American Apparel's RFID Project Doesn't Automate Cycle Counts, It Wipes Them Out

A funny thing happened on the way to American Apparel (AMEX:APP) using RFID to automate the process of inventory cycle counts: The retailer threw out the process. Instead of using RFID technology to let associates count faster—which is what CTO Stacey Shulman was talking about a couple years ago—there's now going to be no associates, no counting, no cycle time. Instead, ceiling-mounted RFID antennas will track inventory continuously. From the point of view of store managers and associates, the process won't be automated. It will be gone.

Technically, of course, it will just be invisible, involve no store manpower, and have cycles too short to notice and too frequent to count. But change a store process that much and it really is gone. That's the kind of transformation that pundits like to talk about, but almost no chains actually do—mainly because retail IT departments are too busy automating processes instead of obliterating them.

And let's face it: Once you grasp the idea of what American Apparel is doing, it seems obvious. Fixed antennas up in the lights? Continuous shelf checks? Of course. But in 2011, Shulman was wondering how fast an associate would be able to wheel an RFID-reader-laden cart down an aisle, automatically counting merchandise on both sides. That was obvious too—obviously a lot faster than the manual process.

And as soon as the data was collected it could be used to update inventory numbers, and the store manager could immediately and accurately see what needed to be replenished on the shelves and what had to be reordered. The new RFID-enabled cycle counts would be a huge improvement over the old process.

Here's Shulman two years ago: "I get reports every day that let me see exactly how many items are missing, exactly which stores need to cycle count, exactly what the trend has been, exactly the in-stock percentage, exactly what stores are having problems." That's what RFID-tagging every item made possible: Shulman and other American Apparel executives could get data on how well the stores were being run on a daily basis. All it required was associates walking the aisles, doing a faster and more frequent version of what they were doing before.

Now the data is real-time and continuous, with associates freed to replenish merchandise and—oh, right—sell it to customers.

This comes to mind because one piece of coverage we saw about the new American Apparel system said it could all have been done in 2003. No, it couldn't. Did the technology exist? Probably. Was RFID tagging in place and being used in stores? Not many. And when it was being used, it was typically to speed up what associates were already doing. That stack looks short? Use the RFID gun to see which sizes need to be replenished. Then maybe (or maybe not) the RFID tag would be used to confirm the price and adjust inventory levels at POS, too—a slightly better bar code replacement.

Shulman's idea in 2011 was to use the data for store management, not just merchandise replenishment. Check everything, not just the short stacks. Track what was moving at the shelves, not just the POS. It was a good enough idea, with good enough ROI potential, to implement. And until it was implemented, no one could have stepped back and said, "Wait, why use one bank of readers on a cart once a day when we could put readers everywhere and get rid of the cart and do it continuously?"

Facepalm. Obvious. But not until then. That rolling-cart automation probably couldn't have been skipped over on the way to cutting out the labor cost, the cycle time and the inefficiency.

Now, of course, anybody can do it. All another chain needs to do is get all merchandise RFID-tagged, implement inventory tracking dashboards, test the technology and train store managers and associates on paying attention to what the system is telling them. The fact that RFID can now actually claim to reduce labor costs, not just empty shelves, might actually get a lot more retailers on board for the item-level tagging that will make everything else possible.

But it's worth remembering that, most of the time at least, until you do process automation, you can't get to process transformation—or obliteration.