How Intel cut internal IT costs
- 09 December, 2009 06:44
IT managers waiting around for that special day when they get budget relief and use it to fix all their problems are setting themselves up for disappointment, according to internal IT at Intel Corp.
"It's not going to get better," said Ron Miller, business relationship manager for Intel IT. "There's not going to be that magic year that comes along where all of a sudden you get budget relief and you can use that one year to fix all the problems you've had over many years."
What you need to do is make that "magic pill" happen for you, he suggested. "You need to spend some time innovating and looking at how to do things new because you can't continue to do them the same way that's worked for you successfully in the past," he said.
Everything changes -- the technology, use models, applications on top of the technology -- and IT managers must change as well, Miller argued. "You're going to have to find some quick, productive ways of doing that that don't involve waiting for budget relief," he said.
Speaking at the Intel Premier IT Professional conference in Toronto, Miller provided some examples of what Intel has done to cut its own internal IT costs. He also presented an IT Business Value model that helps demonstrate ROI for IT projects.
What worked for Intel might not work for everybody, noted Miller. But all IT managers should learn how to measure, manage and base their analytical decisions on the data, he suggested. "It takes the politics out of the game," he said.
1995 to 1998: Reducing the cost of PC ownership and maintenance
From 1995 to 1998, Intel reduced the costs of maintaining PCs by doing three things, said Miller. First, it consolidated all purchases for equipment to be handled through IT. After IT became the central point to purchase PCs in the environment, Intel consolidated its vendors. Then the company consolidated to a single, unified 24/7 worldwide IT help and support organization.
1998: Shifting from desktops to notebooks
Intel made the decision to shift from desktops to notebooks in 1998, said Miller. "We used to be 80 per cent desktops (and) 20 per cent notebooks. We made the flip. Everybody who wanted a notebook could have a notebook with the refreshes going on," he said.
The decision was based on data that broke out the total costs for desktops and notebooks. Although the cost of a notebook was $800 more than the cost of a desktop, IT knew that employees with notebooks were also more productive, he explained.
IT gave 300 desktop users notebooks and monitored their usage, figuring that if the employees put in one extra hour of work each week, the productivity would offset the hardware costs. "What we found was they did an additional three to eight hours of work based on job type and profile," he said.
With the numbers and the methodology behind them, IT could easily justify giving employees notebooks and the project was immediately adopted after a presentation to the executive staff, he explained.
The ROI increases when you go beyond measuring performance and take into account unexpected events and natural disasters that may shut down physical office locations, noted Miller.
The 2003 blackout in Toronto didn't affect the company, he said. Neither did a five-day ice storm in Portland or a SARS scare that shut down the Hong Kong office in 2005. With notebooks, employees can work anywhere and travel to other locations if necessary, he explained. Intel also ships notebooks to employees via FedEx.
Intel didn't need to declare a disaster or key up the disaster recovery centre when a six-inch coldwater pipe break dumped 7,000 gallons of water through three floors of a California building in the middle of the night. The flood affected over 80,000 square feet of office space that displaced 450 users, but in terms of hardware, Intel only lost 12 PCs, he said. Six were at desks and six were in a lab because everyone else had taken their machines with them.
2000: Buying high-end PC systems
Intel was buying low-end PC systems in 2000 based on advice from consultants, who said the company was spending too much on PCs, said Miller. But in less than 18 months, all the systems had to be replaced because the machines could not run the new operating system and application base. "We had to forklift and throw out 20,000 clients prematurely in our environment. It was a bad mistake that we made," he said.
The company now buys high -- but not too high, he pointed out. Intel buys newly introduced machines once they become commercially available around the world. If an employee is in Hong Kong next week and needs the IT group for a repair, the replacement parts have to be available, Miller explained.
Intel extended its refresh cycles to make up for the increases in hardware costs, but IT noticed a large increase in exception requests, high levels of customer dissatisfaction and additional calls coming into the help desk from those with four to five-year-old machines. Intel estimated a hard disk failure rate of about five per cent and then made the decision to switch over to SSDs, he said.
2004: Determining optimal PC refresh cycles
"Now we bought high, we bought low, we kept machines in our environment for up to five years and we refreshed them as early as one year. We had all the hard data we needed to do the perfect analysis for determining our optimal PC refresh cycles," said Miller.
Taking into consideration all the costs associated with a machine -- purchase price, disposal, failure rates, help desk support, etc. -- IT found that keeping a PC in its environment for one year cost $1,023. Keeping the PC in that same environment for two years brought the cost down to $863. The cost decreased further to $803 each year if the PC was held for three years.
Intel performed the calculation again in 2009 to make sure the company was still doing everything right, said Miller. While the numbers have changed with the costs of PCs going down, Intel found that a three-year refresh cycle is still the answer, he said.
2005: Automating the help desk
There is no reason IT should manually run things like a disk defragmenter, checking to see if a machine has been backed up, how memory is functioning on a system, whether a machine has the latest patches or if a virus scan is running correctly, said Miller.
Intel packaged up its help desk services in 2005 and now takes a self-service approach. Users click on a PC health check button for a green, yellow or red light indicating their PC's status with a description of the details. The process takes about 30 seconds -- not the 20 minutes previously required for the help desk to run these processes manually in order to figure out what was wrong with the machines, noted Miller.
After running through the process, users can click on another button that places a call to the help desk, he said. All the information about the user's machine goes right into that call, so the help desk immediately understands the status of the machine, he said.
Intel has calculated that this has saved $3 million, because 70 per cent of the time people are able to solve their own problems without calling the help desk -- and when they do call the help desk, problems are resolved in five minutes, he said.
2009: Refreshing servers
Intel's refresh cycle for servers is four years. IT was asked by corporate executive staff to do without server refreshes this year, said Miller. "We did the calculation ... we were able to go back and say it would cost us $19 million more for us not to refresh this year than if we would refresh," he said.
Intel is sharing Excel spreadsheets that calculate business value, a rudimentary portfolio management tool and other documents online at no cost to those interested in learning more.
You eventually reach an inflection point, said Miller. "The inflection point is where things have changed so much that if you keep doing them the same, it's just going to get worse. You've got to do some innovation," he said.