Evan Schuman's StorefrontBacktalk

Microsoft this week will roll out a new dot release of its Point of Sale package, one that claims better scalability capabilities.

Although that is clearly not earth-shattering, it does cast some light on the age old question of IT organizational strategy when a company is enjoying rapid growth.

The conventional IT thinking has been that, all else being equal, consolidated and centralized is the most efficient way to run a business. All of the data is analyzed globally and purchasing trends can be quickly spotted and acted on.

A less traditional approach is mirrored sites. This allows for the purchasing of smaller systems (geared for small retailers), albeit a heck of a lot of copies.

Microsoft is fond of positioning itself for the small retailer, proudly pointing that its research shows that more than 99 percent of retailers are actually one-location operations.

Of course, if you do a retail survey that counts Wal-Mart as “one” and Sam’s Bait Shop as “one,” you can’t necessarily expect meaningful results. But Microsoft still makes a fair point. (Long-time readers know how crazy it makes me to concede that Microsoft is making a fair point.)

Microsoft is therefore in a position to encourage lots of sites running operations on their own with the data consolidated and analyzed later, just as Oracle and SAP are more inclined to suggest consolidation, as that’s their playground.

Mike Dickstein, Microsoft’s director of POS products, points out that even with chains, one location can be remarkably different from another.

There are some chains whose each and every store has a remarkably complicated and sophisticated SKU selection, such as Home Depot and Sam’s Club, while other chains have very small individual store locations, Dickstein said.

One such chain would be Family Dollar, a $6 billion discount retailer with 5,948 stores in 44 states. When talking with Family Dollar’s CIO, Josh Jewett, last week, he spoke about his chain’s plans to add 400 to 600 news locations every year.

Indeed, dealing with that kind of store growth is one of the chain’s top four strategic initiatives.

Family Dollar’s stores average about 8,000 square feet each, compared with a typical Home Depot size of about 140,000 square feet and about 40,000 SKUs at each location.

Given the small size of their locations, they’ve been working with Wyse thin-client Web-based applications, mostly for kiosks that will take job applications. Those applications?if approved?will automatically be dropped into an Oracle payroll application.

“We are looking at our organizational structure” and trying to “find the lowest TCO [total cost of ownership] way to do it,” Jewett said, although he was very careful to not say how he planned to do so.

“I’m not comfortable commenting about or sharing how we are structured to support growth. We deem our ability to grow well and profitably a strategic differentiator,” he said.

One of the reasons this debate is returning is that the options and the IT landscape have morphed over the last few years.

Consider Metro, Germany’s largest retail chain at about $68 billion in annual revenue. Gerd Wolfram is managing director of the Metro Group Information Technology and also serves as the chain’s chief technology officer.

The chain has some 2,400 stores in 30 countries, employing about 250,000 people.

Although his chain is still solidly in the centralized camp, today’s technology is raising the question of whether that needs to be adjusted.

“Years ago, it was a strong trend to centralization. Today, [the retail IT industry] is going back to decentralization,” he said, talking from the show floor at this week’s NRF event in New York City.

“With a Web services approach, the question is, ‘What is centralized and what should be decentralized?’ There are a lot more options than before.”

Among those options are Web services and an assortment of tools and services that make a remote complex retail environment easier to manage.

Wolfram sees a big decentralized argument being store independence. But he doesn’t mean the ability to alter product mix to local customers. He’s referring to the store’s ability to fully function when servers at headquarters crash, the old single-point-of-failure issue.

Wolfram said the battle of centralized versus decentralized must be decided “on the basis of security, cost of operation and user-friendliness. There’s no final answer.”

A key assumption behind the decision to keep operations centralized is that it will deliver faster analytics, which theoretically enables faster decisions.

Although the faster analytics part is certainly true, the faster decisions part has a human component that, as humans so often do, mucks up the equation.

In other words, real-time data doesn’t help much unless there’s a manager out there willing to give it real-time attention.

Wolfram sees the need for software to address that concern, to act as a sophisticated filter that will flag any exceptional trends that warrant management’s attention.

“There is an issue. Data overload is there,” he said.

So much of retail IT operates on the premise that the more data that is delivered, and delivered as quickly as possible, the higher quality decisions will come out the other end.

But this misses the fundamental issue of context. Take a grocery chain. A store manager sees that a particular product has only been sold 16 times in the past month, which let’s assume is far below the threshold of this particular store manager.

Before replacing that item with something else, does the manager query who bought those items and what their buying history is? What if the manager knew that four of the people who were buying that rhubarb-scented disinfectant were among the store’s highest profit, and largest volume, customers?

And what if the manager then knew that those four customers simply love that product and the discontinuation of it would send them to every other grocer in town, searching for it?

Suddenly, that $1.29 spray can is associated with the potential loss of a lot of repeat business and a healthy slice of profit.

Today’s software is trying in vain to replicate the 1950s-era local grocer who knew all of his/her customers and what they bought. They had CRM databases in their heads that nothing today from the likes of Oracle, SAP or Microsoft can match.

Yes, the reverse is true. Those vendors today offer databases that provide information that earlier merchants couldn’t dream of. But if they’re not using the data, what good is it?

Centralized or decentralized should be based on a lot of factors, but having a realistic sense of how your people make their decisions is a pretty good start.