[deepkut]

This is a great point, and I respect your skepticism.

At this point, we're in the process of confirming our hypothesis with case studies.

We did launch this clustering algorithm for our previous venture, Greekdex.com, in which we determined entire fraternities and sororities given just one user's friend graph. Clusters were "completed" over time as more users signed up.

That said, it's obvious that members in fraternities and sororities tell their fellow brothers and sisters about the site that they just signed up for. Therefore, it's difficult to say that our clustering algorithm resulted in more users signing up.

tldr; we're working on confirming our hypothesis with case studies.

[far33d]

Makes sense. In the context of auto-identifying social networks, this is very valuable - I could see glassdoor/identified/branchout finding value in this as well.

[deepkut]

Thank you. We actually considered building a recruitment platform based on clusters.

That said, we don't know much about the recruitment industry.

What are your thoughts?