Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vivamus convallis sem tellus, vitae egestas felis vestibule ut.

Error message details.

Reuse Permissions

Request permission to republish or redistribute SHRM content and materials.

Big Bang or Phased Rollout?

HR Magazine, January 2004A pair of major USDA implementations illustrates the benefits of each.

In 2000, the U.S. Department of Agriculture (USDA) rolled out PeopleSoft 7, a client-server HR management system (HRMS), in eight phases over nine months. When the phased rollout was completed, 552 HR specialists in 10 agencies were using the software, which also offered limited employee self-service (ESS) features to about 60,000 workers nationwide, or about 40 percent of the USDA workforce.

In July 2002, USDA upgraded to PeopleSoft 8, a web-based HRMS with major improvements, including significantly expanded ESS features. For this implementation, USDA turned the new system on for all HR specialists and 60,000 workers on the same day in what information technology (IT) people call a “big bang” rollout.

Phased rollout for one, big bang for the other. Why? “For PeopleSoft 7, we had to do much more change management and more massaging of data,” says Hans Heidenreich, the project director for both implementations. That’s because USDA was moving from a home-grown system on a mainframe to off-the-shelf applications on a PC network, he says.

With PeopleSoft 8, the HR power users already had experience with a similar product, and moving from client-server to the web was easier than the move from mainframe to client-server, Heidenreich says.

“Both approaches work,” says Bill Henry, vice president for strategy at PeopleSoft Inc., based in Pleasanton, Calif. “The real issue is understanding the degree of change your organization can accept. The bigger the change, the more we recommend the phased approach.”

USDA’s experience illustrates some of the factors that dictate whether a phased rollout or big bang makes more sense. The pros and cons of each apply broadly to any HRMS, not just PeopleSoft systems.

From Mainframe to Client-Server

In the mid-1990s, USDA began a re-engineering project focused on HR, procurement and financial management. Like most large organizations, USDA had spent millions of dollars over the years to develop and support mainframe systems for administrative processes.

USDA decided it was time to move to a PC client-server environment, adopt off-the-shelf applications and get out of the business of developing software.

Agency officials decided to tackle HR first because the “software was ahead of other areas in being ready to implement out of the box,” explains Heidenreich. By 1997, his team members identified 360 requirements that HR software would have to meet, and they contracted a consulting firm to survey major vendors. Based on the survey, USDA had the top three contenders give demonstrations. “PeopleSoft came out as meeting most of our requirements,” he says.

Next came a short period of testing in a laboratory environment, followed by a more rigorous pilot test in an eight-person USDA HR office in Gainesville, Fla. Three months into the six-month pilot, USDA expanded it to HR offices in five other states, says Heidenreich. The pilot, now with more than 40 users, lasted through February 1999.

Next, Heidenreich’s team, which included about a dozen HR and HR information system (HRIS) specialists plus about six IT contractors, drew up a plan for the nationwide implementation, including training and rollout strategies. The team did not consider big bang, Heidenreich says. Rather, it decided to stage the rollout in eight phases, which would make it easier to train smaller groups of HR users and to clean up the data during migration.

In client-server architecture, part of the application sits on a server and part sits on the user’s desktop computer. For PeopleSoft 7, the server software resided on servers in USDA’s national information technology center in Kansas City.

The team began rolling out the software in early 2000 and completed rollout by that September.

Upgrade on Tap

The team immediately started to think upgrade; PeopleSoft 8 was released during the PeopleSoft 7 project. The new version’s biggest difference was its architecture, which was rewritten for the web. In the web-based system, the entire application sits on a server and is accessed through a web browser. No software is needed on the user’s PC, so upgrades and patches are all installed on the server, reducing work for IT.

Because PeopleSoft 8 is built for the web, it also has many out-of-the-box ESS applications that anyone with a browser can use for personnel transactions, easing HR’s load.

The PeopleSoft 8 pilot test was less rigorous. In the spring of 2001, the project team loaded PeopleSoft 8 into a pilot server and allowed any HR user of PeopleSoft 7 to try PeopleSoft 8. The pilot lasted two months, Heidenreich says. The rollout was held up for five months while the USDA upgraded its network. In May 2002, Heidenreich got the green light to upgrade.

The project team held one large training conference for all the HR power users, with more than half of the 552 showing up. They also visited each state for additional training and conducted some web-based meetings. “We didn’t have to do a lot,” Heidenreich says. “The users were really keyed into the new software already.”

In July 2002, the team switched everyone to PeopleSoft 8 on the same day. HR users in each office trained USDA employees in their state on the ESS applications.

Change Management

PeopleSoft 7 involved a lot more change management than did PeopleSoft 8, Heidenreich says. During that first rollout, HR users moved from old-fashioned green-screen terminals hooked into the mainframe to desktop PCs. Many had never used a mouse.

“We did not anticipate having to teach people how to double click,” says Heidenreich. The team changed its training program after the first users went through it. In the second and later phases, training included PC basics—such as using the mouse and navigating.

