Data Science

Is the Tipping Point REALLY Toast - Do Influencers REALLY Matter?

michael-wu.pngMichael Wu, Ph.D. is CRM2010MKTAWRD_influentials.pngLithium's Principal Scientist of Analytics, digging into the complex dynamics of social interaction and group behavior in online communities and social networks.


Michael was voted a 2010 Influential Leader by CRM Magazine for his work on predictive social analytics and its application to Social CRM.He's a regular blogger on the Lithosphere's Building Community blog and previously wrote in the Analytic Science blog. You can follow him on Twitter at mich8elwu.



Last week, I’ve offer up some perspectives on the topic of empowerment vs. influence. But I deliberately steered clear of the discussion on whether influencers exist (if you remember, that was the last argument in favor of empowerment over influence). From the discussions that I have found on this controversial topic, most people seem to reference the 2007 report by Watts and Dodds (W&D) as scientific proof that influencers either don’t exist or don’t matter. I believe this academic report has been popularized in the industry by a 2008 Fast Company article: Is the Tipping Point Toast?. However, to truly understand the results by W&D, I recommend reading the original publication, which can be downloaded here for now (since this is copyrighted material, there is no guarantee how long it will be available).


If you find the paper difficult to read, don’t be too hard on yourself! It’s written by a social network theorist with a physics background (Duncan J Watts) and a mathematician (Peter Sheridan Dodds) aimed primarily at other rigorous academic researchers. If you managed to plow through the technical details, you will find that there is nothing wrong with W&D’s result. Their claims were simply misunderstood by people outside academia. In many cases, their claims were greatly exaggerated, and certainly over generalized beyond their valid regime. Let’s find out why...


W&D Use Steady State Cascade Size to Measure the Effect of Influencers

W&D’s study concludes that the cascade size initiated by influencers is not significantly larger than those initiated by an ordinary person. And guess what, they are right! This claim is perfectly valid under the conditions of their simulation. The key is in understanding how they measure cascade size, which they’ve described in their paper:

When all activations associated with a single cascade have occurred, its size can be determined simply as the total cumulative number of activations.

This is what physicists called a steady state measurement. It means you wait until everything that can happen has happened, until nothing changes anymore, and then you measure the total number of activations (influenced individuals). The implication of using cascade size, which is a steady state measurement, is that you lose all the dynamics and temporal information about the system (because the measurement is made only after all the changes and activities have stopped). This means you cannot say anything about the speed of the cascade, or how fast it reaches the final cascade size.


For example, consider a race of 100-meter dash between the tortoise and hare. If you measure how far they’ve traveled after they cross the finish line and stop running (reach steady state), you would have concluded that both can run the same amount of distance. This conclusion is not wrong, because both the tortoise and the hare have indeed finished the race and have run 100 meters. But without any timing information, you wouldn’t know which one was faster, or who won the race.


So it may be true that influencers and the ordinary people initiate cascades of similar sizes (i.e. they will eventually influence similar number of individuals), but that does not mean an influencer doesn’t matter. Influencers can still matter a great deal, because they can reach that final cascade size much faster than the average persons.


For example, an influencer may trigger a cascade that, on average, influences 10,000 people, but an ordinary person’s cascade can influence, on average, 9,000. This is what W&D claimed, the cascade size triggered by influencers is not significantly larger than those triggered by ordinary people (and they are right about that). However, the influencer’s cascade may influence those 10,000 people in a week, whereas it may take an ordinary person three months to influence his 9,000. This, I believe, is the single most confusing point that has led to the prevalent misinterpretation of W&D’s result.


Besides, one of the reasons that ordinary people can trigger a cascade of similar size as influencers is because the influencers are still in the network. Even though a cascade may be initiated by an ordinary person, this cascade of influence will inevitably reach some influencers, who will continue the cascade with much greater efficiency. No wonder the cascade sizes are not so different from those started by influencers. The difference between influencers and the ordinary people is time (which D&W did not address).


Aside from this misunderstanding, many people also over generalization W&D’s claims. W&D’s result is based on simulations. There is nothing wrong with simulations, and many famous scientific findings came out of simulation experiments in silico. We just need to understand what assumptions they’ve made, and not over generalize their claims beyond these assumptions.


