DETROIT -- In 14 months, a General Motors engineer went from telling colleagues that improving the Chevrolet Cobalt's ignition switch was "close to impossible" to signing off on the very fix he had shot down, documents show.
Why he made such a surprising about-face -- and how it could occur so inconspicuously that virtually no one else at GM knew about it until last fall -- is among the biggest questions that those investigating the automaker are working to answer.
It happened against a backdrop of extraordinary upheaval at GM and Delphi Automotive, the supplier that made the switch in Mexico. Analysts at the time were starting to wonder aloud whether bankruptcy could be in GM's future, and Delphi began a messy, four-year trip through its own bankruptcy in October 2005.
Both companies were bleeding billions of dollars and scrambling to contain the damage by cutting jobs and slashing costs wherever possible. But union contracts compelled GM to keep plants churning out unprofitable cars such as the Cobalt because it had to pay workers whether they were productive or not.
"That's impacting everything you do -- how much you spend, how much energy or resources you throw at solving a problem," said Karl Brauer, senior director of insights at Kelley Blue Book.
"You're producing these small cars basically to give workers something to do that you're going to have to pay either way," Brauer said. "If that's the atmosphere, then it's easy to see how any cost related to those vehicles becomes 'No, we're not doing that.'"
GM CEO Mary Barra told Congress last month that the automaker had too much of a "cost culture" at the time. But the constant announcements of layoffs and restructuring plans, beginning in 2005, also created an environment in which employees couldn't help but worry about their jobs and the company's future.
In early 2005, GM, which had posted a long streak of profits on demand for pickups and SUVs earlier in the decade, began quietly racking up what became its biggest quarterly loss in 13 years: $1.1 billion. Amid the gloom suddenly enveloping GM, engineers were looking into a report from the Cobalt's program engineering manager, Gary Altman, that he had accidentally turned the car off by merely bumping the ignition with his knee.
They quickly identified that the ignition switch needed more torque to prevent the key from slipping out of position. But Ray DeGiorgio, who had overseen the part's design since 1999, rejected the idea.
"One would immediately think that changing/increasing the ignition switch torque effort would be a good solution," Blendi Sullaj, a contract engineer assigned to the GM Technical Center north of Detroit, wrote Feb. 1, 2005, in a document released last month. "After talking to Ray DeGiorgio, I found out that it is close to impossible to modify the present ignition switch. The switch itself is very fragile and doing any further changes will lead to mechanical and/or electrical problems."
On March 9, Altman ordered the inquiry into the problem he had encountered closed "with no action," Sullaj wrote. Altman concluded that all of the proposed changes would cost too much, would take too long to implement and wouldn't fully resolve the problem, Sullaj wrote.
In a 2013 deposition, Altman testified that proposed changes weren't effective enough.
"If this was a 100 percent fix, GM would have spent the money?" he was asked.
"Yes. Oh, yes," he replied.
But at the time, Sullaj recorded Altman's decision this way: "None of the solutions represents an acceptable business case."
One week after the inquiry was closed, GM stunned Wall Street by cutting its profit estimate for 2005 in half -- even the reduced projection turned out to be wildly optimistic, as the company ended up losing $10.6 billion on the year -- and revealing that it was deep in the red during an ugly first quarter. Panicked investors sent GM shares plunging to their lowest level since 1992.
GM cut 2,000 salaried jobs in 2005 and announced a plan to close 12 plants and eliminate 30,000 hourly workers within three years.
In May 2005, engineers re-examined the matter, after employees in GM's "brand quality" department noticed that the company was repurchasing an unusual number of problematic Cobalts from customers.
By the end of May, GM had opened eight claim files based on customer complaints about their Cobalts repeatedly stalling. It ultimately agreed to buy back five of the cars, and a sixth customer was given a replacement car by his dealer, according to a 2013 deposition of Victor Hakim, a manager in GM's field-performance assessment department. Yet in June 2005, GM told Cleveland's newspaper, The Plain Dealer, that it had received "no consumer complaints" about the issue.
Meanwhile, DeGiorgio began seeking more data on the switch, despite having insisted earlier that it could not be modified.
"Ray is requesting this information," a Delphi engineer, Jack Coniff, wrote in a June 14, 2005, e-mail to two coworkers. "Cobalt is blowing up in their face in regards to turning the car off with the driver's knee."
In September, GM approved a proposal by David Trush, the GM engineer responsible for the Cobalt's key cylinder, to redesign the car's key so that dangling items could not make it rotate as easily. "This will assist in limiting the issue but will not completely eliminate it," wrote Craig St. Pierre, a GM engineer designated as a "champion" for the inquiry.
But GM later reversed course. "Canceled," St. Pierre wrote in March 2006. "Business case not supported."
Elsewhere within GM, DeGiorgio had been working with Delphi on changing the switch design. It's unclear, though, who else at GM was aware of it, as none of the publicly released e-mails between DeGiorgio and Delphi engineers show that they were copied to any other GM employees.
DeGiorgio requested and was given test results from 2002, when he had first approved the switch for production. Documents show the torque fell short of GM's specifications from the beginning. DeGiorgio "needs a copy of the test results," Coniff wrote in a Nov. 16, 2005, e-mail. "Sounds like there might have been an issue somewhere."
By January 2006, samples of a new switch design were circulating between Delphi and GM. Delphi engineer Arturo Alcala, shipped a box of 24 ignition switch samples to DeGiorgio, who wanted them for in-vehicle testing, according to e-mails between them. The switches had been modified using "the stronger Catera detent spring-plunger," that they had discussed in December, Alcala wrote at the time.
A Delphi change order dated the week after DeGiorgio received the switches lists the Catera spring and plunger as items that GM requested so the switch would meet its torque specifications. After Delphi finished testing the new switch, DeGiorgio faxed a GM form known as a "commodity validation sign-off" approving the redesign back to Delphi on April 25, though he dated it the 26th.
Delphi, in a document dated May 2006, said DeGiorgio had agreed to make the change without assigning a new GM part number. That decision violates GM policy and "Engineering 101," Barra told a Senate panel investigating the recall April 2.
DeGiorgio, in a deposition last year, denied any knowledge of the change. GM says it had no official record of it until October 2013, when Delphi provided the form signed by DeGiorgio. GM has suspended him and Altman with pay, pending the results of an internal investigation.