As a journalist, I've always been fascinated by stories of business failure: As Time Inc. founder Henry Luce supposedly once said, they're usually much more interesting than stories of success.

As a small-business person, I also ascribe to the view that you learn an awful lot from failure, and that tolerance of failure is a crucial aspect of entrepreneurial business culture.

But at the same time, failure is not something an entrepreneur can give in to very readily. A certain level of blind optimism—even in the face of long odds—is often necessary to build a successful product or company. If you, as the business owner, don't believe, you can be pretty sure your employees, customers, and shareholders won't, either.

Indeed, understanding when to cut your losses and when to go all in—whether it be on a new product or service, strategy, or company—might be the most critical challenge in business.

Marc Pincus, a serial entrepreneur whose online-gaming company, Zynga, is emerging as a powerhouse in its field, touches on one dimension of this issue in a recent interview with veteran tech journalist Bambi Franciso.

Pincus' lesson from his previous company, Tribe, was to understand the numbers, listen to what they're telling you, and act quickly. At Zynga, he says: "We do tons of AB testing and we're clear about our leading indicators. We're much more disciplined about failing fast on many fronts. We try to build new game approaches all the time, but we kill them fast if they don't meet our success metrics."

As Pincus notes, this approach makes a lot of sense in the mass-market consumer Internet business, where success requires massive scale and it's possible to make an intelligent assessment of the odds of reaching goals fairly early on.

In an excellent story on Venture Beat, Mark Goldenson makes a similar point in recounting his 10 lessons from a failed startup. One of the big problems was that his company, also in the online-entertainment arena, wasn't able to react fast enough to market feedback.

Yet the "fail fast" idea goes only so far. There are plenty of cases where metrics alone won't tell you what you need to know or won't tell you quickly enough.

At my previous company, the Industry Standard, we missed our targets by a mile in the first year and almost got shut down. But we then went on to "instant success" (and then dramatic failure a few years later, for completely different reasons). We knew in our gut that we had the product right, even if the numbers didn't show it at the time.

At NewWest.Net, if we fall short of a sales goal, that doesn't necessarily mean the product is flawed, especially in this recession. When we launch a new conference, for example, many of the relevant metrics are by definition visible only after we've made a substantial investment.

We've killed products—including a major new conference—because we've missed the early targets. But in other cases we've stayed the course.

My advice on all this is twofold. First, don't give in easily if you feel the passion. A lot of companies are about more than just making a good return. They're about creating something great, something that the founders really care about and think is important. Sometimes you have to take the view that failure simply isn't an option.

Second, for any new idea that could be a "bet the company" investment—and in a small business it's easy to find yourself here—put it to a two-part test. Is it central to your passion? If it isn't, define a "go/no-go" point early in the process, even if the data aren't complete at that stage. And stick to it. There's no shame per se in killing something, even if you feel that it's "too soon."

On the other hand, if the initiative in question is central to your vision, and if the early results are ambiguous but not disastrous, then take a deep breath and stay the course. If you succeed, it will be that much sweeter. And if you fail, well, you'll have learned a lot. And if the entrepreneurs above are right, you'll be even better positioned for your next venture.

(Photograph of a casino table by Christian Thomas/fStop/Getty Creative Images)