Influencers, the influenced and being half pregnant

Posted by on Jan 24, 2008 in Uncategorized | 16 Comments


Gareth was straight out of the blocks in alerting me and the rest of our community to this great piece around Duncan Watts work
on how behaviour spreads through a population and how the Tipping Point model looks much less robust than has been suggested

I’ve discussed his stuff here and here before and Gareth himself does a pretty neat summary of the implications but for those of you new to it (or who’ve been asleep at the back) here’s a short approximation of Duncan’s position:

1. Behaviour spreads through populations in a very different way from that which we assume:

NOT as Gladwell’s epimediological and information network through some individuals who are “Influential”

BUT RATHER randomly:

“If society is ready to embrace a trend, almost anyone can start one–and if it isn’t, then almost no one can”

2. The causal influence chain seems to make sense in retrospect (as the Black Swan points out) but the thing we don’t ever consider after the fact is the host of unsuccessful trends – the armies of records that didn’t sell, of fashions that never take off, of the new products that disappear without trace, the churches that never got built, the careers that never were

3. If this POV is right then the expensive and fiddly approach indicated by the suave Gladwell and the charming Ed Keller – with Influentials individuals shaping what “the rest of us” do is a waste of time and money: the revolution can start anywhere…

Now, I must declare an interest here: not only do I think Watts is right but I’ve been working with his British based counterpart to beta test our own version for marketing and communication. And we’re getting pretty close to something pretty robust (more of this later on, I promise)

Why am I so convinced he’s right? Well, beyond the difficulties in making the numbers (or the mechanism itself) work I think there are 3 or 4 broad reasons, most of which fall under this heading: the assumptions about human society and behaviour behind the Influentials kind of model just don’t stack up

1. No-one person or class of person influences all or most of an individual’s behaviour: Stephen and Fion‘s classic study showed how personal influence works in a number of different ways and varies by market and…for **** sake, few of us spend our lives stuck in one small social group away from the other folk in our society (except members of isolated religious communities and remote jungle tribes). In anycase, human social networks are not fixed (like wires on a board) but constantly shifting: each of us plays many different roles in different social contexts all of the time

2. The alternative “copying” model seems to be based in a much firmer grip on human beings and how they learn to do stuff. In particular, it places mimetic learning (learning by copying – one of our most important learning strategies from infancy on) at the heart of its explanation of how behaviour spreads. Recent work on mimesis and mirror neurons seems particularly relevant here: what makes us so “clever” is our ability to learn from each other; the same mechanisms seem central to our ability to live such a sophisticated social life

3. The influencer model reliance on verbal behaviours and what is said (recommendations, word of mouth etc) and the transmission of information seems more than is sensible or even necessary (we’ll all been told on training courses what a low percentage of a presentation is the verbal bit).

No, it’s how folk around your behave (or you perceive them to) that shapes your likely response. As Watzlawick and co articulated long ago, analogic communication (that is behavioural) tends to be a much more powerful lever to generate behavioural response than digital (i.e. information-transmission).

4. In particular, I think the Influentials kind of models appeal to us because they reflect a web of other ideas we have – they fit some of the received ideas we have about humans (rigid heirarchies, active verbally driven influence etc) and some of our longstanding marketing practices (targetting on a priority audience, send them messages etc etc). I also suspect a big part of the appeal is that we’d like to think of ourselves in the Influential boxes one way or another, certainly the agency folk might fall in this trap…maybe? And finally that they give the impression that most marketing behaviour changes as a result of what the brand does to consumers (directly as in the old advertising model or indirectly as in WoM). But just because we feel comfortable with them doesn’t make them true, does it?

I remember being told by someone at the NYC book launch that “he saw what I was on about but just preferred to think about things” the Influentials way. Similarly Jon – an otherwise enlightened fella – comments today

“At the end of the day, it’s probably six of one, and half a dozen of the other”

I beg to differ. They can’t both be right. One – the Influentials one – still sits within the old world view of human behaviour as essentially a mechanical and complicated thing and worries about the levers to pull to ensure a particular thing happens; the other embraces a pile of learning about our species and in particular highlights the complexity of humanity and human behaviour: it quite happily acknowledges the truth we all resist – that it’s really hard to make human beings do a particular thing.

And whether you “prefer” one model or another is not really the point, is it? It’s the facts of how each works (for example, the numbers of folk like Duncan) and the ease and utility of applying these models in the real world that we should be considering, isn’t it? How you’d “like to think about it” isn’t going to get you to the best result now, is it?

