Craig Newmark came by last week and spent an hour with us and the leaders of several of our portfolio companies.
I kicked off the conversation by describing how Charlie, who was short two softball players for a game last Monday, posted on Craigslist that morning and by game time that afternoon had two ringers and a number of alternates lined up. I was astounded by that kind of liquidity - Craig, characteristically, smiled cryptically and said that he was happy to hear that the list was useful.
It was a broad ranging conversation where we talked about the challenges of managing a "community" the unusual business model of offering it for free until your users insist that you charge them, and the remarkable efficiency of user generated web services.
I spent most of the time trying to ascribe motive, direction, and strategy to the choices Craig made in the eleven years since he started the list. Craig refused to take the bait, insisting, as he always has, that he, Jim and the rest of the 21 people at Craigslist are reacting to the community, not acting on a plan.
So with little help from Craig, I am going to take a shot at why we admire Craigslist. We admire Craigslist because it is a remarkable combination of scale, relevance, and efficiency. Craigslist serves 200 (soon to be 300) cities, in 35 countries, where is presents 8 million listings to 10 million users, every month. Its relevance to any specific user is, in part, due to the network effect of this scale, but it is also driven by many subtle choices in interface, utilities, and business model. These design decisions, which Craig would argue, were driven by his need in the early days to support a growing number of users with few servers and no employees, are also the key to Craigslist's efficiency. Craigslist still employs only 21 people. For a system of its scale, I can think of no other service that comes even close to that level of efficiency.
Many of the web services we admire seem to have this combination of scale, relevance and efficiency. Google, del.icio.us, and Last FM, come to mind. They all get there in slightly different ways, but they all get there; leading me to wonder if getting this combination of things right is the critical success factor for web services.