(The number I quoted in June is now out of date. The combined value of the top 21 companies we've funded is now slightly over $7 billion, and the average value of all the companies we funded up to summer 2010 is thus about $33 million.)
PG - What is the median valuation of those fundraising rounds? Averages are not a great metric because the few big valuation winners (Dropbox, Airbnb) skew the average big time and bring up the "normals" a ton.
If Airbnb is valued at $1B and if there are 200 YC alums who've raised, that adds $5M to the "average valuation" of each YC startup. (I know those #s are not right but just for purposes of the example).
Not sure I understand your point. I'm asking for median valuations as averages often distort reality.
Plus, "median series A returned a loss" - huh? Can you clarify?
If 5 companies have valuations of $5, $10, $15, 20, $1000, the avg is $210 million. The median is $15 million. I'd argue the median is more representative of valuations received than averages. And if you're a startup founder, the median is more useful to gauge the program as that is more likely what your valuation will be near than the average.
Average is the important metric in startup investing, because the distribution of outcomes is extremely skewed, and the median is likely a net negative outcome at re series A stage. It's those few blockbuster hits that make VC investing work.
New founders might care more about the median, but big investors, not so much.
Absolutely. But it seemed the original TC article was trying to suggest which program was best for startups. So if trying to figure out which program is better for general partners/limited partners, I agree average makes sense. For startup founders, median is what matters.
Out of a group of 10 companies, the average YC company has a top valuation of about 220 million. Paul does not give the distribution other than 'a power law' distribution.
>> If 5 companies have valuations of $5, $10, $15, 20, $1000
A more likely range of valuations out of 10 companies is:
$0, $0, $0, $0, $0, $0, $3M, $11M, $55M, $220M
The YC average is quite decent. The median result is likely to be $0.
If this probability distribution scares you, it is time to rethink startup companies.
The usual equation for startups is "This has a small chance to change the world." Fiddling around with the non-world-changing outcomes is premature optimization, since the best exits come from failing to change the world and merely creating a viable business (look at Paypal; they wanted to upend the world's financial system, and instead they had a billion-plus exit facilitating Beanie Baby trades).
The median is more representative of what a founder's expected financial return is, but many folks aren't optimizing for that. If they are, you need to look at the off-balance-sheet asset: the six-figure job offer from Google, FB, MS, Yahoo, etc. etc. etc. that is generally available to people of the viable startup-starting caliber.
Since the mean startup return is so driven by a small number of massively successful outliers, it makes economic sense that the median outcome will not look so great. (If startups presented a good chance of being as good as the next best option, plus a small chance of F-U money, nobody would work anywhere else.)
The median probably matters more to individual founders. The average matters to rational portfolio managers. As far as I know there is no portfolio strategy that rewards you proportional to your median investment. So, saying that one summary statistic is more representative than the other is only reasonable if you have some implicit reference frame.
Are there any studies that demonstrate the amount of correlation between money-raised and exit valuations?
Yes, many. That's what any study of the returns of venture funds is measuring. And since the returns of the top venture funds are consistently at least positive, we can be fairly confident that $7 billion is a lower bound on exit valuations.
While I appreciate the attempt, this was pretty shoddy data analysis. First, Crunchbase's data is pretty bad. I saw a spreadsheet floated around some time ago that had a lot more YC companies listed on it (~200) so this is def not comprehensive. I'd imagine their TechStars data is even less complete given they tend to cover TS less than YC.
In terms of the analysis, some attempt to normalize the data would have been good.
Also, time-series figures would be more interesting as it would help show which program might be gaining or losing momentum.
In general, the idea of total funding being the best metric is laughable given how a few outliers skew the data.
It would have been interesting to see how quickly companies raise after the programs conclude. In a sense, analyzing by vintage/class would be more useful.
And then to conclude with the following "Y Combinator beat TechStars in many of these metrics, but none of these numbers translate to which (if either) is the best fit for your startup. That’s for you and them to figure out."
If your going to do some data analysis, try to make it actionable/useful and stand by it or take an opinion vs just a shallow attempt at data analysis which you neuter with caveats.
That is the better question. Unfortunately the revenues of startups are usually trade secrets. Funding events are based on investors looking at revenues, so they're the best indication available.
First off, that data isn't (and will never be) available for new-ish startups.
Young, high growth startups often have irregular revenue and no profit. The early years of Facebook would've shown weak revenue and no profit. Most smart people knew that it had a disproportionate shot at big returns long before it had meaningful revenue.
It's probably a pretty good proxy. The charts in the article are dominated by a few huge investment rounds that resemble privately offered IPOs. If you think going public would be a good indicator of success, that's closer to what's going on in those charts than it appears at first.
Neat little comparison and nice tidbit of info derived.
But as a fellow data scientist, the choice of bar graphs to represent this compact bit of info really just feels like padding to the article which could really use a bit more analysis.
Obviously measuring success by capital raised is silly. Is there a better alternative? Some function of pageviews (for more 'social' companies) vs. money raised? Tough to say.
HN is more populated in recent weeks with TC titles which would make it less read by some. Maybe the name should change from Hacker News to Crunch News or Hacker Crunch. :: eyeroll ::
That is the real test, and we've published those numbers:
http://ycombinator.com/nums.html
(The number I quoted in June is now out of date. The combined value of the top 21 companies we've funded is now slightly over $7 billion, and the average value of all the companies we funded up to summer 2010 is thus about $33 million.)