This post was written by Paul Graham of Y Combinator

This is a talk I gave at the last Y Combinator dinner of the summer. Usually we don't have a speaker at the last dinner; it's more of a party. But it seemed worth spoiling the atmosphere if I could save some of the startups from preventable deaths. So at the last minute I cooked up this rather grim talk. I didn't mean this as an essay; I wrote it down because I only had two hours before dinner and think fastest while writing.

A couple days ago I told a reporter that we expected about a third of the companies we funded to succeed. Actually I was being conservative. I'm hoping it might be as much as a half. Wouldn't it be amazing if we could achieve a 50% success rate?

Another way of saying that is that half of you are going to die. Phrased that way, it doesn't sound good at all. In fact, it's kind of weird when you think about it, because our definition of success is that the founders get rich. If half the startups we fund succeed, then half of you are going to get rich and the other half are going to get nothing.

If you can just avoid dying, you get rich. That sounds like a joke, but it's actually a pretty good description of what happens in a typical startup. It certainly describes what happened in Viaweb. We avoided dying till we got rich.

It was really close, too. When we were visiting Yahoo to talk about being acquired, we had to interrupt everything and borrow one of their conference rooms to talk down an investor who was about to back out of a new funding round we needed to stay alive. So even in the middle of getting rich we were fighting off the grim reaper.

You may have heard that quote about luck consisting of opportunity meeting preparation. You've now done the preparation. The work you've done so far has, in effect, put you in a position to get lucky: you can now get rich by not letting your company die. That's more than most people have. So let's talk about how not to die.

We've done this five times now, and we've seen a bunch of startups die. About 10 of them so far. We don't know exactly what happens when they die, because they generally don't die loudly and heroically. Mostly they crawl off somewhere and die.

For us the main indication of impending doom is when we don't hear from you. When we haven't heard from, or about, a startup for a couple months, that's a bad sign. If we send them an email asking what's up, and they don't reply, that's a really bad sign. So far that is a 100% accurate predictor of death.

Whereas if a startup regularly does new deals and releases and either sends us mail or shows up at YC events, they're probably going to live.

I realize this will sound naive, but maybe the linkage works in both directions. Maybe if you can arrange that we keep hearing from you, you won't die.

That may not be so naive as it sounds. You've probably noticed that having dinners every Tuesday with us and the other founders causes you to get more done than you would otherwise, because every dinner is a mini Demo Day. Every dinner is a kind of a deadline. So the mere constraint of staying in regular contact with us will push you to make things happen, because otherwise you'll be embarrassed to tell us that you haven't done anything new since the last time we talked.

If this works, it would be an amazing hack. It would be pretty cool if merely by staying in regular contact with us you could get rich. It sounds crazy, but there's a good chance that would work.

A variant is to stay in touch with other YC-funded startups. There is now a whole neighborhood of them in San Francisco. If you move there, the peer pressure that made you work harder all summer will continue to operate.

When startups die, the official cause of death is always either running out of money or a critical founder bailing. Often the two occur simultaneously. But I think the underlying cause is usually that they've become demoralized. You rarely hear of a startup that's working around the clock doing deals and pumping out new features, and dies because they can't pay their bills and their ISP unplugs their server.

Startups rarely die in mid keystroke. So keep typing!

If so many startups get demoralized and fail when merely by hanging on they could get rich, you have to assume that running a startup can be demoralizing. That is certainly true. I've been there, and that's why I've never done another startup. The low points in a startup are just unbelievably low. I bet even Google had moments where things seemed hopeless.

Knowing that should help. If you know it's going to feel terrible sometimes, then when it feels terrible you won't think "ouch, this feels terrible, I give up." It feels that way for everyone. And if you just hang on, things will probably get better. The metaphor people use to describe the way a startup feels is at least a roller coaster and not drowning. You don't just sink and sink; there are ups after the downs.

Another feeling that seems alarming but is in fact normal in a startup is the feeling that what you're doing isn't working. The reason you can expect to feel this is that what you do probably won't work. Startups almost never get it right the first time. Much more commonly you launch something, and no one cares. Don't assume when this happens that you've failed. That's normal for startups. But don't sit around doing nothing. Iterate.

