4 ms·

I find the methodology kinda dubious. For instance, I would like to see some substantiation that living within the bay area or new york are key indicators of s

by MCRed 12y ago

I find the methodology kinda dubious. For instance, I would like to see some substantiation that living within the bay area or new york are key indicators of startup success. The primary method by which startups fail is running out of capital, that's pretty much the definition of failure. Yet these two locations are some of the most expensive locations in the country. If you take a filter of all software or hardware technology startups across the US, across all states in the past decade, and then look at their failure rates, I bet you will find that the odds of success are higher outside of these two locations. NY and SF probably have a lot more total startups and thus total winners -- but that's also because they both have spigots of money flowing into a lot of bad ideas. Many of those bad ideas last long enough to be acquired, often for the team or for tools, without ever having been a success in the marketplace. These exits would be considered "success" by this analysis, but they aren't actually success. So, look at how many entities are still in business 10 years later. I bet outside of California and New York, mortality rates are actually lower. Of course, I can't prove this. But here was an opportunity to actually do some analysis and they didn't.

7Figures2Commas 12y ago
This is just a rehashing of old news. Earlier this year, Mattermark, the company that Bloomberg Beta worked with, was making the bold claim, "Mattermark Founder Prediction Is 25X Better Than Chance."[1] As I noted at the time: > Predicting who in a group of 1.5 million technology professionals is likely to start a company presents an unsupervised learning problem. Short of contacting all 1.5 million people and asking them, there is no way to confirm whether the predictions the system made are correct, so you cannot make claims about efficacy. > There are approaches used to deal with unsupervised learning problems, but there are no details in this post even indicating that the folks involved recognized they were dealing with an unsupervised learning problem in the first place. Instead, we just have claims like "While we believe the future founders group has a 17% chance — 25x higher..." for which no further information is provided. Beyond the unpublished methodology, it would be interesting to understand the data they modeled against. "Mattermark then mined publicly available information on social network sites such as Twitter, Facebook and LinkedIn to find information on one-and-a half-million professionals who were in some way connected to technology start-ups" is awfully ambiguous. What were the "in some way connected" criteria? Was there any validation of the data? Was de-duplication attempted, and if so, how was it performed? How was age determined when it often isn't displayed on public profiles? [1] https://news.ycombinator.com/item?id=7473203 https://news.ycombinator.com/item?id=7473203

by MCRed 12y ago

7Figures2Commas 12y ago

This is just a rehashing of old news. Earlier this year, Mattermark, the company that Bloomberg Beta worked with, was making the bold claim, "Mattermark Founder Prediction Is 25X Better Than Chance."[1] As I noted at the time: > Predicting who in a group of 1.5 million technology professionals is likely to start a company presents an unsupervised learning problem. Short of contacting all 1.5 million people and asking them, there is no way to confirm whether the predictions the system made are correct, so you cannot make claims about efficacy. > There are approaches used to deal with unsupervised learning problems, but there are no details in this post even indicating that the folks involved recognized they were dealing with an unsupervised learning problem in the first place. Instead, we just have claims like "While we believe the future founders group has a 17% chance — 25x higher..." for which no further information is provided. Beyond the unpublished methodology, it would be interesting to understand the data they modeled against. "Mattermark then mined publicly available information on social network sites such as Twitter, Facebook and LinkedIn to find information on one-and-a half-million professionals who were in some way connected to technology start-ups" is awfully ambiguous. What were the "in some way connected" criteria? Was there any validation of the data? Was de-duplication attempted, and if so, how was it performed? How was age determined when it often isn't displayed on public profiles? [1] https://news.ycombinator.com/item?id=7473203 https://news.ycombinator.com/item?id=7473203