The W&D Model is a Simulation of Interpersonal Influence

W&D stated themselves that their simulation is not a model for media influence where a single influencer has the potential to affect a large population. Quote from their paper:

Related to the distinction between personal influence and media influence are the various manifestations of Web-mediated influence, such as that exerted via Web-logs, social networking sites, online forums, and recommender systems. Although individuals can indeed gain considerable exposure for their views by expressing them online—a number of individual bloggers, for example, have gained large followings—the influence of the blogger seems closer to that of a traditional newspaper columnist or professional critic than to that of a trusted confidant or even a casual acquaintance. Thus, although the question of how different forms of influence—including traditional media, Web, and interpersonal influence—compare and interact with each other is indeed an interesting one, it is outside the scope of this article, which deals only with interpersonal influence.

W&D initially assumed that the top influencer in their network is only 4x more influential than the average person (low variance network). In an attempt to make their result more validity, they have repeated the study with a 40x difference (high variance network) instead of 4x. However, in social media, the most influential person can often be hundreds and thousands times more effective than the average.


The Influence Network in W&D’s Model is Random

W&D used random networks (both low and high variance) in their simulations, which are completely unrealistic. Both interpersonal and social media influence are constrained by who we interact and communicate with, and real life communication networks on social media are never random. W&D realized this too and said in their paper:

Another objection to the basic model is that random networks, whether of low or high variance, are rarely considered good approximations of real social networks.

But their attempt to address this object is to artificially create a group-based network by randomly assigning people to a number of groups, and then randomly connecting people within and between groups. None of their simulations were based on the structure of a real social communication network.


Binary Decision + Positive Externality + Threshold Rule

Finally, the biggest problem with W&D’s simulation is how they simulate the process of influence itself. In technical jargon, they’ve used a binary decision with positive externality under the threshold rule. But what that means in plain English, is that they’ve assumed that people will be influenced and adopt something (say a product) when the fraction of their neighbors who have adopted the product exceeds a certain threshold. For example, if my threshold is 0.5, then I will adopt the product when half of my friends adopt the product. Each person may have a different threshold, but when that threshold is reached, adoption always occurs. Clearly this influence process is unrealistic, because the target doesn’t seem to have a choice. Regardless of everything else, a target is automatically influenced when a sufficient fraction of their neighbors are influenced.


W&D noticed this problem as well, and they tried to address it by using a completely different model, known as the SIR model, to represent the influence process. But this model only addresses the threshold rule by using a concave and monotonically increasing influence response function. This model still exhibits positive externality, where the target’s probability of being influenced is determined by the number of influenced neighbors.



Since W&D’s simulations were conducted under many unrealistic conditions, their results should not be applied beyond the scope of their simulation assumptions (or similar conditions). Trying to generalize the limited scope of their claims to social media in the real world is, in actuality, a huge leap of faith. However, within the scope of W&D’s simulation condition, there is nothing wrong their claim that influencers trigger cascades of similar size as ordinary people. Just remember, this doesn’t mean influencers don’t make any difference. They do matter.


Even though ordinary people can eventually influence roughly the same amount of people as influencers, influencers can achieve that result much faster. This can bring tremendous value to a firm due to the time value of money. We have published a whitepaper that quantifies the WOM value derived from influencers. Compared to random people, influencers can bring 50% more values to a firm through their WOM influence.


So, do influencers exist? You bet!

We see them in our data all the time, because we have ways to find them reliably.


Do they matter? Definitely!

They may not matter in terms of the eventual number of people that they can influence, but they definitely matter in terms of how fast they can influence these people. So if you only care about the amount of people being influenced, and not how long it takes, then influencers may not matter that much. But if you don’t want to wait a life time for the word-of-mouth to reach your target, you better seek out the influencers for help.


