Reputation Graph vs. PeopleRank

flamingoLast week TechCrunch ran a great post on Quora’s development of an algorithm to determine and rank user quality. In it, they used the term PeopleRank and credited my good friend Shervin Pishevar with the coining of the term (Go Shervin!). A number of people have asked me whether I think PeopleRank is the same thing as the reputation graph and so I wanted to write a short post explaining what I feel the differences are between PeopleRank and the reputation graph.

PeopleRank is, according to Shervin’s definition, a measure of influence. The classic PeopleRank company is Klout, which is attempting to measure the degree of social influence a user has. It’s still early for Klout (and measurement of social influence in general) but I think there is a huge amount of potential here. Marketers for years have been trying to figure out who has influence and to effectively reach them and services like Klout help dramatically with this. I see this on a daily basis with my wife as she runs marketing for one of the top video game franchises in the world. Tools like Klout are starting to have a very big impact in the world of brand marketers.

Quora’s PeopleRank has the ability to provide an even deeper set of data about a person’s influence. Quora will have the ability to start sorting expertise in the different verticals. Want to know who has the most knowledge about venture capital? Oh look, it’s Keith Rabois, Michael Wolfe and Mark Suster. Now those guys aren’t necessarily the three absolute smartest people on the planet when it comes to VC but most people knowledgeable with the industry would rank them pretty high up the list. And as more people join Quora and answer more questions a fairly accurate influence hierarchy in many verticals will be built.

So how does PeopleRank interact with the reputation graph? In my version, the reputation graph consists of all of the things we know about the people who we know. Who’s smart? Who works hard? Who is a slacker or dishonest? PeopleRank appears to be both a superset and a subset of the reputation graph. It’s a superset because PeopleRank attempts to build a global hierarchy of various forms of influence (e.g., who is the smartest in the world when it comes to advice for training for a marathon) whereas reputation graph is a more local measure that refers to what you know about the people around you. It’s a subset because a lot of the items in the reputation graph are not forms of influence. For example, Quora or Klout probably aren’t the best vehicles for determining who the most creative people or most effective salespeople in your hometown are. That will likely come from other people who are building the reputation graph as those aren’t really measures of influence or topic-specific knowledge.

This isn’t to say that hierarchies won’t emerge from the reputation graph though. Where it gets interesting is when social circles (and hence, reputation circles) overlap. I think it will be possible to build hierarchies across a wide range of dimensions (indeed, some people are already starting to do this). The hierarchies that could be built from the reputation graph will more likely be bottom-up rather than the more top-down nature of the hierarchies emerging from PeopleRank-based systems like Quora or Klout.

So in the end, I think PeopleRank and the reputation graph are two highly correlated concepts but not one and the same thing. PeopleRank is an extremely useful and powerful way to measure influence on a general or topic-specific level while the reputation graph is a mapping of some or all of the things that people think about the people who are in their social graph. Companies like Klout and Quora are the best examples of PeopleRank while companies like Honestly, Mixtent and CubeDuel are the best examples (at least at the moment) of the reputation graph in action.

26. January 2011 by Jon
Categories: Reputation Graph | 11 comments

  • Sol Eun

    Really enjoyed reading your post.

    I am building a service to do a series of behavioral analysis through graph constructed from scattered information in different social networks. Reputation graph is one of the critical piece to understand how a person behave with their peers or community.

  • http://twitter.com/jonbischke Jon Bischke

    Thanks! I'd love to chat. Can you email me at jonbischke at gmail?

  • http://www.nosnivelling.com daveschappell

    another way i've started to think of it, in terms of fast-rating services like CubeDuel, is as a distributed and crowdsourced conjoint analysis — i always loved the idea of CA, but hated the prospect of completing (or asking customers to complete) massive CAs. with tools like CubeDuel and behind-the-scenes systems such as Quora's PeopleRank, you get the benefits of CA without the tremendous sit-down-and-do-it form completion

  • http://azeemazhar.com/ azeemazhar

    Hi Jon

    Really enjoyed your discussion about PeopleRank and the reputation graph, something we find very interesting — as it is our business.

    In terms of the nomenclature – before the notion of PeopleRank gets too popular — it might be wroth figuring out the provenance.

    When we were naming the core PEER Index algorithms back in 2009, (which we we unimaginatively refer to as PEER now) we investigated PeopleRank as a name and we found it was being used and referred to, not least by my friend Ajit Jaokar who wrote about it in 2006 (http://www.opengardensblog.fut… ) .
    Perhaps Ajit should get the credit? ;)

  • http://www.briansolis.com briansolis

    Azeem, you did well with the name of the company. It is at the heart of what you're truly providing…an index of peers amongst their peers. I would like to discuss the idea of two things here of which I will also take to stage at LIFT.

    Though Ajit used the term PeopleRank in 2006, the context was different. In January 2008, I used the term in relation to Social Capital, ” It boosts the value of the content, extends the lifespan of great ideas, and also contributes to the Social Capital of all those who engage. It's people rank versus page rank evolving the democratization of content by giving people a shared voice and platform exactly where the conversations are taking place.” I later introduced the idea of the human algorithm to refer to all technology that attempted to measure or rank our places within society.

    On the second note, we should evaluate the use of “reputation graph.” The two graphs that I believe deserve our focus are those of “social” and “interest.” They are two graphs that are bona fide and actually identifiable. People are connecting with those they know, want to know as well as those who share common interests. I don't see individuals connecting solely based on reputation, nor are we tracking ties based on reputation. This is a linear vs. hierarchical discussion that I believe warrants deeper study. Reputation and stature are more closely tied to currency and capital rather than connections. I would like to talk about this further with you while I’m at LIFT.

    Jon, very well thought out post. I believe you are absolutely correct in dividing social capital (PeopleRank) and reputation. In fact, I am writing a report on this subject where I break out ranking platforms this way:

    Popularity (reach)
    Social Capital (stature)
    Reputation (interest/focus)
    Influence (cause/effect)

  • http://twitter.com/jonbischke Jon Bischke

    Amen. You've nailed it here Dave!

  • http://twitter.com/jonbischke Jon Bischke

    Thanks Azeem.

    As for the coining of PeopleRank, I make no judgments. :)

  • http://twitter.com/jonbischke Jon Bischke

    Thoughtful stuff Brian. I guess this begs the question as to whether the use of the word graph connotes the causation of relationships. The word graph does imply linkage (http://en.wikipedia.org/wiki/G…)) but I'd stop short of saying that graph implies causation of that linkage. That seems up for debate. As for people not connecting on reputation, I'd argue (in a friendly way!) that one as well. I think people do connect based on reputation and that reputation graph will make those connections more efficient. In the end, I'm not sure that reputation graph is necessarily the best term but it seems to work pretty well in communicating this concept and the definition of reputation graph being “what you know about who you know” seems straight-forward enough.

    As for your four ranking platforms, can't wait to read the report! :)

  • Pingback: Alexander6

  • http://www.brandvois.com/reputation-management/ Reputation Management

    One of the Web's largest opportunities.

  • http://flavors.me/stctechnologies/ STC Technologies

    Good informations.