Evan Schuman's StorefrontBacktalk

Suffice it to say, the use of RFID did not evolve in that way; nor did it happen that quickly. The required infrastructure changes within retail were too expensive. The cost of individual tags, though much publicized, was not the controlling factor. Manufacturers would have had to add the tags to many different types of packages, and these packages could interfere with remote reading. Plus, readers would have had to be replaced at checkouts and in portals, backrooms and backdoors.

There were also warnings that consumer privacy would be compromised. Although most tags could only be read at short distances by standard devices, it has been shown that modified devices could be used to read tags at much greater distances. Removable tear-off tags were developed, but at a greater cost to the manufacturer.

Even as this greater complexity versus cost savings was debated, a small number of manufacturers quietly started to add RFID tags to selected items. This tagging was typically done with higher cost items, where tags could easily be integrated within packaging. And in some cases, laws required the close tracking of pharmaceuticals. Considerable experience was gained in understanding integration methods, cost of the infrastructure needed and reading accuracy.

But in many more cases, item-level tagging was replaced by efforts to tag cases and pallets. Retailers–including Wal-Mart and Metro–strongly pushed these moves and manufacturers–including Procter & Gamble and Colgate–participated in a major way.

And that is how Wal-Mart got into the situation it’s in today.

What should Wal-Mart have done instead? The best way to start RFID tagging–as Wal-Mart is now doing–is with specific categories that have natural complexities. Apparel, for instance, is an obvious choice. Clothing has a complex inventory structure; it is also seasonal and relatively difficult to inventory.

Display execution compliance and diminishment of out-of-stocks also need to be measured if an RFID deployment is to have substantial value. Many RFID attempts are examples of trying to measure the behavior of a very large and complex system, but one in which too many changes are made all at once. Any attempts should focus on two key aspects: Choosing categories that are likely to be of substantial value, and considering the effect of operational execution within those categories.

Another crucial difference now is money. In the early tests, manufacturers had to absorb a large portion of the costs. The new item-level tests, however, include more marketing funds from Wal-Mart to support the complexity of the execution required to perform the needed tests. As a result, more manufacturers will be able to participate, which will in turn provide a better experimental base and extend that base to midsize companies.

So Wal-Mart’s choice to open with a limited selection of clothes and underwear makes perfect sense. As I mentioned, clothing has a complex inventory structure, is seasonal and is relatively difficult to inventory. So the inclusion of tags read by store employees with handheld readers–and even by readers on shelves–could simplify and make inventory more transparent. This inventory data could then be aggregated chain-wide to do predictive analysis of sales by region and time, to improve ordering and even to design processes. In addition, business intelligence could be linked to multiple sensors within the supply chain, resulting in a smarter and more inexpensive system overall.

Wal-Mart will not be reading these clothing tags at checkout, so there will be no connection to personally identifiable information, which alleviates the privacy concern to some degree. Still, there will likely be some contention that any tags connected to an article of clothing could be remotely read and might lead to some loss of privacy.

After a very long RFID trail–which Wal-Mart was a huge player in paving–Wal-Mart has now started on the right track.