Census Screwups--Further Thought

From a Government Executive article on the failure of Census project to use handheld computers for 2010 census.

"[the contractor rep] ascribed the program's problems to continuously changing requirements on the part of the Census Bureau, which delivered 400 new or refined requirements to Harris in January, two-thirds of the way through the development process. "We were informed that there was concern that the requirements given to us for nonresponse follow-up may not be complete," she said. "We have designed what the Census Bureau asked for, but what they have asked for may not be what they need."

Census Director Steven Murdock admitted that the bureau could have done more to ensure that the contract was executed on time. "Clearly, we didn't do everything we should," he said. He admitted that the bureau didn't scope the requirements for the project fully or effectively communicate with the contractor as much as necessary.

The decision to return to paper was driven largely by the bureau's comfort with the old way of doing things. "Census Bureau officials are more comfortable addressing any such challenges through the paper process, with which they are more familiar," Janey said.

Murdock said reverting to paper would provide more flexibility and minimize risk, and he cited the bureau's knowledge and experience with the paper-based system as a justification for the decision.

The changing user requirements is a familiar cause of problems (how many times I heard that from our IT type--you can't change these requirements!). Bad requirements means they didn't have the right people involved in specifying them and hadn't trained them in what's involved in IT. Which probably relates to the culture problems in the last two paragraphs. A familiar story, as with the FBI.

What I don't understand is two things:

• Lehrer News Hour had a piece on the professional shoppers for Commerce (checking prices of products for computing cost of living index). The woman was merrily punching away on her laptop computer entering the data. So obviously Commerce has experience with hand-held applications that work on a distributed basis.
  
• Given my faith in the 80/20 rule (80 percent of your work comes from 20 percent of the cases)--has census and the contractor ever considered a rather simple handheld application that would do most of the cases, with a fall-back to paper for the complex ones? Eat the elephant one bite at a time??