from Hacker News

Show HN: Invitation Engine for Facebook Apps

by deepkut on 7/16/12, 3:05 PM with 42 comments

We spent a lot of time working on this, including the demo, and we'd love some feedback.

It might take a little bit of time to load if lots of people are trying the demo, so please try to be patient. We spent a lot of time to ensure you'll be impressed.

Feel free to email us at tony or chuck @graphmuse.com if you'd like to discuss further!

  • by patrickambron on 7/16/12, 3:39 PM

    I personally just don't like app invites. I don't use them, and I literally never pay attention to ones I get.

    That said, the clusters are really impressive. All of my clusters were basically just groups of people from different times/events of my life (highschool people/friends, college friends, colleagues, people I knew when I lived in Syracuse, etc). It's an obvious way to separate my groups of networks, and it was fast. That's useful in so many contexts

  • by habosa on 7/16/12, 7:49 PM

    Hi Tony.

    I just demoed the app and it's awesome, it basically identified all of my friends into perfect groups, with only a little confusion. Obvious groups like "people I met when I spent a summer at Brown" were 100% accurate and blew me away. High school and college also got their own clearly defined groups. Way to go, I hope other people are as interested as I am.

  • by far33d on 7/16/12, 5:31 PM

    It's not clear to me that the main hypothesis of this engine is correct - namely, that request recipients who are closely clustered to other users are more likely to accept the request. Do you have data that supports this hypothesis?

    That said - request recipient optimization works, and offering it as a service could be a successful business. I think your primary value proposition (higher CTR's on lower volume) is solid, but would like to see verification that your approach (using clustering) is valid for that use case.

  • by engtech on 7/16/12, 8:36 PM

    Wow. You built 10 clusters, and I could have clearly given a name to each cluster.

    As part of your demo, you should let me build a list based on each cluster.

  • by verganileonardo on 7/16/12, 5:09 PM

    I tried the demo and received this message:

    Uncaught Error. {"errorCode": 6, "errorStr": "Internal operation failure, unable to reticulate splines!"}

  • by patja on 7/16/12, 5:57 PM

    How will this be monetized? I can't imagine it being free in perpetuity if it catches on.

    What about security? Passing Facebook user access tokens over http opens them up to being sniffed on the wire, which would allow anyone to impersonate a user of your app. Maybe ssl is supported, but all of the api documentation examples use http.

  • by KingOfB on 7/16/12, 7:03 PM

    Another method of productizing the technology could be an automatic group generator. I'd pay 5$ to have all my friends merged into mutual groups. I have 500 friends and there's no way I'm going to create groups for all of them. But having them grouped would be nice.
  • by mikebo on 7/16/12, 11:49 PM

    Can you give more info on how you are clustering friends? AFAIK Facebook doesn't allow you to fetch friends of friends, so curious how you're finding overlapping connections.
  • by caser on 7/16/12, 9:49 PM

    I could see this as a better way to organize my news feed. Facebook is trying to do this a bit with "Groups @ xxxx", but this has grouped people better than that.
  • by chuinard on 7/16/12, 4:19 PM

    Nice work using arbor.js. I've been trying to get a grasp myself on how to use it. What do you think about the ease of using the framework?