Think this might be the year we have to choose…

UPDATE: Seth (hat tip to Mark) and Guy both have interesting things to say here. And David muses on the fact that Guy alerted him to the piece (which leads to even more interesting stuff…)

Further UPDATE: Greg nails it in his comment on David’s post


  1. John Dodds
    January 24, 2008

    I herd a stock strategist speaking on Radio 5 at 5.46 pm (you can listen again) describing traders as a bunch of lemmings who may react to catalysts but ultimately copy their colleagues. Sound familiar?

  2. Matt Moore
    January 25, 2008

    One thing about the influentials model is that it is hierarchical. At the top sits the marketer, then below them in the information pyramid are the influentials, then below them are the morass of consumers.
    The influential is like a little marketer homunculus who internalises then spreads the marketing messages.
    A non-influential model has no such hierarchy. The marketer is simply surrounded by consumers who may (or may not) respond to actions by the marketer – and more importantly by each other.
    The Alexrod & Cohen stuff on complexity and agent behaviour is very relevant here…

  3. Jon Howard
    January 25, 2008

    Implying a lack of enlightenment is a little harsh!! Wasn’t fence sitting tho – clearly behaviour is way more random and collective than the Influencer model would suggest. But at the same time, some people are clearly more influential than others (whether because of the passion, knowledge, fame, charisma, contacts or whatever). So we can’t completely dispense with it as a way of thinking. How you find those people, in marketing terms, is another matter entirely, tho.

  4. Mark Earls
    January 25, 2008

    Thanks, chaps for comments. Most useful.
    First, JD is on the money here with traders example. It’s worth remembering that this copying mechanism leads to both big fat pension funds and the panic we’ve seen in recent weeks.
    Second, Matt Is also spot on with the comment on the value judgements built into the heirarchy of influentials type models. (I edited this point from an earlier draft, I’m afraid). Think this tells us a lot about which deeply embedded ideas this approach plays to and why it’s so difficult to let go of. Top tip to the Axelrod lit btw. I’d also recommend Paul’s Ormerod take on the implications of stuff eg. Why Most Things Fail (05)
    Finally, though Jon – no offence intended but my point (and of course Duncan’s) is that it’s not born out by the data that some people are more influential than others. Most behaviour propagations are explained well enough without having an influencer in there.
    Indeed, even when we see what you might call a “prestige bias” to random copying (where random copying is distorted by copying the behaviour of individuals of high social status, let’s say) it’s not these individuals who do the influencing but the converse: its the copiers who distort their behaviour. An important distinction, I reckon.

  5. Jon Howard
    January 25, 2008

    Not disagreeing – and you know I think you’re/Watts theory seems more intuitively correct.
    Nonetheless, there must be shades of grey (rather than being a binary either/or choice)…which is something Watts seems to accept.
    Whatever the data might say, it just makes intuitive sense to me that there can be (although maybe less often than we might think) times, places and situations where some people do wield more influence, whether starting, propogating or turbo-charging an idea. The problem, I would argue (and why the Influencer theory is flawed as a marketing tool) is that this influence is completely random (where Watts comes in): there aren’t fixed and eternal influences who always influence.
    After all, if my mum had been the first champion of Herd theory and posted it on her blog (if she had one), I’m not sure it would have captured the imagination and taken flight…even if right…and even if the world of marcomms was waiting to hear.
    Whereas, your good self is better known and wields (if I’m allowed to use the I word) more influence with comms types. So people were more likely to listen.

  6. Badger Gravling
    January 25, 2008

    Doesn’t Watts own findings point to influences having an effect – just not that of starting trends?
    The Fast Co article says “in cases where an Influential touched off the trend, it spread much further”.
    And increasing the power of the influencer model makes the trend go far further.
    Anyone can start a trend. And for a trend to carry it will need certain conditions. I think Watts work and models are important in underlining that it’s not just down to the key influencers.
    But I do think his models need more refinement, and comparing, for example, 1967 letter writing with 2001 email communication doesn’t really seem a like-for-like experiment. Super connecting gateway keeps won’t have the same important to communication because the ways to reach an individual are more open than ever before.
    20 years ago, you might have found me in a phonebook, electoral roll, or via my work.
    Now you can find my blog, my contact details on countless forums and lists, my details on LinkedIn, Facebook, Myspace etc. So you don’t need to find me via a friend when a google search will let you connect.
    But if your message is reinforced by someone whose opinion I value, then it’s more likely to be read, rather than be filed as spam.

