Inquiring minds have asked for more insight into the math behind Vanno’s reputation trending.    To illustrate this, let’s take a closer look at Kellogg’s precipitous drop from Vanno’s top 10 after a series of strange affairs involving peanut butter, bongs and Pop-Tarts.

But first, a little refresher on what Vanno does – and doesn’t – do.      Measuring reputation involves collecting and processing opinions.    Pollsters have done this for years, but their game only works if they carefully select the people they survey.  Online, you don’t have that luxury.  If you try and put online communities to work crowdsourcing, you need a better way to process what you collect, since more often than not the crowds are actually mobs. 

Vanno sees the mechanics of online mobsourcing as akin to gossip.  People don’t, for example, wait until they’ve heard sqrtN/N <0.05  things about a person before reacting.   They start forming an opinion with the first thing they hear, and in parallel run it by their social networks – asking others what they think about the “evidence”, and from whom it came.  This forms a belief base, against which new evidence and opinion about the person are weighed.   Iterating beliefs and evidence seems hard-wired into the human brain, and has formal mathematical expression as Bayesian belief revision.  It’s how people think, it’s helped revolutionize spam and fraud detection, and it’s what underpins how Vanno measures company reputations.

So, now from the sublime beauty of Bayesian statistics to the (slightly) ridiculous case of how much peanut butter, an Olympic star gone bad, and Bill Maher’s rant against the nutritional content of Pop-Tarts damaged the reputation of the consumer products company Kellogg.

Prior to the Jan. 2009 salmonella in peanut butter revelations, Kellogg had one of the best overall reputations at Vanno – ranking consistently in the top 10.   Kellog stories at Vanno, starting during our 6 month private beta and continuing through our public launch in Nov. 2008 were very positive, mostly citing credible 3rd party sources. And Vanno users voted strongly in agreement with this positive take on Kellogg’s reputation.  The peanut butter salmonella stories had a small overall effect on their reputation – submitters and voters seemed to assign more blame to the peanut producers and regulatory agencies than Kellogg.   The math reflected this, as the belief base on Kellogg was only slightly affected.    But Kellogg dropping Michael Phelps, and the subsequent focus on the moral/nutritional equivalency of pot and Pop-Tarts, drew more negative attention and passion from users, and the belief base took a much bigger hit, dropping Kellogg to #16, and then to #68.

Now, from top 10 to #68 (and as of this writing, #83) might seem like a big drop to some people (Kellogg management, for example), but remember that we track over 5600 companies.  The power of Vanno’s approach is best seen in the ability to reveal such real-time micro-trends.   If we were using classical statistical methods, we would simply not have the statistical power in our sample of stories, votes and comments to say anything reliable about micro-trends in Kelloggs reputation.    The combination of an active online community and our Bayesian methods allows us to squeeze much more real-time insight from the stories, votes and comments we get from our users.