Amazon randomness

Amazon had an aha! moment somewhere during the evolution to what it is today, which this video explains quite well. It runs contrary to conventional thinking about warehouse organization, because Amazon uses computers and barcodes extensively.

When Amazon receives a product from a supplier, they break the case apart and store each item randomly all throughout the warehouse, which means when a person orders more than one item, the walking time to gather the items goes down. Even if it is a newer warehouse where robots are doing the moving, it still saves time. Not having fixed locations for certain merchandise means the space adapts to what people are currently ordering.

This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to Amazon randomness

  1. briand75 says:

    Fred – similar experience with procurement process improvement in power generation: One of the parts of the process involved receiving Purchase Orders – many of which did not have complete information – for example part numbers, budget accounts, etc. The service-oriented folks in Procurement would spend hours trying to complete the P.O.’s which was crazy as they don’t have access to the information needed. So these orders became “clunkers” that had to be manually dragged through the automated system which cost an amazing amount of money and time. The correct solution was to reject the incomplete P.O. and keep doing so until it was completed properly. They weren’t being service-oriented, they were creating a shadow system that was inefficient, not effective and costly.

  2. briand75 says:

    Geek is right. Being an engineer, I love the orderly outcomes from the “random stow”. Fascinating t0 see some of the aspects of queuing theory in action.

    • Fred Stiening says:

      They can get away with this because the merchandise doesn’t deteriorate (much). At US foods, the expiration date was the most important criteria for which pallet to pick from, and shortening the walk time only came into play where the pallets had the same expiration date.

      Another video suggests that the putback routine evaluates the remaining items on the carrier and pushes it toward the back of the warehouse if nothing interesting is left on it.

      When I was working on a much more primitive system for auto parts at GM, one day I ran off a report of parts that had been in the warehouse for more than 90 days – the design goal as a JIT buffer for a crowded plant was replenishment every 4 hours and time in warehouse of one day. My report caught a number of screwups, including the warehouse had received parts destined for the chevy truck plant on the other side of Flint. If a truck showed up without an ASN (electronic Advanced Shipment Notifucation), the office people would just create one by hand so they could receive the truck – without apparently noticing that the Deliver To: address was across town

      The facility was pretty impressive, but was plagued by old school warehouse thinking. While the computer knew what was in every storage location, they still did a reconciliation of the physical inventory every night (using bar code scanners). They didn’t trust the computer. The reason was they would do things like stop the sortation conveyor system and use a forklift to remove the pallet from the middle of the conveyor, then remove the pallet carrier with the bar code. The empty pallet would arrive at the putaway station, and the putaway operator would physically remove the carrier from the conveyor and never complete the phantom putaway and emergency shipment.

Leave a Reply