Evan Schuman's StorefrontBacktalk

Two things about that deal jump out. First, Accenture got the deal by promising that it would cut costs by getting rid of customized systems and replacing them with “packaged vanilla solutions.” Or as (non-Accenture) consultant Bob Lewis later rephrased the pitch: “In exchange for hefty fees, we’ll rip out your sources of competitive advantage and replace them with generic alternatives.”

And second, Accenture began the rip-and-generic-replace in 2004. By 2006, Best Buy’s slide began. Just when Best Buy needed competitive advantage against Amazon, that’s exactly what it didn’t have.

Now Best Buy wants to get that advantage back. That’s good. What’s worrisome is how Joly said he plans to go about it: a big, highly customized, multiyear project being launched only 18 months after Best Buy started rebuilding IT.

Understand, this is nothing like the plain-vanilla systems that Accenture built for Best Buy. That’s the whole point—Best Buy already knows how that turns out. To connect up all the websites, provide highly targeted search and do the other things Joly wants, there’s going to have to be a certain amount of invention and a lot of experimentation to get it right.

That can be done. But Best Buy hasn’t managed a project like that at least since Accenture took over. And building it in isolation and then rolling it out two or three years later to replace the existing collection of Best Buy websites? That sounds like another Target.com disaster in the making.

Big IT projects fail—a lot. Only a few launch cleanly. Many more can be rescued if they have time to get the kinks worked out after they’re rolled out. But E-Commerce projects don’t get that type of breather. Customers just aren’t understanding about things like that.

But big IT projects face another problem: Two years is a long time in corporate time, especially for troubled retailers. CEO Joly may not be around that long. If his successor looks at this project 60 percent of the way in and says, “This project is expensive, we have no guarantee it will do what we want, and we already have websites that work. Tell me why we should keep spending this money,” what is anyone in the IT or E-Commerce groups going to be able to say?

The solution to both those problems is the same: Break up the project, design an architecture, build a foundation, then replace one of the existing Best Buy sites with a site that plugs into that foundation. That simultaneously starts real-world testing early and begins delivering ROI fast.

It also gives developers experience with how the new platform works, where workarounds are necessary and what to avoid completely. It eases customers onto the new platform (customers only hate one thing more than an E-Commerce site that works badly, and that’s a site that somebody suddenly changed on them). And it gives executives an answer to the “why finish this project?” question.

In short, it eliminates a lot of surprises for everyone. As other Best Buy sites are built and plugged into the platform, more kinks will be worked out. By the time the main E-Commerce site is ready to go live, the risk—and the unpleasant surprises—will have been cut to a minimum.

At a time when Target’s problems are still fresh in many people’s minds—and when Best Buy may not survive that type of disaster—let’s hope that’s what Joly has in mind after all.