Is RFID Another Word for GIGO?

By D'Anne Hotchkiss, Editor -- RFID News & Solutions, 8/24/2005

Tons of raw RFID data is coming back through the supply chain to Wal-Mart’s top-100 suppliers. Now what? Do what with it? Manage it? Avoid garbage-in, garbage-out?

Companies are evolving from the physics of making RFID technology work—from standards and tags and readers—into a new phase. “The physics is the first wave, and wave one is cresting. Wave two is the data wave, the beginning of the benefit wave. It’s starting to swell,” says Richard Beaver, director of global retail and RFID for Teradata, a division of NCR.  “Companies are starting to ask, ‘what do we do with the data?’”

Data from RFID tags can overwhelm an information architecture. Tags are read at a dozen or so points, from dock door to case box disposal. The EPC number, associated times, dates and locations are relayed as block data via Wal-Mart’s Retail Link portal to its RFID-compliant suppliers. The EPC number permits associations back to other data at the supplier or the retailer.

“It’s early to try to figure out how to use the data. Each company has to map it into an existing hierarchy. Most systems stop at the SKU level, not a unique ‘each’ level,” notesBeaver.

So companies should not load the data into a data warehouse “until you know what you want to do with it,” Beaver advises. “It’s just as important to keep bad data out as it is to keep good data in.”

Still, product data analysis requires that data reside in one location. “You can’t do analysis on the local level,” advises Peter Riemen, executive vice president for T3Ci. Companies are “all over the map, from very advanced to not. Some people have succeeded in pulling this together into a centralized system. Others are just housing it.”

T3Ci creates a system of records, brings the data into its own system, and draws conclusions. VeriSign  operates the RFID equivalent of the Internet’s root-name servers with the Object Naming Service (ONS), a directory of all the electronic product codes assigned to tagged objects. Companies can read the data in each location and then use the ONS to look up the associated information.

For RFID data to have value, it needs context. “It’s difficult to create context only based on local information. Much of the value is understanding the whole path something has taken,” Rieman says, calling that understanding “knowing the tag’s inheritance.”To know the context requires creating a data structure at the lower levels. Where is the reader located? Where was the item the last time it was read?

Managing this level of detail requires that companies add a layer of data to the model they’re currently using [see accompanying story, “RFID Data Now Pinpointing Supply Chain Performance Issues”]. “That means defining the processes that will generate a payback and early in the pilot. Be sure to involve someone from your data warehousing group,” advises Beaver. “Consider all the data impact. That may direct you to a different set of solutions than otherwise.”