TechStars companies’ hosting decisions

The other day I stumbled across a post on Hacker News titled Chart of YC companies’ hosting decisions. The post caught my attention as it’s always interesting to see the decisions other smart founders are making. Being a TechStars alum, I immediately reached out to the creator, Joel Franusic and asked if we could make a similar chart for TechStars. He agreed and after finding a list of active TechStars companies the new charts are ready.

The charts were generated by an open-source tool built by Joel called domain-profiler. It’s a neat tool and surfaces publicly available information in a really interesting way. It’s also interesting to compare the TechStars graph with the Y-Combinator graph. For example, notice how MediaTemple has a deal with TechStars and gets a much larger slice of the pie for TechStars than YC. Some other things of note about the TechStars graphs:

  • Smaller percentage use Google Apps for email
  • Greater percentage use SSL certificates
  • Greater diversity in web hosts

Joel is a startup evangelist for Microsoft and I look forward to following his work with domain-profiler and other projects.

  • Josh,

    This is definitely interesting, and something I have wondered about in the past. I organized an analysis of this type early on for Boston '09, and also included software and tool choices; these and related decisions are significant expenses in time and money for a fresh startup. I'm inclined to work on organizing this information for the benefit of future teams in both programs. Is this something you would also be interested in contributing to?

    One inconsistency: a lot of the TechStars teams were using EngineYard, which is represented as Amazon in the charts. I wonder if there is any way to resolve this issue, considering that EngineYard uses Amazon's infrastructure/IPs.


  • @jf


    Heroku is also misrepresented by my tool. To address that, they gave me the list of the IP Addresses for their front-ends. I haven't updated domain-profiler to make use of that information, but presumably I could use the same technique for Engine Yard. There might be other clever methods to use too. For example. Heroku serves an SSL certificate for "*" – maybe Engine Yard does something simlar?

    I'm also making more updates to domain-profiler. Beyond a straight-up survey, I can't think of good ways to accurately detect software and tool choices. You can make some assumptions from URL structures, Javascript library choices, etc. I've also considered scraping job posting sites to "detect" that information too…

    For as much as is practical, I'd love to contribute to your survey!