I like Paul Buchheit's suggestion of trying to make something that at least someone really loves. As long as you've made something that a few users are ecstatic about, you're on the right track. It will be good for your morale to have even a handful of users who really love you, and startups run on morale. But also it will tell you what to focus on. What is it about you that they love? Can you do more of that? Where can you find more people who love that sort of thing? As long as you have some core of users who love you, all you have to do is expand it. It may take a while, but as long as you keep plugging away, you'll win in the end. Both Blogger and Delicious did that. Both took years to succeed. But both began with a core of fanatically devoted users, and all Evan and Joshua had to do was grow that core incrementally. Wufoo is on the same trajectory now.

So when you release something and it seems like no one cares, look more closely. Are there zero users who really love you, or is there at least some little group that does? It's quite possible there will be zero. In that case, tweak your product and try again. Every one of you is working on a space that contains at least one winning permutation somewhere in it. If you just keep trying, you'll find it.

Let me mention some things not to do. The number one thing not to do is other things. If you find yourself saying a sentence that ends with "but we're going to keep working on the startup," you are in big trouble. Bob's going to grad school, but we're going to keep working on the startup. We're moving back to Minnesota, but we're going to keep working on the startup. We're taking on some consulting projects, but we're going to keep working on the startup. You may as well just translate these to "we're giving up on the startup, but we're not willing to admit that to ourselves," because that's what it means most of the time. A startup is so hard that working on it can't be preceded by "but."

In particular, don't go to graduate school, and don't start other projects. Distraction is fatal to startups. Going to (or back to) school is a huge predictor of death because in addition to the distraction it gives you something to say you're doing. If you're only doing a startup, then if the startup fails, you fail. If you're in grad school and your startup fails, you can say later "Oh yeah, we had this startup on the side when I was in grad school, but it didn't go anywhere."

You can't use euphemisms like "didn't go anywhere" for something that's your only occupation. People won't let you.

One of the most interesting things we've discovered from working on Y Combinator is that founders are more motivated by the fear of looking bad than by the hope of getting millions of dollars. So if you want to get millions of dollars, put yourself in a position where failure will be public and humiliating.

When we first met the founders of Octopart, they seemed very smart, but not a great bet to succeed, because they didn't seem especially committed. One of the two founders was still in grad school. It was the usual story: he'd drop out if it looked like the startup was taking off. Since then he has not only dropped out of grad school, but appeared full length in Newsweek with the word "Billionaire" printed across his chest. He just cannot fail now. Everyone he knows has seen that picture. Girls who dissed him in high school have seen it. His mom probably has it on the fridge. It would be unthinkably humiliating to fail now. At this point he is committed to fight to the death.

I wish every startup we funded could appear in a Newsweek article describing them as the next generation of billionaires, because then none of them would be able to give up. The success rate would be 90%. I'm not kidding.

When we first knew the Octoparts they were lighthearted, cheery guys. Now when we talk to them they seem grimly determined. The electronic parts distributors are trying to squash them to keep their monopoly pricing. (If it strikes you as odd that people still order electronic parts out of thick paper catalogs in 2007, there's a reason for that. The distributors want to prevent the transparency that comes from having prices online.) I feel kind of bad that we've transformed these guys from lighthearted to grimly determined. But that comes with the territory. If a startup succeeds, you get millions of dollars, and you don't get that kind of money just by asking for it. You have to assume it takes some amount of pain.

And however tough things get for the Octoparts, I predict they'll succeed. They may have to morph themselves into something totally different, but they won't just crawl off and die. They're smart; they're working in a promising field; and they just cannot give up.

All of you guys already have the first two. You're all smart and working on promising ideas. Whether you end up among the living or the dead comes down to the third ingredient, not giving up.

So I'll tell you now: bad shit is coming. It always is in a startup. The odds of getting from launch to liquidity without some kind of disaster happening are one in a thousand. So don't get demoralized. When the disaster strikes, just say to yourself, ok, this was what Paul was talking about. What did he say to do? Oh, yeah. Don't give up.