Wednesday, October 22, 2008

influence influenza

Whether Influencer Listers are egoists or not, the comments on Peter Kim’s post last week questioning the self-promoting nature of said lists raised some interesting assumptions about what we expect from the construct of Influence.

To sum up the commentary:
Influencer lists provide no value because they’re simply popularity lists.
Influence is undefined and ambiguous.
Influencer marketing is ineffective, or diluted because of size and the ambiguity of lists

I’ve blogged before about my dislike of measuring something for the sake of measurement. Specifically, I’ve been pretty harsh when people make magic formulas combining a hodgepodge of variables then call it Influence. In my opinion, these approaches go wrong for 3 reasons.

The formulas:
• Lack objectivity
-- arbitrarily involve variables simply because they’re available (e.g. # friends)
Lack reliability - incorporate variables that measure the same thing multiple times (e.g. friends on Facebook + followers on Twitter + connections on LinkedIn)
Lack Validity - fail to show that they predict a meaningful behavior (e.g. “real influence,” sales, good content, etc.)

Without going into detail on psychometrics, I think others would agree there’s an abundance of digital breadcrumbs available to us… we have to start to show how they relate to meaningful constructs; influence, arguably, being one of them.

I think the call to arms today is mainly about validity: we need evidence that people are measuring what they’re trying to measure—that “influence” algorithms predict something meaningful (e.g. widget adoption?).

To be clear, our expectations for influence, influencers and influencer lists probably vary as widely as the ways they are being measured. Transparency will be key.

All this, and I still haven’t come down on the practice of influencer marketing…

16 comments:

Marc Meyer said...

I'm completely down with this. The problem is the criteria for lists moves more than amoeba. And is completely subjective-which means how much weight are you to put into someones influencer list? Again, completely subjective. I would put more stock in a list compiled by an influencer except, you don't how many people on the list are personal friends like you said. That in fact I cam into contact with last week.

kate said...

Thanks, Marc. I agree, subjectivity is an issue and exactly why knowing what type of ecological validity a list has is the solution.

As an overly simplistic example, it wouldn't really matter if people on a list were friends if they got there because they all scored high on volume of conversation about Topic X, which in turn predicted new product success in that category. In that case, it would make sense that they're friends-- similarity (subject matter affinity) breeds liking.

If we begin to figure out 'the other side of the equation', or things Influencer lists predict, we'll be able to weight lists dependent on our own objectives.

BFrench said...

Kate,

I'm so pleased that you're continuing the conversation from Peter's post. You've steered it in a good direction.

I'd like to raise two points.

First, influencer lists in social media suffer from the same problems as "top lists" put together by the traditional press, professional associations, and other vested interest groups. Social media influencer lists have inherited some bad practices from the earlier forms of media and community. As we near the end of the year, we'll see a slew of these lists (e.g. Baseline's top 50). I don't think that social media lists are any worse -- or any better -- than many of these.

2nd point: I think you've highlighted what has to be the most crippling of the short-comings -- basing lists at least in part on a credible indication of how much of an effect a person/group exerts in the world.

I can think of several ways to measure influencers in that context. But all of them mean extra work -- difficult work at that. There needs to be a compelling upside for someone to do that extra work.

kate said...

great points, thank you. Your first point, I'm sad to say I agree with. In many ways 'social media' has followed suit of traditional media and it's gotten us into lots of trouble: measurement, monetization, etc. The rules just don't apply; we've been forcing new constructs and models into old ways. Luckily, I think this is changing.

Per your second point, I think the brand monitoring research firms (Nielsen Online, TNS/Cymfony, etc.) are well-incentivized to crack the code and busy at work doing so. You're right-- it's no easy pursuit. I'm curious about your ideas. What types of real-world outcomes would you recommend?

Duncan Brown said...

Hi Kate,

I feel we’re slowly getting to the meat of the discussion on influence, so congratulations on posting some good insight.

There’s a danger that we throw the baby out with the bath water in rejecting the notion of influencers and/or Influencer Marketing. Influence can be measured, within the constraints you provide. It can be compiled independently of those on the list, it can be measured objectively (through surveys and techniques such as cluster analysis), and it can be meaningful, if set in the context of a real world decision.

On this last point, always ask: influential on whom? A list is pointless not only by virtue of the criteria (or lack thereof) used to compile it, but also because it lacks a reference point. By setting the context specifically (buying a car, buying a global CRM system, etc) influence has a useful purpose. You can then ask, who influences people that make this type of decision?

Of course, knowing who is influential is only half the battle – engaging with them is the next, and harder, step. I’m looking forward to your thoughts on Influencer Marketing, because I have some great case studies to show its effectiveness – if done correctly.

Regards,

kate said...

right. first off, thanks, duncan. influential with respect to what and "on whom," as you say, understanding that there are many dimensions to "whom." I would argue resonance makes for influence. Clearly, I skew toward the ‘Wattsian’ view on network receptivity. In other words, I think it’s about click more than pure leadership.

Furthermore, this applies to engaging as well as identifying.

Really looking forward to hearing your case studies... feel free to ping me at k.niederhoffer@gmail.com if you want to discuss in more detail.

DrMcFunkyFunk said...
This comment has been removed by the author.
Miles Ward said...