Finally, I must say that I have great respect Dr. Watts. He and his advisor, Steven Strogatz (another great mathematician), formalized mathematically and made famous the small-world effect (a.k.a. the six degrees of separation). W&D’s claims were not wrong. They were simply misunderstood and over-generalized due to media exaggeration. So, do you still think influencers don't matter? Although the content of this post is rather deep and involved, I certainly hope I’ve clarified some of the confusions out there. If I missed anything, please don’t hesitate to ask me. I welcome any comments and discussion as always.



Great post, as always. Had a detail question though - you introduce the idea that an influencer can impact 10,000 people in a week whereas an "ordinary person" can influence 10,000 over a much larger span of time (let's say 10 weeks). Wouldn't that mean that in the same span of time, the influence could reach 100,000 and the ordinary person 1/10th that amount. 


Is the data here centralized around influence on a particular topic/message as a limiting factor?


I need to reread the original paper... will do so later today.

What a great post. Getting your network 'infected'  with the ideas of information that you (or your organisation) want, is the main question. Could you help people or organisations that want to model their network to spread the word/ideas?

Data Science

Hello Kammerait,


Thank you for the excellent question.


If the velocity of influence propagation is perfectly linear, what you said would be true. However, the speed of influence propagation is usually not linear. Consider an ordinary person who just slowly influencing his neighbors, they themselves influencing their and so on. This speed may be quite steady for a while, but when it reaches an influencer, it he will spread it very fast to his neighbors, and you will see a big jump there. 


Also the network may saturate at some point. Consider a network that forms a connected component, so that everyone can eventually reach everyone else. Then the maximum number of people that such influence can propagate is the population of the entire network. If that is 10,000, then there is no way to go above that. 

This limitation may be due to the precise structural configuration of the network itself, or could be limited by topic (as you mention), interest, etc.


So if you seed WOM with influencers, you get a quick big burst, and the slowly saturate as it reaches the periphery of the network. But if you seed WOM using ordinary people, you get a slow growth, then few smaller sporadic bursts (depending on how your seeds are from the influencers). And then you saturate eventually.


This is typically what I see in both simulation and real WOM propagation in our community data.


Thanks for the great question. Let me know if you have more questions after reading the W&D's original paper.


Wow, Michael. Lots to digest. Great post!


While I agree that it's possible for the rabbit and turtle influencers to have equal reach when it's all said and done, I'm not sure I agree in every case. Because of a rabbit's (or influencer's) very nature, they know how to "run" very well, and you assume they're more apt to understand the dynamics of a race. A turtle (or average user) on the other hand isn't necessarily motivated in the same way the rabbit is, and could go off on leisure tangents without putting much effort into the goal.


I also don't think the results would turn out equal if you compare based on motivation. When comparing a motivated influencer against a casual user, they'll have different goals and that alone can account for a huge difference in POTENTIAL. I know this isn't going to be true in every case, but all things equal, setting goals and strategy lays a better foundation, and with that the proactive ability to connect disparate points that reactive influence can't reach. I'd think a turtle with a goal and strategy would end up beating a casual influencer, and a casual user would be no match for a strategic influencer. Regardless of the time factor. What do you think?



In many instances, there are multiple influences/influencers that drive people to take action.  While it may be hard to sort out the relative value of each influential interaction, there's no reason to deny the importance.  Most influencers impact how a group of people think, believe, and act around a very specific subject matter.  I remember a day recently when Robert Scoble published a testimonial for an espresso coffeemaking machine  ... without much result, other than his typical followers/listeners being amused.  Influencers matter, but only with certain target groups thinking about very specific issues and in moments of greatest significance.

Excellent, excellent post. Misinterpretation of this research was becoming very problematic.  Now I know why the results seemed in conflict with what many of us have personally observed over and over again. 


I now have a go-to URL for anyone who brings this topic up in the future, thank you.


Anthony Nemelka


Teleplace, Inc. 

Data Science

Hello Miriam,


Thanks for the comment. But I am not quite sure if I understand your question completely.


In my opinion, finding the influencers is really the easy part. Keeping them engaged and keep talking about the things that you want them to talk about is the hard part. I think this is what you meant by getting your network infected with ideas you want right?


