to. And, if you're in the States, remember to be kind to the census worker when he or she comes to your door next year.