Month: November 2012

I just read a blog post entitled “You can do it alone” as well as the ensuing discussion on Hacker News. Paul Graham says:

You can certainly start a business without a cofounder. What’s hard to do, empirically, is to start one that gets really big.

Well, it’s always hard to start a business that gets really big regardless of the number of founders. Paul doesn’t address the implied question: how much harder would it be for a single founder?

Regardless of the answer, I know first hand that it’s possible to be a successful single founder of a startup. Obviously it has advantages and disadvantages when compared to having co-founders: more effort, more potential reward, always on call, no safety net, etc. Ultimately it’s a personal decision, and I won’t give a generic opinion (although I have no interest in ever doing it again).

However, investors have their own reasons to prefer startups with more than one founder. These reasons are not always obvious to founders, or even to investors themselves in some cases. Let me put on my dunce investor hat and tell you a couple of mine:

1) Single point of failure. As someone who built network infrastructure for a living, I tried to avoid having one component that would take down a whole system if it failed. When a startup depends on a single person, it is much more likely to die if this person gets abducted by aliens (unless it’s just for a couple of days, and the memories of the experiments are erased. But I digress). I often wondered what would happened to IndexTank if I disappeared, and I believe it’s very unlikely that it would have survived. Two or more co-founders are like having your online business running on several availability zones of a cloud provider. I was like the US-East (N. Virginia) AWS data center for IndexTank. We eventually distributed the service across regions, but I couldn’t clone myself.

2) Higher motivation for an early exit, or not selling. If a company with three co-founders needs to sell for X so that they all feel they’ve “solved the money problem,” then one of those founders would be happy to sell for X/3. This is clearly a risk for investors; it lowers the expected value of the investment even if the companies prospects no different than if it had multiple founders. The converse is also true: if the single founder happens to be a megalomaniac world overtaker, then it will be harder to convince him to sell even when it would represent an excellent return for investors.

To sum it up, a startup with a single founder is a riskier investment, and there is no reason to believe the return will compensate for the risk. It follows that rational investors must try to convince a single founder to get a co-founder.

This is just one more example of ways in which investors and founders can have misaligned motivations. Being honest and open about this issue is good for both parties.

 

If you use Twitter or Facebook, you may occasionally see status updates that look like: “I crushed a 7.2 mi run with a pace of 7’24” with Nike+ Running App. #nikeplus.” About a month ago I started collecting those tweets into a database, to see what interesting stats I could extract. Here are the results.

I collected about 300k tweets, of which about 85k had both the distance and the pace of the run. Some were in miles and some in kilometers, so I normalized everything to the metric system. I cleaned up some obvious errors or misuses of the system, e.g. people who “ran” at inhuman speeds such as 50 mph. Cheetahs :). I used the remaining data to plot a few charts. The first one is a histogram of distances people ran, split into weekdays and weekends:

A few insights from this chart:

• As expected, longer distances quickly become infrequent.
• People run longer on weekends. In fact, the average distance for a weekend run is 8.75 km compared with 6.47 km on weekdays.
• Certain distances seem to be particularly popular during weekends, and they look like typical race lengths: 10k, 16k (10 miles), the half marathon and the marathon.

Here is a chart highlighting the difference between weekends and weekdays. For the comparison I normalized both datasets as if I had 100k runs of each, even though obviously I collected more weekday data than weekend.

How about speeds? I tried to see if people run faster on weekends, it turns out that on average they don’t. I did expect the run length to make a difference in speeds, and it does. Here are the speeds for 5k and 10k runs, 5k runs are obviously a little faster on average.

The average speed for people who track their runs with #nikeplus seems to be around 10 or 11 km/h, which is a bit slower than I had guessed. Of course, I have no idea if this self-selected group of people who share their runs on the internet is representative of all runners. However, if you’ve ever run a popular race this curve will seem very familiar to you: the middle of the pack is always crowded, the elite athletes are few and spread out. Same for the slow runner / walkers who finish as the roads are being reopened to traffic.Can you think of more insights you would extract from the data? My code is on Github if you’d like to play with it.