What's Y Combinator's conversion rate?

By small exit, I believe the parent is referring to an acquihire with a multi-year earn out for the founders. The point of such exits is to return investor capital.

Now successful companies can provide "minor" liquidity events for founders and key employees. Depending on the level of success, the "minor" amount could still be larger than some exits.

If anyone could inform us more about these pre-exit liquidity events, such as conditions and amount for different roles, many people would be interested.

Er no that's not often true. I guess it depends on what you mean by "better" and "small" exit but often an acquihire or low cash exit isn't better than if the company ipos. Consider liquidation preferences, ratchets etc.

This seems to have changed over the last several years. YC used to be very pro-founder in this regard, from an outsider perspective.

There's a pg essay that I can't find, but in essence states: Our rule is that it's up to the founder. If they want to aim small, that's fine. We ourselves aimed small with Viaweb. We didn't want to work on it the rest of our lives. It would be pretty lame to push founders into doing something that we didn't want to do. (EDIT: Found it: http://www.paulgraham.com/swan.html#f1n)

Contrast that with the recent Paul B interview: https://blog.ycombinator.com/paul-buchheit-on-lessons-learne...

> Jasper: Sam Altman has said that the only criterion Y Combinator uses to evaluate applying companies is, “Can this be a $10 billion plus company?” like Airbnb and Dropbox. While this model works great for a fund, there is an Early Exits movement that suggests individual entrepreneurs have a much higher likelihood of success when they raise less capital and target exits in the $20 million range. What do you think this view?

> Paul: The math does not support this strategy but if other investors want to try it that’s fine. Also, it is not just returns we are looking for but really impactful companies. When you sell too early you don’t realize the full potential.

> For example, Facebook had an offer from Yahoo for a billion dollars, which everyone told Zuck to take. Fortunately, he said no. Had he said yes, it would have been another failed Yahoo acquisition and Facebook would not have nearly as much impact. The reason we have these big and influential companies is the founders believed in a long- term vision.

I think all YC companies have always aimed big, but some, like Zenter, were fine with being acquired by Google for a smaller amount. And YC was fine with that outcome in the sense that there weren't any institutional forces set up to discourage founders from pursuing it.

YC's signals now seem to say, "A small exit is equal to failure." And while that's true for YC as an investor, I'm not sure it was true for YC's institutional forces until the last few years. If you're in an environment where all of your peers consider you a failure if you sell for a few million, are you going to want to resist that peer pressure? Especially when they're your friends and mentors?

I want to be very clear that all of this is an outsider's perspective. Hopefully someone in YC will issue a smackdown if it's mistaken. I was just giving some supporting data for why people are feeling this way.

Like blizkreeg said in another parent post, minor/small exit likely means acquihire. Which is technically better than nothing but sometimes the founders get nothing out of it except maybe a pay bump compared to peers if they remain at the acquiring company long enough. That's a potential negative along with the positive.

Most small exits are in fact managed failures. Our feelings about VCs and founders don't really change this.

A metric that counts failures as successes and successes as failures is a bad metric. That's all Paul was saying.

This is absolutely false.

I am the founder of Zip Phone (S14), the only company from that batch to have had an exit so far according to yclist. This is my quick story - joined YC in the summer of 2014 with a product that I had been working on in my spare time. Spent my 3 months in Mountain View, but just did not have the traction to make an impact on demo day. Discussed it with my group partners (qasar, pb, dalton etc.) and they recommended that I defer my presentation to the next batch. Came back to India to continue working on the product, and meanwhile got an acquihire offer from a local company. The offer wasn't impressive by YC standards at all, and I reached out to the partners for more advice.

This is what they basically said - Do whatever you want, we'll back you up. No ifs and buts. No conditions. Pretty much unconditional support.

The acquisition finally did go ahead, and YC essentially just got their money back. But at no point did I feel pressured to do anything. Even when things weren't rosy, YC was always there for support, but never any undue pressure.

Needless to say I suppose, but I'm a fan. I'm working on a new startup now, and hopefully will apply soon again to YC!

The OP said:

> whereas even a minor exit (acqui-hire)

How is viaweb anything like that? First off. Compared to the avg modern startup, in pretty sure viaweb took little in funding so founders likely had most of the ownership. And they got $50M in stock of a company that would keep going up for another 1.5 years. And $50M back then is like $75M or more in today's value. That's not an acquire or even a small/minor exit.

Also you said fyou money. A minor/small exit wouldn't give you that. A million or two isn't fyou money.

The dynamic you're talking about is intrinsic to the mathematics of funding startups. It's not a value judgement.

YC publishes their own database for this. http://www.ycombinator.com/companies/

Sure, but that error rate is probably relatively consistent. Probably exits aren't listed for earlier years either. Unless everyone is getting more secretive about exits, which is interesting in itself.

Anyways, maybe someone from YC will comment. They know the truth and it's a worthwhile figure to publish.

Then why doesn't the metric account for companies that have not shutdown and not exited, i.e. have grown beyond a certain value? Isn't that closer to the most favorable founder outcome?

You're excluding brand dilution as a function of class size. Knowing you are going to be one of 100+ companies going through the program each semester makes you less enthusiastic to join if you already have a very promising or successful company, and it makes it harder for VCs to determine signal; "It's from YC" stops being good enough.

A big reason for the class size explosion is almost certainly for network effects, a fleet of ready customers who support YC and drive growth out of the gates.

It creates the space for another brilliant investor to launch a rival YC that caters to a small size of elite founders. Isn't that almost inevitable?

Sure, great for VC lousy for founders. I agree with the post above. Why would great founders join ycombinator?

At least one issue is that since the survival curves are different for each batch (as established by your final 'Y Combinator conversion rates by vintage' section), your data is not i.i.d, the K-m curves are changing over time, and so the overall survival curve and its accompany CIs isn't really meaningful: it's sort of the weighted average of the nonparametric K-M curves of each batch further weighted by right-censoring of more recent batches... or something like that.

You could try instead showing stratified curves by year, or directly modeling how the hazard rates change over real time. 'survival' claims to support the latter by the 'crazy clock' method (but when I tried doing that in https://www.gwern.net/Google%20shutdowns I never could figure out how it worked).

One thing I wonder is whether YC is scraping the bottom of the barrel and this is why the conversion rates get so much worse, or if it has become more difficult in general to be acquired; if the former, perhaps including the size of the batch as a covariate would be a meaningful test? Data quality might also be an issue from the other comments, and you could try auditing it by picking a few supposed non-exits and researching them intensively to estimate how much measurement error you have.

It might also be interesting to see if the survival curves look more like type I, II, or III curves or what parametric distribution might fit best - looks like it could just be exponential/memoryless which is interesting.