We have posted a few times about how miserable it can be to work in an Amazon fulfillment center. (See for example here.) We have also had a few posts on Kiva robots — both before and after Amazon bought the company in 2012. Kiva produces automation systems for fulfillment centers. These are essentially robots that bring shelves to pickers who select what is needed to complete customer orders. At the time Amazon bought them, Kiva’s clients were firms like Crate & Barrel that while significant catalog/web retailers had far less variety than Amazon. Indeed, one of our posts on Kiva was basically asking when the robot hordes were coming to a fulfillment center near you.
According to the Wall Street Journal, those hordes have now arrived (Amazon Robots Get Ready for Christmas, Nov 19). Back in May, CEO Jeff Bezos claimed that they would increase their number of robots from 1,400 to 10,000 over the year. What difference does this change make?
At a 1.2-million-square-foot warehouse in Tracy, Calif., about 60 miles east of San Francisco, Amazon this summer replaced four floors of fixed shelving with the robots, the people said.
Now, “pickers” at the facility stand in one place and wait for robots to bring four-foot-by-six-foot shelving units to them, sparing them what amounted to as much as 20 miles a day of walking through the warehouse. Employees at some robot-equipped warehouses are expected to pick and scan at least 300 items an hour, compared with 100 under the old system, current and former workers said.
Obviously, tripling the number of items picked per hour is a nice productivity boost. However, it is not clear that this touches every part of Amazon’s business.
Running an Amazon fulfillment center has to deal with two realities. First, they carry a mind-blowing range of stuff. Second, their sales volume is highly seasonal. The first point suggests that it may not be worth having every item they offer accessible to the robots. The second calls into question whether they should size the robot horde for their busiest day — that would be helpful on that busy day but it also means a lot of idle capital the rest of the year. In many ways, it would be cheaper to deal with peak days as they have been dealing with peak days, hiring lots of temporary workers.
So how is Amazon going to use its robotic minions? One suspects that they are putting their most popular items (e.g., hit video games) on shelves for the robots to schlep. If it is easier to get temps up to speed on picking stuff brought to them, the seasonal workers staff the Kiva stations while the experienced picking pros roam the hinterlands of the fulfillment center for the more rarely ordered items. This configuration would give very cheap costs on a large number of items while minimizing the impact of odd-ball orders.
4×6 foot shelving units are brought to the picker, allowing them to select the right item. That implies that most (90%?) of what’s being moved isn’t wanted, but moved anyway. How much energy is being wasted moving many tons of unwanted merchandise around the warehouse?
Thanks for the update, Marty.
I actually discuss this in the elective (after the Peapod case)—I don’t have WSJ subscription (should get that!); would you mind emailing me the full article? thanks Professor Jan A. Van Mieghem Kellogg School of Management at Northwestern University
From: The Operations Room <comment-reply@wordpress.com> Reply-To: The Operations Room <comment+lai2or7n0okbfgo0u1ab_3@comment.wordpress.com> Date: Thursday, November 20, 2014 at 10:37 AM To: Jan Van Mieghem <vanmieghem@kellogg.northwestern.edu> Subject: [New post] Automation at Amazon
Marty Lariviere posted: ” We have posted a few times about how miserable it can be to work in an Amazon fulfillment center. (See for example here.) We have also had a few posts on Kiva robots — both before and after Amazon bought the company in 2012. Kiva produces automation “
Dave B – I have been mulling over your comment about energy consumption. I have been considering the minimum amount of energy it would require to operate a robot, and the maximum that can be accomplished with with that energy. For example, if it took a robot a certain amount of energy to lift a small bin of tee-shirts and drive it over to the picker, how much more would it take to lift a 5’x5′ shelf and bring it over? Also including that the pick-per-shelf on a warehouse-wide basis would be greater than 1, the wasted energy would be reduced. Just a thought!
As a sidenote: Kiva has automated not only order fulfillment, but also inventory placement. Its based upon a linear program that places the high turn items closest to the picking stations, and the low turning items further away. Thus over time, the system can become more energy efficient.
Great write-up!
Very cool to see automation taking hold behind the scenes– these are humble beginnings, but it will be cool to see how the technology continues to evolve over the years.
Woah! I’m genuinely taking pleasure in this template/theme of this website. It’s easy, however successful. A lot of times it’s severely to get in which “perfect balance” between fantastic excellent along with looks. I need to say you’ve carried out an amazing employment on this.