If that is the question, we definitely do have strategies to help keep influencers engaged. I've analyzed many situations where influencers programs fail. And I found that in most cases, the influencers are too far apart from the other influencers, and the friends that are within the immediate reach of the influencers are not passionate enough about the same topic as the influencers to keep the WOM propagating. They may spread the words once or twice, but it usually very hard to keep it going for a long time unless these friends are also passionate about the same topic.


So the key success factor really comes down to shortening the virtual distance between the influencers. Once the distance between them is short enough, they can essentially talk back and forth among themselves. And these ongoing conversation is what keeps the influencers engaged and keep talking. This creates an echo chamber effect that, in essence, amplify and transform the topic of discussion, because new ideas are generated from conversations all the time. Then listeners around them can intermittently pick up relevant new ideas that interest them and spread the ideas themselves.


That is the theory, and there are many way to implement the shortening of distance between influencers. Holding an event, hosting an online community, etc are all possible. So the question to your answer is YES, we definitely have strategies that sort of "get a company's customer network infected with ideas you want to spread."


Thank you again for the question. I hope I've address it properly. If that is not what you were thinking, please let me know, and I will try again.


Data Science

Hello Michelle,


Haven't seen you for a while, thank you for stopping by an commenting.


What you said is very true. I only use the rabbit and turtle race figuratively to explain the concept of steady state measurement in W&D's report. In W&D's paper, they certainly didn't distinguish influencers from ordinary people as rabbits and turtles respectively. They were simply people who have more connections and people that have fewer connections on the influence network. And W&D also didn't say that they the cascade size are the same in every case either. Their claim was that the average cascade size started by influnecers was larger, but not significantly larger than those started by ordinary people. They ran the simulation many times and the result they've claimed was based on the average. I totally agree that understanding the dynamics is of prime importance, but that was not addressed by W&D.


Concerning the point of motivation, what you said might very well be true, and it definitely matches with my intuition. But I cannot say that for sure, because I have no data to back that up, and data that quantifies people's motivation is very hard to get. But certainly from the fable "The Tortoise and the Hare" we can see that a tortoise can beat the hare if he's focused and prepared and if the hare is not motivated. But in the real world, who know? Maybe the unmotivated people never end up as influencers the first place. I just don't have any data to prove it either way.


I think it also depends on what you consider as an influencer. If you simply consider influencers as people with lot of follower, then certainly, what you say may be true. But we don't consider people with huge follows influencers. They have to meet ALL 6 factors in our influence model. Under such a stringent selection process, it is possible that the unmotivated people will never end up in the final list of influencers anyway.


Thank you for the comment, and hope to see you next time.


Data Science

Hello VineBuzz


Thank you for the comment.


What you said captured several factors in my 6-factor influence model. Namely: domain credibility, content relevance and temporal relevance. Of course, to be truly influential, an influencer need to be also have high bandwidth, channel alignment with the targets, and the target has to have confidence and trust him.


It is great that you are thinking about the target, because most people out there are not even thinking about the target when they talk about influencers.


Thank you again for stopping by and commenting. Hope to see you next time on Lithosphere.


Data Science

Hello Anthony,


Thank you for the comment.


I'm glad I have a chance to clear up this confusion and mis-interpretation. Many people asked me about the D&W study, and I've explained it over and over again at talks and conference. So I am very glad that I can write it up too.


I'm glad you find this post useful and thank you for spreading the words about the truth. Smiley Happy


Dr. Wu (or Michael, whichever you prefer),


I am afraid you have committed the same sin you preach against: oversimplification -- but to the other extreme.  I do know I am over my head here, as you are far more versed and knowledgeable that I could ever be in this topic but that never stopped me before.  Feel free to shame me back into my hole.


You are making the statement, if I understand correctly, that time matters in dissemination - else, the steady state measure would be valid.  Alas, why does dissemination over time matters (as in shorter time is better) is what you are not quite exploring in detail.  For an idea, a business campaign, or anything to reach 10,000 (or any other number) of readers or recipients - is time really the matter? If a business gets 10,000 new customers - does the time it takes to reach them matter? Well, in our biz world most people would say yes, but the question, in all seriousness, is does it?


