Fulfillment: Man vs machine

There is a really fascinating story in today’s Wall Street Journal on how e-tailers are organizing their fulfillment centers (Holiday Help: People vs. Robots, Dec 20). In a nutshell, the question is to what extent should retailers rely on automation instead of humans. Robots, of course, can be a lot more productive. As stated in the video below (but not in the printed article), Crate & Barrel has found that introducing robots has allowed one person to do the work of six in the old system. But introducing robots requires a multi-million dollar investment.

The robots in question come from Kiva Systems. The video below shows that, like the Drells, they can dance just as good as they walk.

At one level, this seems a classic example of a tradeoff between a manual and an automated process. The standard argument would suggest that the bigger you are the easier it is to justify labor-saving automation. Put more bluntly, Amazon should love robots. But they don’t. As the article notes, Amazon relies on humans to roam the aisles and pick the goods.

Meanwhile, just outside of Phoenix, Amazon.com Inc. takes a more human-powered approach. There, employees walk 18 to 20 miles a day down aisles lined with shelves, filling library-style carts with the latest orders and carrying items back to packing stations. For the holiday season, this Phoenix facility—one of about 20 that Amazon has across the U.S.—quadruples its staff to about 1,200 to handle the holiday rush, with teams working 24 hours a day.

Which approach is better is a matter of debate in the 15-year-old e-commerce industry. Amazon has avoided robots at its fulfillment centers, aside from systems that came with recent acquisitions such as Diapers.com. Yet Kiva Systems Inc., the company that makes the robots used by Crate & Barrel, now has systems in place at several top online retailers in the U.S., including Gap Inc., Staples Inc. and Gilt Groupe Inc.

So why has Amazon eschewed automation?

There seem several factors at play. One is variety. The Crate & Barrel fulfillment center discussed in the article stocks 8,000 different products. That is enough to assure a wide selection of place mats and wineglasses but is nothing like the variety that Amazon handles. Presumably, having too much diversity in picking slows down robots enough that they lose some of their advantage over humans.

Another issue is volume variability. That is, how much difference is there between slow and busy times and can one justify sizing an automated system for the busiest day of the year? Take the numbers from Amazon. If their volume quadruples during the holiday season, can they really afford to have huge amounts of capacity sitting idle for 10 months out of the year? Relying on temporary workers and over time seems a reasonable option. This seems like a more compelling argument than variety considerations. It also, however, raises an interesting question: If you can use temporary workers, why can’t you use temporary capital?

Kiva said it is testing a program to rent out robots seasonally, so that companies could beef up their robotic staff for the holidays, just as firms like Amazon do now with humans.

That is, Kiva is considering servicizing its equipment. That’s an elegant solution but it is potentially complicated. If nothing else, it will seriously alter their balance sheet. I have to think that Kiva currently carries little more than demo models and builds almost everything to order. Moving to renting machines would put (potentially) millions of dollars of robots on their books. They may well get a decent return on that but as long as their customers are primarily B-to-C sites whose sales spike at the same time, they are merely making their customers’ problem their own.