Kate, thanks for opening a topic I’ve been stewing on hard lately.
The “objectivity” component you mentioned is really important, and easy to overlook. There’s way more data in this big network we’re interacting with than people expect, or is available in easy, parsed, manageable formats. More Data is often really helpful in maintaining objectivity, because if you arbitrarily pick some subset, you’re often unwittingly tilting your results.

Duncan, what you said here is critical:
“can be meaningful, if set in the context of a real world decision.”
Context, perspective, and scope are all key. This system I’ve helped architect uses conceptual topics as the lens through which influence is understood, so you’re always aware of the conceptual context in which influence is asserted. Focusing on the data around the real, salient discussion on who recommends, who answers questions, who impacts other author’s subsequent sentiment or activities has always been more effective in our view than bulk "popularity".

Kate, "validity" is critical, you’re absolutely right. I don’t think the information needed to make valid predictions of *all* desired outcomes is available here in social media. However, there are some very tangible impacts of efficient engagement by targeting influencers within a conceptual context, which I’d love to share from a case study standpoint with you at some point.

Thanks for a great discussion, I’m excited to hear what other folks have to say!
-Miles

Warren said...

Kate, as usual great analysis and very thought-provoking. I agree that the idea of influencer lists are so subjective and usually just popularity contests.

Related to this discussion, What do you think of WOMMA's Influencer Handbook? From WOMMA, "The Influencer Handbook aims to provide practitioners with the following information:
• Definition of an influencer and influencer marketing
• Types of influencers
• Methods to engage and thank influencers
• Guidelines for influencer self-regulation
• Bibliography of influencer communication research and practice."

http://womma.org/influencerhandbook/

kate said...

Hi Duncan - I'm interested in this system you mention and would love to hear more about your definition of conceptual context. Those questions: who recommends, who answers questions, who impacts sentiment are spot on, in my opinion and naturally difficult to operationalize. Certainly all information is not available-- and furthermore that which is, needs to be filtered by a lens like 'topic'. Please do share your case study-- would very much like to discuss it.

Warren... need to get back to you on that. I'm naturally skeptical, given my stance on Influence as more of a "click." I'm critical of the overall notion of influencer marketing as it tends to perpetuate the broadcast model by merely trying to amplify a signal. I'd be more for it if "Influencers" were optimized for dialogue...

Robin Seidner said...

Kate,

As someone who comes from the measurement side (like you), I have a lot of discomfort with these lists. Influence is subjective -- just like there are certain friends whose opinion means more than others, and even more narrowly, their influence is only about certain things/topics. And, does their influence mean I am more or less likely to do something?

I am not discounting that influence - social influence - is real. I do think that online influence measures are hard to parse, and they are not as simple as these lists make it out to be.

kate said...

Hey Robin-- agreed, definitely not simple, but indeed people (we) are occasionally influenced. The trick is narrowing down the parameters so that we can start to figure out under which conditions, how, and where influence occurs. One really good way to start would be to pick an outcome and work backwards-- which variables predict the majority of the variance??? That said, let's be clear: there's the person, the environment and the interaction...

InformationSpan said...

Hi Kate - it's worth saying that there is a solid academic research bais to influencer networks, but I haven't seen it mentioned here.

Look up "Collaborative Innovation Networks" or "Dynamic Social Network Analysis", research by Prof Peter Gloor of the Center for Digital Business at MIT (start at http://www.ickn.org/index2.html).

I heard this presented eighteen months ago when I took a group of then colleagues to MIT on an IT "Field Trip". Among other things, Prof Gloor's results are based on hard analysis of blogs, scientific literature, patents and so on to discover real influence networks. I won't try to encapsulate the research, but it's well worth a look.

kate said...

Hi there and thanks for the great resources. I've been researching collaborative knowledge networks (CKNs, Open innovation, organizational networks, etc.) for the past month or so and am fascinated by this line of work. One important distinction I would make between these 'literatures'-- Influence and CKNs, is the unit of interest as the person vs. the meme. In influence, as we're discussing it here, the goal is typically about node identification; and, are often believed to be static. In CKNs, simulations are used to follow information dissemination, across many nodes and links. What's amazing about these simulations is that they account for the nodes being adaptive, or 'learning' over time, which is really applicable to the evolution of social networks online.

Of course this distinction doesn't always hold true-- I've come across several papers on Influence using epidemic models of flu vaccination, for example. Duncan Watts' work would also be worth calling out here.

I was not familiar with Gloor's research, but just checked out his site and blog and will continue to immerse myself in his pubs.

Thanks again-- and if there's still any question, I strongly believe academic research should serve as a foundation for our study of related constructs. I also love field trips...

Laura "Pistachio" Fitton said...

My BIGGEST questions about "influencer" lists are simple:
...influence WHOM?
...influence them to do WHAT?

Like many aspects of marcom measurement, influence measures seem trapped in the amber of an "audience reach" framework, when they really could be stepping into the modern world and looking at behaviors, actual changes and business results.

All this to say THREE CHEERS for this post and discussion. While I guess it's flattering that my handle is starting to show up on some of these lists, it has all seemed like a bunch of ego-stroking to me. Like "gee, thanks, but I'm kinda busy over here trying to figure out how to actually make things work out better..."

Well written, much needed post.

kate said...

Thanks, Laura-- Happy to know you're oriented towards behavioral outcomes. Would love to hear your thoughts on the outcomes of microsharing (recently read your report). Have you witnessed any impressive cases of business efficiency, cost savings, or otherwise due to microsharing?