Everything is miscellaneous
Yesterday in Las Vegas I spent a great afternoon at the headquarters of Zappos, which is the Amazon of shoes. They have what may be the largest in-warehouse inventory in the world--750,000 unique products, amounting to more than 1.5m pairs of shoes in the Kentucky distribution center (needless to say, this is a great Long Tail retail example, although I'll save that analysis for a later post).
Having just read David Weinberger's great new book, after which this post is named, one of the things that struck me most about Zappos was the way it organizes its massive warehouse. It doesn't. The shoes are placed randomly on the shelves.
Sure, Zappos experimented with various taxonomies and organizational systems, but none of them scaled. New shoes came in and old shoes went out, seasons changed and so did styles. When a vendor creates a new line, you can't just shift everything down to make room for it next to the old line. When a line is discontinued there's no way to pull everything up to fill in the space. Short of reorganizing the entire warehouse with every twitch of the marketplace, no ordered system was better than any other at accommodating change.
So Zappos just stopped bothering. It made peace with messiness. The shoes are logged in by UPC when they arrive and assigned a spot wherever there is room available. When it comes time to pick-and-pack, the computers tell the warehouse staff where to go. No single trip is optimized, but the system as a whole works as a minimum-effort machine. Just as random access works best for bits in disk drives, it turns out to be great for atoms in warehouses, too.



I think that method is called random pick, and it dates back at least 20 years, according to the person who explained its efficiency to me back in 1996, and who had been involved with the method for at least a decade.
Amazon implemented it in the late 1990s after using a ridiculous non-random pick.
Posted by: Glenn Fleishman | January 17, 2007 at 10:04 PM
When I was programming robots for the Federal Reserve's East Rutherford Operations Center, that's exactly how the containers of money were stored. Banks would deposit containers of cash with any mixture of currency: say $10 million worth of 50 dollar bills, $30m in 20s and $10m in 10s. The computer would record the configuration and location of that container in the warehouse. When another bank wanted to withdraw that same amount of currency in the same denominations, the computer would look up that container and direct a robot to get the money and bring it to the armored car depot.
Posted by: Patrick Di Justo | January 17, 2007 at 10:49 PM
I wonder if they ever have to defrag the warehouse...
Posted by: Jamie | January 18, 2007 at 04:28 AM
We built a variety of automated storage and retrieval systems (ASRS) as early as 1983 with such an approach. The analogy with a disk drive is apt ... we stole the seek optimization algorithms for the ASRS from disk firmware.
Posted by: UncleKenny | January 18, 2007 at 08:00 AM
There was a great article recently (I can't remember where)that talked about the new library systems being used, where books are sorted in the same way - as long as the computer knows where the book is, it can automatically pick it and return it to the desk quickly and efficiently..
Posted by: Harry DeMott | January 18, 2007 at 10:31 AM
It appears that I'm the only person in the world who _didn't_ know this already :-)
I've already forwarded on the link to several friends whose companies have warehouses!
Posted by: Mark Harrison | January 18, 2007 at 10:31 AM
How about the merger of eBay tickets and StubHub? What happens when Long Tails mate? Their combined warehouse is the accumulated drawers of every season ticketholder in America :)
Posted by: SuperFan | January 18, 2007 at 11:59 AM
I dunno Jamie, but I'm sure they have to do periodic garbage collection.
Posted by: HDC | January 18, 2007 at 02:38 PM
Amazon does do the same thing. It's called "random stow."
Here's another example of the same phenomenon: A friend of mine has a file on his desktop called "notes" (or something like that). Anytime he wants to remember anything--an address, a phone number, a great pizza place--he just appends it to the file. Then, when he wants to find it again, he just searches the "notes" file. It works fine.
Another example, of course, is Gmail, which substitutes searching your mail from sorting or filing everything.
It turns out that organizing stuff (physical or virtual) in storage is just a means of finding it again. If you have some other means--a database mapping item numbers to locations, or a fast search--you don't need to organize.
Posted by: Jason Crawford | January 18, 2007 at 11:22 PM
I'm a long tail fan....for sure. I think I get it...but this post confused me. I look at the picture of the warehouse and then the 1.5 million pairs of shoes inventory and I think...that's not long tail. It's a LONGER tail for sure. The long tail of shoes would be more like a place to put a picture of shoes and then users put links to where you can buy them...maybe getting a commission of sales....zappos of course would be on 1.5 million of those listings...but I'd guess there are many more from many countries. Am I missing something? That's like saying Amazon is the long tail of books.
