From the IDA Distinguished Infocomm Speaker Series
Singapore
August 15, 2006

Michael Dell on Demand-Pull.

QUESTION: I was studying a couple of balance sheets in this industry and I noticed that companies like IBM and HP use very high equity, like $40 billion and I thought Dell is using maybe just $4 or $5 billion. How are you able to generate such high profits? What are you doing that the other companies are not, the big ones?

MICHAEL DELL: If you look at our balance sheet, it’s fairly different from other companies. Probably one of the most obvious differences is the amount of inventory we have. We generally have about 4 – 5 days of inventory whereas it would be more typical to see 40 or 50 or 60 days of inventory or sometimes quite a bit more, and that adds up to be a lot of billions of dollars real fast. How does that happen? Well, what it works is that when a customer orders a product from us, and this could be a big multinational corporation or a small customer on the Internet. We receive orders everyday, about 150,000 or so computers and within a minute of so of getting the order, we translate that into the bill of materials for all of the ingredients that go into that order. Those signals are passed to our suppliers and our supplier’s suppliers, and they deliver materials into our factories every 90 minutes based on those orders. So if you went into our factory and you said, where are all the parts? Well, those are the trucks coming in with the parts. Then there is the production line where they are building the computers.  Where are all the computers that are finished? Well, they are on the trucks that are going out. 

Essentially we have a demand-pull system instead of a supply-push system and that has all sorts of advantages. One of course is that when you’re changing technologies from GPRS to EDGE to HSDPA or from one processor to another, or one operating system to another, whatever it may be, you can end up with the wrong one very quickly by the permutations and combinations with memory and network and IO and everything else.

So in the competing model, there is this process of guessing what people are going to buy and then delivering those to a store or a warehouse, in anticipation of demand. So, it’s basically guessing. And of course let’s say if you have 5,000 nodes and you guess what the demands are going to be and you put them all out there, invariably you’re wrong, right? You’ve got the wrong ones. So then you have to mark them down or move them down from one place to another, and this doesn’t work very well. And when you finally sell them, was it what the customer wanted? We just have a totally different system.