What's Best for You

There are no rules for which approach to integration companies should take. There are, however, some trends.

Companies with fairly straightforward business processes without a lot of specialisation are good candidates for a single-instance project. So are companies that don't consider themselves IT leaders - maintaining best-of-breed solutions takes more programming skills than doing a single-instance project.

On the flip side, companies that feel as though either best-of-breed applications or their programming skills give them a competitive advantage should consider going with (or waiting for) Web services. They should also consider going with Web services if they have business units whose processes would be difficult to conform to a traditional ERP system.

It's worth noting, however, that no matter what path a company takes, its CIO will most likely have to do a little of both. Andrew Macey, a vice president with IT services company Sapient, compares integrating best-of-breed applications to cleaning out a closet; it's an opportunity to decide which applications to keep and which to get rid of. Similarly, it's unrealistic to imagine that you will be able to rely on a single ERP instance for every business need. In fact, most ERP vendors partner with smaller specialty vendors for such tasks as tax calculation. "You are going to wind up with some integration chores no matter which path you go down," says Macey.

Web services-based integration and single instance are comparably sized projects. Each will cost millions of dollars and take at least a year. The nature of your business will determine which one you choose. "There are pros and cons to both paths," says Macey. "There's no easy answer."

SIDEBAR: A Cautionary Note on Web Services

Whether you take the plunge or wait for standards to emerge, the costs and risks will be the same.

Most integration efforts today don't employ Web services. Rather, they work with more established XML-based messaging middleware, such as MQSeries and ActiveWorks. But there is general agreement among experts that Web services is a viable long-term integration answer. In fact, Web services is already popping up in some companies, inside the corporate firewall. But Cap Gemini Ernst & Young chief technologist for the Americas John Parkinson says that these projects to date have mainly been small pilot-type projects, and that while they were challenging, they were nowhere near as challenging as integrating core applications is going to be. "People have an overly optimistic view of Web services," he says. "Companies are underestimating how complicated it's going to be."

Companies that start a Web services-based integration effort today are assuming more risks (see "Calculated Risks", page 96). They're basically coming up with an idea of what they think a Web services architecture will look like in two to three years and aiming for it, says Parkinson. If they miss their guess, they may have to do it all over again if they can't make changes to what they've developed.

The alternative is to wait for standards and best practices to emerge. By waiting, the project will be cheaper and follow a more prescribed path. But you may miss out on a couple years of benefits.

Parkinson expects that the ROI on both approaches will ultimately be the same, with the quicker return of the latter (waiting) making up for the increased cost of the former (taking the plunge). The bill for both, he believes, could be similar to a single-instance ERP project.