Posted by: David Armstrong | January 19, 2007 at 08:50 AM
This random theory is used in filling up the planes too. Airline industry in the past(even today) tried various methods like the last rows filled first but a new study reveals that the industry can save millions by allowing passengers to fill the seats randomly on first come first serve basis.
Posted by: Aditya Gupta | January 19, 2007 at 03:04 PM
this is the same methodology used by commercial business records storage facilities. you just place the item in the next available free slot. The computer tracks where it was placed. it doesn't matter where the item is placed just so long as you can retrieve it
Posted by: pak152 | January 20, 2007 at 11:23 AM
I am sure businesses that store items in a specific location still enter data about the item and the location in a database. What they gain by ordering is security: when the database breaks down, they will still be able to
So how do Amazon and Zappos take care of this eventuality?
Posted by: Branko Collin | January 22, 2007 at 12:25 AM
I just finished reading A Perfect Mess, which you mentioned in an earlier post you didn't like. It's interesting that less than a week later you highlighted a story that would have fit perfectly into that book's theme of disorder often being more conducive to getting things done. Sometimes the organizing takes more time and money than the benefits of organization would deliver.
Posted by: Adam Jusko | January 22, 2007 at 09:56 AM
I programmed such a scheme for a manual warehouse for a British Leyland supplier in (IIRC) 1982 or 83.
It is KEY that the computer allocates the location, and the fork lift driver does as he is told. If the driver tells the computer where he put the pallet, you drown in data errors in about 3 weeks!
Then you have to take stock of everything!
Posted by: Ian | January 23, 2007 at 08:39 AM
i used to manage a small warehouse and shipping/receiving department for a CD-ROM company. The immediate thought that came to mind when I read this was, how do they do inventory? No matter how disciplined you are, the numbers in the computers eventually get out of sync with the numbers on the shelves, and you have to count. so unless all items of a particular SKU are stored contiguously, which introduces a nasty defragging problem as pointed out above, it would be very difficult to know for sure how much of a particular item you had without counting everything in the warehouse. and i hate to think of the computer-to-shelf-and-back-again gymnastics that would be required for the aforementioned defragging ...
of course, i'm sure they have this figured out somehow, i'm not exactly a rocket scientist in this department. perhaps there is some order to things - not entirely random. Like each long shelf has a "stack" of different skus which can be compressed and collapsed, and when you go to fetch something you have to count up from the beginning of a shelf, if that makes any sense at all ...
Posted by: Chris M | January 24, 2007 at 11:25 PM
Apparently, it works for documents as well. I remember a SMB document management system which files mail according the the date they come in, and a PC logs what is stored. Makes filing real easy.
Posted by: Chui Tey | January 25, 2007 at 05:29 AM
Robert Heinlein's characters used the same technique in his 1986 book "Number of the Beast". They couldn't fit everything they needed in their multiverse-jumping flying car so they stuffed things where they fit and let the car's computer remember the details.
Posted by: Trey Boudreau | January 25, 2007 at 05:40 AM
Chris, the 7 CD Long Tail audiobook had a profound impact on my business thinking. The large image of the above shoe warehouse allows me to realize the long tail of the products that I offer.
Thank you for your always intuitive and outside the box thinking.
Posted by: Jeff Knize | January 25, 2007 at 02:13 PM
About two years ago I worked at a Target store doing overnight stocking. It sucked, but one thing I that really struck me was that they did the same thing with extra stuff. Just stored it wherever it would fit (although they did have some general guidelines about what kinds of things would go where).
You just scanned the bar code with a hand held scanner, then scanned the bar code of the 'slot' and that was it.
It's not really that uncommon.
Posted by: Chad Okere | January 29, 2007 at 12:50 AM
I think...that's not long tail. It's a LONGER tail for sure. The long tail of shoes would be more like a place to put a picture of shoes and then users put links to where you can buy them...maybe getting a commission of sales....zappos of course would be on 1.5 million of those listings...but I'd guess there are many more from many countries.
Posted by: home gym equipment | March 27, 2009 at 11:06 AM