A clue to MNsure's problems found in a similar software precursor

Before MNsure, the state's new online health insurance marketplace, there was HealthMatch, an expensive, problem-plagued software project that foreshadowed MNsure's trouble a decade ago.

"It was a complete failure," said longtime state Rep. Tom Huntley, a DFLer from Duluth. "We hired a company to manage it and set it up...and nothing ever happened."

The HealthMatch project dragged on for five years, and consumed more than $20 million of taxpayer money to develop software that was never usable. The Office of Legislative Auditor Jim Nobles monitored the HealthMatch project. Nobles said the state got nothing out of the project except valuable experience.

"The one thing we could have gotten out of [HealthMatch] are lessons. And so when I look at MNsure I just don't necessarily see that people are applying lessons of what not to do."

One of MNsure's central functions is to determine if someone qualifies for MinnesotaCare, a subsidized health insurance program. MNsure's software is supposed to verify a person's income and review eligibility requirements nearly instantaneously. But the code has been unable to do so. MNsure's software also has inaccurately calculated federal tax subsidies for people buying commercial insurance.

In December, Gov. Mark Dayton blasted IBM for delivering a defective product that did not properly determine eligibility.

It was the second time in less than a decade that a computerized eligibility system had been a big expensive flop.

Back in 2003, the Pawlenty administration took up an idea conceived during the Ventura administration. The state contracted with SSi North America to produce, "HealthMatch." The computerized system was supposed to apply digital precision to the eligibility process and bring down a high error rate on MinnesotaCare applications. Minnesota's Department of Human Services (DHS), which administered the programs, had been using two dated mainframe computer systems to determine eligibility, neither of which was specifically designed for that purpose. The process was arduous and inefficient. A 2006 DHS audit found enrollment errors in 10 to 14 percent of cases and incorrect income determinations in over 30 percent of cases.

But early on, the HealthMatch project started to spin out of control. Numerous reviews of the debacle indicate:

The HealthMatch project dragged on for five years, and consumed more than $20 million of taxpayer money to develop software that was never usable.

• The state expanded the program's scope twice.

• Project leaders underestimated the complexity of the work.

• The state's requirements for the software weren't clear.

• A sequence of corporate mergers roiled the contractor, which had significant quality problems and senior management turnover.

There was another big problem, said David Kirchner, who did an analysis of the Health Match project for the Legislative Auditor's office in 2007. Deadlines were wildly optimistic because the wrong people set them.

"There was a sense that it needed to get done, and it needed to get done within a specific period of time," Kirchner said, "but then those decisions were made by people at a very senior level who were not involved in building the system."

As deadlines passed, communications deteriorated, with both sides blaming each other for delays.

On Jan. 23, 2007, then-assistant commissioner of health Care at DHS, Brian Osberg, complained to the contractor's president Rob Marchant that key company staff members were "inaccessible at times when their expertise was critical...and staff absence from meetings has delayed numerous decision-making processes." Osberg told Marchant to instruct staff to, "check their voice mail and email regularly."

Here's that letter:

Nearly three weeks later, Marchant accused Osberg's management team of "obfuscating a number of issues in the hope of avoiding culpability in the project delays." He added that "the state has clearly been under-resourced in critical design areas and is over a year behind in its last commitment..."

Here's that letter:

By March 2008, Human Services Commissioner Cal Ludeman had had enough.

"I met the executives of the company and some of the technology folks; I conferred with the chief information officer of the state and I made a decision to just simply cancel the project," Ludeman said.

But that wasn't the end. The company sued for breach of contract. The state agreed to pay an additional $7.25 million to settle. Total cost for HealthMatch, about $30 million.

When it comes to troubled government I.T. projects software often gets the blame said consultant Michael Krigsman, who analyzes why I.T. projects fail and how to prevent them from going wrong.

"It's easier to talk about technology--that the software didn't work, the vendor delivered bad software and therefore if we get a new vendor then the problem is going to be solved," said Krigsman.

But the problem often lies beyond the coding, in "the human realm, as opposed to the technical realm," he said. "The human realm is where the solution lies for most of these types of project failures, but human relationships are messy."

But there's a major difference between MNsure and HealthMatch, said former state Sen. Linda Berglin, long a central player in Minnesota health policy. Unlike Mnsure, HealthMatch never got off the ground, she said.

"So the things we're seeing now with MNsure are things that we can see need to be fixed. But it isn't like the whole system isn't working at all," Berglin said.

Still, former human services commissioner Cal Ludeman said the lesson to draw is not how to manage a big software project. He said the lesson is, don't try to do electronic income verification--it's just too complicated. "Like trying to solve a Rubik's cube while it is in flight," he said.

The problem is both state and federal law require electronic income verification.