I've been thinking a lot about start-ups lately, probably because I'm involved in three of my own. I've written about two of them--Black Gold Suites in North Dakota and the Frolic! play space in Brooklyn, New York. The third is a restaurant called Kobeyaki. It does for Japanese food more or less what Chipotle has done for Mexican. I have three partners in this one: Brian Konopka, who heads operations; the chef, Sal Barrera; and Brian Kelly, the business guy.

All three ventures have reminded me that start-ups are, in many ways, experiments. If you recognize this going in, you can design the experiment to ensure you'll get the information you'll need to make good decisions about the next steps.

Take Kobeyaki. Our goal is to turn it into a national chain, but a lot has to happen before that becomes a realistic possibility. We first had to determine whether the concept was any good. Although I left the menu and the mechanics of the restaurant up to my partners, I insisted on having a veto over the location. And I used it, rejecting a couple of the spots because there wasn't enough foot traffic. I was willing to spend more to be on a block that had a lot of hungry people walking around at lunchtime. Why? Because we were testing the concept, not the location. If the restaurant failed, I didn't want to be left wondering if we would have succeeded in a better spot.

We settled on a place in Midtown Manhattan, across the street from the Fashion Institute of Technology. I'm glad to report that the restaurant was a hit from the start. But the learning experience hardly stopped there. Within the first few days, we were able to identify a number of mistakes we'd made. The lunch hour, for example, was busier than expected, which meant long waits. And tracking inventory was tougher than we'd anticipated, which hindered our ability to produce timely financial statements. When we finally got them, we discovered that the restaurant was losing money, despite a high volume of business. So we began making all those little changes that determine whether you operate at a profit or a loss--such as adjusting portion sizes and using our people more efficiently.

We now know that Kobeyaki can be successful. The next step is to determine whether it can work in a slightly different location, one populated by tourists rather than students. That will be an experiment as well.