Tuesday, March 29, 2011

In the last post, I discussed why we're in a bubble. In this post, I'll discuss why people keep founding startups.

Why people start startups

From an engineering perspective, you might rationally work as an employee in a startup under at least one of two conditions (I'll explain this below):
  1. you feel that the founders have significant skill or resources that you lack
  2. you are receiving a market-reasonable salary
I should clarify that there are a few other reasons you might reasonably work at a startup. Note that I said "reasonably", not "rationally". I'm leaving "rationally" to mean something that's purely economical.
  1. you feel that the company has a great product which will change the market
  2. you really like working with the team
  3. you enjoy working on this product
  4. you can't get a job anywhere else (unlikely)
Those "reasonable" reasons aside, let's get back to being rational. Let's review the economics of being a startup employee. You might expect two things when you're offered a job at a startup -- a salary and stock options. Typically, you might expect your salary to be reasonable -- and a percent or less of stock options; or you might forego a competitive salary in exchange for a few more stock options. It's not reasonable (I learned) to expect more than a percent or two in equity. Even if you're the first engineer, with a very low salary, you shouldn't expect more than 5% in equity. So let's go with half a percent, which is reasonable in a startup with 20 people.

If you're very lucky, the company might be sold for $100M. When you start, the company is probably already worth a fair amount -- say $20M, so the option income might be $400k. Tack this onto a slightly-below market salary of, say, $80k (for someone graduating with a BS), and you've made $180k per year. And again, that's if you're lucky. In contrast, it's not unlikely that you'd make almost this much as a late employee at Facebook, which is going to offer restricted stock units, a much more reliable way of making decent income. To be more clear: there's also a decent chance that the startup will be worth nothing in two years. Then you're out of work, you have Netscape on your resume, and you've made below-market wages for a while.

To be sure, startups aren't a bad lifestyle: you have flexible hours, you learn a lot, and there are a lot of reasonable reasons (second list, above) to work for them. They're particularly great for new grads.

But if you have the same skills and resources as the founders, or at least think you do, then it's completely irrational to work for them, when you could be taking a higher level of equity from the start. A 20% share of that $100M company is now $20M, or $5M/year. With even a 10% success rate, you're taking home an average of $500k / year, compared to your $120k from before.

My suspicion is that this explains the exploding number of startups we're seeing: many reasonably-skilled founders are finding that they don't want to work for Mark Zuckerberg: they want to be the next Mark Zuckerberg.

How will this turn out? My hope is that founders will start giving considerably higher equity to non-founding employees -- co-ops, effectively. Unfortunately, I don't have high hopes for this. First, it doesn't make sense to give high equity to employees whom you barely know. Second, distributing a company more widely gives much less control to the founders. This distributed control might lead to all sorts of infighting, indecision, and bad decisions.

3 comments:

Chong said...

nice thought:)

justanotherblogger said...

Interesting - have you ever thought about founding a start-up yourself? There are plenty of business-ideas that come to mind for your topic modelling algorithm that was mentioned in the Economist this week.

Sean said...

@justanotherblogger:

Absolutely. There are some interesting business areas already using topic models. Chomp is doing neat things with them, for example, and Xerox (not exactly a startup at this point) has done some neat work in the area of document services.

I think there's a lot of room in this area, but I'm also not married to topic models as tools for a startup.