Nielsen: "Our influencers, Ourselves"

Kate Niederhoffer, Director of Measurement Science at Nielsen Online spoke about a current work in progress - defining "influentials."  Is there a hotter topic in CGM for marketers?  Well yes, ROI, but we're closer to figuring this one out, or at least more people are trying.

Kate gave an academic explanation of the history of influence - from propoganda to personal influence to a renaissance of interest via online amplification.

Defining influencers happens in four steps:

1. Factors, focusing on three dimensions:  authority, popularity, virality
2. Characterization of influencers: factors, independent dimensions (low inter-item r-squared), unique combinations
3. Ensure algorithm is flexible enough to allow additional parameters, e.g. persuasiveness, linguistic style.
4. Check for face validity, e.g. peer review

The result is a final score, an "ideal conduit" that can and should be weighted depending on client goals.

Does it go deeper than this?  Probably.  But I don't have a PhD like Kate to explain everything properly here.

In a more simplified approach, Ken Cassar outlined a way for media planners to make influencers identification actionable.  Using data from the NetView panel:

1. Find users who visit niche/esoteric, i.e. "expert" web sites on topics
2. Users who spend more communication time than average (e.g. time spent, posting comments, etc.) on these sites are most vocal
3. Identify other sites visited where you can reach these vocal experts (behavioral targeting of sorts)

In an example for TV broadcasters, 4.4% of the at home population were in the influencer category.  65% of the sample was female, indexing at 124 vs. 74 male.  The 25 - 34 age segment was most influential in age demographic cuts.  The result?  USA Today rates highest for site most likely to be visited by TV watcher influentials (684 index).  [...?]

*****

I had to get back to Boston and the afternoon sessions couldn't be blogged anyway.  But during the day I met a couple people I've seen online:  Emily Riley from Jupiter and Rohit Bhargava from Ogilvy PR.  Saw Henry Copeland from BlogAds around, tweets too.

Comments

[Graham Hill]

Peter

I am beginning to wonder whether the search for the influentials isn't just more inside-out thinking from agencies in trouble.

Three core reasons.

Firstly, as Watts' work has shown, it may be that the readiness of the social network to be influenced by others is just as important as the influencers. I don't see much work being done on the characteristics of meme-ready networks nor on the mechanisms needed to make them so. And this in spite of all Rogers' work decades ago on social diffusion.

Secondly, as Kumar's work has shown, not all those likely to be influencers actually do as much influencing as is expected. His recent study looking at WOM propagation in Financial Services and Telecoms customers showed that only a minority of potential influencers actually did any influencing at all, that they won very few new customers and that the majority of the new customers won were actually unprofitable.

Finally, as Reicheld's work has shown, not everyone who might be a natural influencer is a raving fan of a company, its products or its service. That either means reducing the size of the pool to customer advocates who are also influencers, or resorting to marketing trickery to try and generate messages that influencers can play with. This doesn't make much sense in a marketing world increasingly dominated by the need to be open, honest and authentic.

I would like to see a more rounded approach to social media development, rather than the advertiser-driven approach that seems to be the dominant model at the moment. Let's not forget that the tragedy of the social marketing commons is only a few poorly-thought through social media campaigns away!

Graham Hill
Independe CRM Consultant
Interim CRM Manager

[kate niederhoffer]

Thanks for the coverage, Peter. I would just add that our goal in researching and presenting the history of theoretical work in the field was to determine whether in fact we are tapping into a legitimate construct; or, as Graham is (and others have)suggesting(ed), searching for something elusive, "inside-out," and holy grail-ish. In identifying the main dimensions of influence, we wanted to be sure we were measuring unique constructs that stand on their own legs-- i.e. things that are meaningful to bloggers, advertisers, marketers, etc. on their own: traffic/ time spent (Popularity), citations (Authority), how quickly one's meme spreads(Virality). I wholeheartedly agree with Graham in citing Duncan Watts' ideas on the readiness of the social network-- something we're looking into in terms of the "click" or linguistic synchrony between bloggers and their comments. We (at Nielsen Online) are trying hard to take a very comprehensive or rounded approach to this idea, given the type of data we have access to. Stay tuned and keep the feedback coming; we'll keep you up to date on our progress.

[Ken Cassar]

Peter,

I resent the suggestion that my methodology was simpler than Kate's. I employed BOTH multiplication and division! (kidding)

Seriously, though, the key takeaway that I wanted to leave was that some sites are more heavily composed of 'amateur influencers' (non-bloggers that are expert in subjects, and that are communicative) than others, and that panels can allow us to spot those people and identify the mass market Web sites that are inordinately composed of these amateur experts.

The thinking is that in categories where word of mouth matters, agencies should think about incorporating this kind of thinking into their media planning efforts. In the example that I presented at the CGM Summit, across the entire online population, only 4.4% of people qualified as amateur influencers in the TV category. Certain sites, though, contained much higher percentages. On USA Today, for example, visitors were 584% more likely to be expert on TV programs and highly communicative than a random member of the online population. Thus, when a TV network is advertising a program on USAToday.com, they reach a particular audience, but they also potentially reach the people that those people talk to.

To clarify the index...

((percent of visitors to site x that are 'amateur influencers')/(percent of the entire online audience that are 'amateur influencers'))*100

kc