Slashdot Mirror
Wal-Mart's Faltering RFID Initiative
itphobe writes "Baseline magazine has up an in-depth look at Wal-Mart's years-old RFID initiative. Things apparently haven't gone so well for the retail giant. 'The lack of any obvious concrete gains has raised questions as to whether Wal-Mart should delay or freeze its RFID plans. For now, however, Wal-Mart says it will stay the course ... By January 2006 the company hoped to have as many as 12 of its roughly 130 distribution centers fully outfitted with RFID. That effort stalled at just five distribution centers. Instead, the company is now focusing on implementing RFID in stores fed by those five distribution centers so it can gain a bigger window into its supply chain.' Overall the article focuses on the original intentions of the RFID project vs. their implementation. It also discusses several of the technical elements required to adapt RFID for the US juggernaut."
That's why the industry wants a multiple-nines read rate on tags. Missing a tag is a big deal. On some items (like a pallet of bikes) getting 100% almost every time is easy (a bike in a box is mostly air). On other items (like cans of soup) it's extremely hard. Wal*Mart is unlikely to demand that individual soup cans get tagged, but they might want cases tagged, but even then it's hard because it's soup -- mostly metal and water, two things that don't play well with RFID tags.
One thing to remember is that these companies aren't run by complete idiots. If they pay 6c per can of soup they won't demand that every can be tagged. They also won't trust that the number of RFID tags they've scanned is the number of items shipped. Instead they'll have a shipping manifest that says "300 widgets". If the RFID scanner says it found 300 individual RFID tags, then they can be pretty confident that they read all the tags and that their order is complete. If instead it says 293 they'll know they either have to try to scan it a few more times, or if that doesn't work they'll have to disassemble the pallet and figure out if there really are only 293 widgets or if there are 7 that aren't getting read. If the system works well enough that most of the time it says 300 widgets when there really are 300 widgets it could be useful, but 300 widgets == 300 tags == $30-$60, which is a lot, depending on what's actually on the pallet.
I write retail and warehousing software for a company every American knows.
Have you ever seen work at a warehouse, or at the back of many retail stores? The number of mis-scans, duplicates and such can be pretty significant. Companies account for this by doing physical inventories, which have a substantial labor cost. And those physicals end up disagreeing with sanity check recounts by up to 2%! In a store that has a significant cost per item, a 5 cent tag would be a cheap price to pay to get rid of most of the physical inventory costs and increase the efficiency of inventory control. At $20c, it eats into margins too much for most.