The New Amazon Distribution Model

Recently, Kiva Systems told its existing clients how its acquisition by Amazon would affect them. But the news also revealed something about Amazon’s distribution strategy.

So here is the news: Kiva will not be selling any new systems to new customers for the next one and a half to two years. Instead, Amazon will be filling its new one million square foot distribution centers (DCs) with Kiva robots. In its second quarter financial statement, Amazon said that it had opened 6 DCs this year, with 12 more scheduled for the remainder of 2012 (17 were built last year).

I wrote a strategic report in 2009 titled “Warehouse 2025” (available to ARC clients only) that looked at the economics of the new generation of distribution robots.

Amazon appears to practice mainly pick-to-cart. Each cart has multiple totes. Its warehouses have mezzanine levels; workers often have to push their carts onto a lift to change levels. Once a tote has been completely picked, it is placed on a conveyor and transported to pack stations. Amazon’s labor goals are roughly 160 picks per hour.

In contrast, Kiva Systems leads to higher productivity per worker. Workers work at combination pick/pack stations and the Kiva bots deliver the inventory to them to pick. The pick rate will depend on what is picked — if it is small items picked in multiple quantities, you might go as high as 1,500 items per hour; apparel is closer to 200 items per hour. In my report, I used 600 line items per hour as an average for Kiva empowered workers.

So what would this mean in terms of workers? My initial estimate was that Amazon would employ 45 percent fewer workers per DC. But one of the contacts I interviewed for my strategic report told me that number was too high; his warehouse employs 30-40 percent fewer people by using Kiva.

If you conduct a Google search on “Amazon distribution center,” you come across announcements over the past year for DCs in Virginia, New Jersey, and South Carolina. According to these articles, warehouses costing roughly $50-65 million will employ 1,500 to 2,000 workers.

But then I came across an article about a new Indiana DC. This DC is costing $150 million and the company is only promising 1,000 new workers. Could this be one of the new warehouses that will be using Kiva? The employment numbers seem to add up. This DC will employ at least one third fewer workers.

The total cost of the new Indiana warehouse also seems roughly right. Kiva is sold as a system – robots plus their warehouse control system software. But if you take the total cost of the system and divide it by the number of robots, I estimate that the cost per robot is about $20,000-$30,000. There will be about 6 to 9 robots serving each station. With a two shift operation, and the understanding that the great majority of workers will now work pack stations, you can see how the Kiva robots would add $40 million to the cost of the new DC even if Amazon buys the robots for half of the purchase price of the general market.

Let’s tackle that last assumption, that Amazon can buy the robots for half price. Part of that is based on the fact that Kiva Systems sells premium-priced, high-margin product, which Amazon will now buy at cost. But I also see falling unit costs based on a huge increase in production volume. Word on the street is that Kiva did about $100 million in revenues last year. Based on my back of the envelope calculation, one of Amazon’s new DCs will require more than 80 percent of Kiva’s total production from all of last year. If even half of the unbuilt DCs scheduled to come online this year use Kiva robots, we are looking at 5X production ramp.

Again, doing a little rough math, and depending on what you think a fully loaded warehouse worker costs and exactly how many workers become unnecessary, Kiva warehouses could save Amazon more than $20-35 million in wages and benefits per warehouse per year. If it implements Kiva at 6 DCs this year, and 20 more over the following two years, the company will get its payback for the Kiva investment.

There are some other advantages. Without buying Kiva and ramping production volumes, Amazon could never implement Kiva across its network of new DCs. Further, the company’s backlog prevents other e-fulfillment competitors (that are not Kiva’s existing clients) from copying what it is doing for at least a year and a half.

Finally, as Adrian Gonzalez pointed out in a recent posting, Amazon appears to be on a quest for same day deliveries. Kiva robots provide a blend of automation and flexibility that traditional forms of heavy automation cannot match.

What at first glance looked like an absurdly high acquisition price for Kiva Systems becomes completely rational. But the success of this investment will depend on how quickly and successfully Amazon can ramp the production.

Leave a Reply

Your email address will not be published. Required fields are marked *