  7. Johnnie Moore's Weblog
    January 25, 2008

    The myth of someone in control

    Mark Earls has another good post on the pitfalls of seeking the “key influencers” as a way of instigating change. I share Mark’s scepticism about over-rational models for explaining how change happens, and also about the hero-centric notions of leaders…

  8. Mark Earls
    January 25, 2008

    Badger – thanks for comment.
    I think this is what DW would say (certainly our work would lead us to this POV
    1. Most behaviours spread through a population on the basis of largely random copying. This means that we all are influenced randomly by those around us.
    2. However, some copying behaviour is not random – it is distorted by prestige or biological biases and so doesn’t reflect the purely random copying we see elsewhere
    3. That said, because of the shifting nature of human social networks and the sheer number of folk we come across in our daily lives, the one individual C whose behaviour influenced individuals A and B at time T1 in context 1 does not then influence D and E because e.g. other individuals and behaviours distract them
    4. Again, it’s after the fact that the specific causal chain makes sense but things are much more complicated than they seem in retrospect. What – again – of the Dog That Didn’t Bark? The influences you didn’t act on? This is something to bear in mind around your multiple identity point.
    5. I also think we think all too often about human networks as if they were information systems: channels down which information flows which somehow(!) effects behaviour. I find this increasingly unhelpful because it enocurages us both to hold onto the “fixed wiring” metaphor and at the same time back-fill our modles with the assumption that the way to change behaviour is to give folk information rather than copying (and therefore what we need to do is give folk information – better information, in better or more impactful expressions or more targetted and timely ways).
    As you can see a lot of ideas seem to collide here and as we unpick one, the others sneak back in. It’s fun tho!
    Thanks for stopping by.

  9. Badger Gravling
    January 25, 2008

    Thanks for the response. As I commented on the Fast Company story, but forgot to put here, I think DW’s work is really important, even if I disagree to an extent, because it provides an counter-argument, and gets us somewhere closer to the truth.
    Personally, I definitely feel Influencers, are actually Magnifiers. Humans do copy what they see around them, but (and without research at this stage), are likely to consciously copy those they see as successful in some way.
    If friend A is unattractive and unsuccessful, would I copy their clothing?
    But if friend B happened to be an attractive guy, with loads of friends, and a hit with the ladies, would I copy their clothing?
    But as I said, I think it’s cloudy. For instance, the DW version would work well, I imagine, with accents. You don’t consciously adapt to talk in a certain way, but even if you’re not a fan of an accent, you’ll pick up elements of it over time.
    The only thing I actually strongly disagree with at this stage is that the effects of a Magnifier can’t exist alongside the random spreading of a trend. They don’t necessarily need a Magnifier, but it may be that the stage at which a Magnifier is involved is the stage of greatest growth and expansion.

  10. Henrik Vejlgaard
    January 27, 2008

    Thanks for alerting me to the Fast Company article. I think that it is very naïve to say the least to think that it is the same process that takes place when we see changes in style and taste take place and when changes in values, ideas, and knowledge take place. In my research I have focused on the trend process = a change in style and taste. In my new book (Anatomy of a Trend, McGraw-Hill 2008) I document both how a trend starts and how it gets momentum. In the early stages of a trend the process is non-verbal. It is about observation. I introduce the terms mono-social groups and poly-social groups: mono-social groups are groups of people who only socialize with people who are like themselves; poly-social groups are people who socialize with people who are different from themselves. There is an overrepresentation of people who are poly-social in some groups, for instance, designers, creative artists, performing artists, wealthy people, and gay men. When something new and innovative starts in one of these groups, for instance, among creative artists, the other poly-social people that the creative artists socialize with will be introduced to the new and innovative style and observe it and some will imitate the style. Then afterwards designers, performing artists, wealthy people, and gay men among their social circles will take the new and innovative out into their respective social circles where it will also be observed and imitated. Of course, there will also be talk and verbal buzz but this play a minor role. As Duncan Watts correctly has pointed out a person can only talk to so many people. But hundreds of people can observe and imitate a single trendsetter one evening, and it is because they are in a poly-social social setting they can each go out and influence (through being observed and imitated) hundreds and hundreds of other people. This “non-verbal buzz” is one of the important characteristics of the early stages of the trend process.

