What now? That's the hard question. When an IT project is in trouble, it's easy to ask what went wrong and who's to blame. Easy and popular. And fun, if you're not on the hot seat. But what to do to save the project? That's harder — a lot harder. Especially when, as with the US Census Bureau's "paperless census" project, it can't be killed and can't be delayed.
This month, the US Government Accountability Office reported that the bureau's plan to use handheld computers for much of the 2010 census is in trouble. The GAO noted cost overruns and project management issues that it has pointed out for years.
But the big problem is a single device: the custom handheld designed to be used by 525,000 "enumerators" temporarily hired to track down the estimated 100 million Americans who won't return their census forms.
The device, made by HTC, is 15cm long and weighs about 0.4kg. It contains a GPS locator, maps, wi-fi, a cellular device to transmit encrypted census data, an iPhone-size touch screen, a fingerprint sensor for security and an extra-large battery to run it all.
In short, it's big, heavy and stuffed with gadgetry — not exactly what you'd choose for the retirees who will make up the majority of those half-million enumerators.
And who devised and championed this wrongheaded, user-unfriendly "solution"? A bunch of IT dweebs with no clue about the real world of census-taking, right?
Nope. It was Jay Waite, the career bureaucrat who ran the 2000 census. Waite came up with the idea of using handhelds in 2001, hoping the devices would make the 2010 census faster, less costly, more complete and more accurate.
Then what went wrong? Waite's original idea of using off-the-shelf handhelds fromHewlett-Packard didn't pan out. In early 2006, Harris Corp won the contract to create the custom device. After that, scope creep, poor communication and iffy project management took their toll.
So did ordinary bugs. Data uploaded too slowly. Too-big data files wouldn't upload at all. The handheld's security software locked users out for 15 minutes when their fingerprints weren't recognised. Some users quit during the test last year — too complicated, they said. Pretty much all the testers had trouble making the devices work.
But amazingy, as recently as January, there was no contingency plan in case the handhelds just couldn't do the job. Census Director C Louis Kincannon repeatedly insisted that the handhelds would work, and that's all there was to it.
No wonder the GAO described the project as "high risk".
That's what went wrong. Who's to blame? Kincannon. Waite. Harris Corp. The Census Bureau's IT staff. Take your pick — they all contributed.
So much for the easy questions. What now?
In January, Kincannon retired. He was replaced by Steve Murdock, former head of statistics for the state of Texas. One of the first things to land on Murdock's desk was a special team's report on the paperless-census project. Its conclusion: Things were worse than anyone thought. Murdock launched a task force to figure out what to do. Within a month, he had a set of contingency options.
The upshot: The fancy custom handhelds might work. But if they don't, the Census Bureau will use paper instead.
That's an ugly, low-tech kludge. A paperless census is visionary. It's the future. Let's hope those handhelds work.
But if they don't, let's be glad that, when Census Day 2010 arrives, we won't all be asking, "What now?"