There was quite a bit of resistance to PeopleSoft 7, says Dave Balke, an HR development specialist in the Kansas City computer center who has been involved in training efforts. “What’s wrong with the old process?” was a common refrain, he says. The phased rollout helped overcome this. As the first groups used the new system—and liked it—they created a buzz that encouraged later groups to adopt, Balke says. Ironically, some groups began to complain about not getting it sooner.

The phased approach also gave Heidenreich’s team a chance to identify and fix unanticipated data problems before further rollouts. Most of these—about a dozen significant ones—were discovered in the first three phases, he says.

For example: When planning the data structure, the team assumed that one employee could not work for more than one USDA agency at a time. However, when the team rolled out to the second group, they found that one employee did work for two agencies. With a bit more research, the team identified 200 such cases among 60,000 employees, fixed the problem in the database and identified the other exceptions before the next rollout.

By the time the team reached the last two rollout groups, they had solved all the problems. That ability to gradually tackle problems is a big advantage of the phased rollout, Heidenreich says. “With each phase we learned something new and improved it. You deal with them one or two at a time,” he says. “With a big bang you would have been dealing with 10 or 12 all at once.”

The biggest problem with the phased approach was maintaining the mainframe system and PeopleSoft simultaneously for several months. During this period, when an employee called the HR support center, the staff first had to determine whether the records were in the old system or the new system.

There were other minor aggravations. Some employee data were moved to PeopleSoft in an early phase. Then the employee transferred to a job whose HR office was not on the new system. The employee data had to be transferred back to the mainframe. “Their records had to be moved back and forth each time they moved,” says Heidenreich.

Fewer Data Problems

The big bang was not considered for PeopleSoft 7, but the project team studied and debated both strategies for PeopleSoft 8, including HR users in the discussion. The more they looked at it, the simpler the big bang seemed, for several reasons.

First, by then every HR user was accustomed to the PC and to PeopleSoft. The differences between the two versions were not visible to users. Even the non-HR employees were accustomed to doing some limited ESS on the corporate network.

By the time the PeopleSoft 7 rollout was finished, the team was confident that the data was—in IT parlance—properly scrubbed and cleaned. This would mean fewer data-related problems with PeopleSoft 8, which had tools to automate the data migration from PeopleSoft 7.

The two data structures were different enough, however, that maintaining the two in a phased rollout would have been costly, says Deneen Fox, project manager for Unisys Corp. of Blue Bell, Pa., the USDA’s IT contractor. “Maintaining both would have been a configuration management and control nightmare,” she says. During the first implementation, the old mainframe staff kept doing what it had always done. The new PeopleSoft 7 support staff maintained the new environment. Fox says she would have needed two separate staffs for a phased rollout of the upgrade.

She got a little taste of the problem anyway. Remember that five-month delay for the network upgrade? Fox’s IT staff had to upgrade, patch and maintain data in both versions during that time. No one was using PeopleSoft 8, but the data and applications had been loaded for the pilot, and the IT staff didn’t want to fall behind with changes while waiting to start the rollout.

When they did roll out, there were data-related problems in the big bang due to some 10 custom applications the project team wrote, which users identified quickly after the new system was turned on, Fox says. The team prioritized and fixed those problems.

By the time they got ready to upgrade, PeopleSoft 7 had won wide acceptance, and it would have been almost inconceivable to make some groups wait for PeopleSoft 8 in a phased rollout, says Balke. “The processes were the same, the icons were the same, the name of the application was the same. PeopleSoft 8 was just enhancements,” he says.

“When the community is ready for change, the big bang is probably better,” Heidenreich says. “The advantage is you get it done and have it out there. Everyone’s on the same system, and you move forward.”

Global Implementations

The USDA’s experience illustrates the major issues in choosing between a phased or big bang approach, says PeopleSoft’s Henry.

One exception: When rolling out to global operations—whether initially or during an upgrade—the organization may want to lean toward the phased approach. If HR processes differ from country to country, a geographically staged phased approach is more manageable, Henry says.

It can also make sense to use a phased rollout by HRMS module, rather than by geography or office group, Henry says. For example, some adopters first roll out the modules that will provide the biggest benefit.

Rollout decisions are increasingly dollar-driven, Henry says. CIOs with limited budgets may not be able to support a big bang rollout. During the economic downturn, more companies opted for some sort of phased approach just to spread the costs over more than one budget year, he says.

“I would say in the past 18 to 24 months we’re seeing most people use some type of phasing,” he says. ”Back in the 1990s, when people were more willing to commit a lot of money up front, we saw a lot more big bangs.”

Bill Roberts, technology contributing editor for HR Magazine, is a freelance writer based in Los Altos, Calif., who covers business, technology and management issues.


​An organization run by AI is not a futuristic concept. Such technology is already a part of many workplaces and will continue to shape the labor market and HR. Here's how employers and employees can successfully manage generative AI and other AI-powered systems.