  11. Mark Earls
    January 28, 2008

    Thanks, Badger, for the follow-up.
    I think I’d express what you’re saying this way:
    we mostly copy those around us but sometimes, in certain circumstances those we see as particularly high status/attractiveness seem better models for us than the rest of our peers.
    Is that right?
    If so, I think the data would support this: the most important factor is random copying but sometimes it is distorted by other factors (you might call this “selective copying”) so that it’s still copying just no longer quite random (social status and sexual stuff are examples of cultural and biological biases respectively)
    I would suggest however that your emphasis on “conscious” copying is misplaced: we know from endless psychological and neuroscientific studies that most of what we do, we do without thinking about it.
    But again, it’s interesting to compare our “hunches” with the data (something I’d encourage all in this debate to do); our hunches are often misplaced.

  12. Mark Earls
    January 28, 2008

    And thank you, Henrik.
    I’m afraid I’m not in a position to comment in detail on your comment as I’ve ordered your book but not yet read it – I trust you have done the same…;-) but these thoughts occur to me all the same:
    1. I completely agree about the non-verbal behaviours being much more important than the verbal
    2. What do you make of the fact that there are always lots of wierd minority behaviours emerging and then equally quickly disappearing at the edge of any population? Why do most not get copied but a tiny minority do? Why did Crocs get everywhere last year when so many other similar shoes fail? Is this a function of the thing/behaviour or is it a function of the network?
    3. Are you falling into the Black Swan error of tracking successes/outliers only? Or in DW’s terms of looking for the significantly large tree as a cause of a forest fire?

  13. Alfonso Guerra
    January 29, 2008

    They can most certainly both be correct as they are observing different aspects of a similar phenomena: signal propagation.
    Watts’ observations of his experiments are valid only for end-to-end propagation of a *single* signal. For him to frame it as the manner in which all messages are distributed through multiple networks is both erroneous and hubris.
    Gladwell’s observations on social networks more accurately describes signal propagation with characterizations of transmitters (law of the few), signals (stickiness factor), and the transmission medium (power of context). He just didn’t correlate them with the better understood field of telecommunications.
    How did you fail to see the irony in claiming Gladwell is wrong when his model describes how information about Watts’ work got to you? Watts -> FastCompany -> Gareth -> You.
    Put another way: I am here because David Armano <- Paul Isakson <- You.

  14. Mark Earls
    January 29, 2008

    Thanks Alfonso for your interesting comment but I think we’re at cross purposes:
    1. At some level you are right about these two approaches trying to understand similar things and you could – if you wished describe the similar thing as “signal propagation”.
    However, my point here is that there’s a world of difference between information and behaviour/influence cascades. Tipping point posits that the latter follows the same mechanics as the former (hence the reference to the much debated 6-degrees experiment). They don’t; they are different kinds of things. Many people in the tech world think they are the same, but the truth is otherwise.
    2. Also, whilst practitioners may feel they have a ‘sense’ that the influencer-type model is how things work, they struggle to describe them quantitatively. Indeed, some of the best and most committed market researchers struggle to build a quantitative model for it that works at all. By contrast, the Watts-type model works well in quantitative terms and it explains an awful lot of phenomena very well.
    3. I think you’re proving my point above about distorting post-hoc: I mentioned Gareth because he posted first – the FC piece come up on my rss feeds but I didn’t have time to post on it immediately. (real A-listers like Guy Kawasaki took a few days, also)
    As it happens, I know Duncan’s work well and have been talking about it a lot before this piece; in fact have been preparing a panel session with him in NYC in the Spring. The FC piece was a good excuse to talk about the themes again. That’s how I got to him in this post.

  15. [ paul isakson ]
    February 12, 2008

    Duncan Watts vs. The Tipping Point

    Note: This post is a bit late coming. It has been sitting a draft since the second day of February waiting for a couple final thoughts. Sorry about that. I’m sure you’ve already read quite a bit on the article that spurred it, but rather than delete it…

  16. Gavin Heaton
    February 12, 2008

    This seems to me to be an argument of head and heart. In my heart I like to think that the influencer model works (ego and all that) — it makes life easier and allows me to cascade messages, court the “influential” and so on. But the Watts model makes sense metrically. I can grasp it intellectually. It is just not as seductive.
    But what are the implications for celebrity focused campaigns? I wonder how this can inform our strategic use of brand ambassadors? It seems to be a can of worms 😉