Is time such a critical element in this measurement that if I get 10,000 new customers in 3 months at a lower cost of acquisition versus in one week at a higher or much higher cost of acquisition the former is considered a failure?  I think that time is not the only thing that matters in dissemination, but in a business environment the cost for dissemination and the trade-off (if I use an influencer for one campaign, do they lose credibility, reputation, or influence in successive campaigns?) between using "people like me" with higher credibility versus people with lower credibility but faster reach and dissemination should be considered.


As I said, I am probably making something out of nothing -- but I don't like any extreme of any concept, I prefer to focus on the correlation between metrics and KPIs to prove the value of an idea in business.


What do you think?

Data Science

Hello Esteban,


Thank you for the question. It’s a great honor to get your attention and time among your busy schedule to comment here.


Let me say that my intent here is only to clarify the misinterpreted result of W&D. The popular belief seems to be that W&D’s results suggest that influencers don’t matter. What I want to say is that it can still matter because the velocity of their influence can be very different. Whether that different is important to a business or not is up to the business to decide.


If all things being equal (including cost), I would say that time does matter. This is because the value of money is discounted over time. To illustrate this point, we can look at a more extreme case. A dollar today actually has different value 10 years ago, and 10 years from now. A dollar 10 years ago can buy a nice dinner at a restaurant, whereas a dollar today can only get you a hamburger at McDonalds (not to imply there is anything negative with McDonalds). And in the future that very same dollar will lose more of its buying power due to the natural course of inflation.


Does this matter to business? If you are a small/medium business, maybe not. But large business can gain substantial value by taking advantage of the time value of money. Some strategies include pre-payment for subscription service. Because pre-payment allows earlier and much more re-investment of the available cash, this can have significant downstream impact financially. Another example is PayPal. Although I am not an expert on this, but I heard this from some of my friends in the finance industry, so please correct me if I am wrong here. PayPay holds onto the funds in transaction for each customer for just couple of weeks. But if they do that for a large number of customers, they can essentially maintain a fairly stable cash holding that is large enough to make them a lot of money just on interest alone. There are probably a lot more of these cases, but I am not that financially savvy.  Smiley Happy


That being said, I think your point about cost consideration is very valid. If it cost a lot more to engage the influencer than the average person, then companies would need to balance the cost and benefit as with everything else. But cost wasn’t considered in this particular study by W&D either.


I certainly hope that I didn’t give the impression that time always matter. It can matter, and if all thing equal, then time does matter, especially for large enterprises that deal with large amount of funds.


I’m confident to say that you are not making thing out of nothing. If you suspect something weird, then probably other people are thinking the same too. So I’m glad you ask the question, and I hope I’ve addressed it. Thank you and hope to see you around Lithosphere later.


Hi Michael,


This is a good post and addresses what could, without proper attention, be viewed as an existential threat to the emerging science and business of social network analysis.


Here in the UK we've encountered similar debates that have originated from other contexts, most particularly the field of archaeological anthropology (believe it or not!).  Cultural anthropologists have developed a concept called 'cultural drift', which they use to explain the spread of cultural artefacts and technological innovations through palaeolithic societies.


Some of the precepts of this work have been used to analyse modern cultural data sets like Top 40 pop record lists in the US and the persistence of market leading brands or companies in a range of sectors.  The work has demonstrated that the propagation and persistence of leading cultural artefacts or memes is predictably regular and directly related to the number of competitors in a 'market'.


The general conclusion of this work seems to be that only a limited number of top 'slots' are available in any 'market'; the cultural form of any 'winner' is pre-defined and only changes or 'drifts' slowly over time; when there are a lot of more-or-less equally qualified candidates for a top slot the emergence of any particular winner is a pretty random process; the impact of so-called 'key nodes' is therefore irrelevant over long periods of time or indeed their very existent is questionable (see Bentley, University of Durham et al.)


In reflection of your central point against commercial interpretations of the Watts & Dodds' work, I think the issue of 'time', or speed of propagation, is the key weakness of the cultural drift criticism of social network analysis.  Moreover, cultural drift exponents also overlook the fact that key nodes will be operative in a network at any stage of a meme's propagation.  


A current network analysis project we are undertaking for an airline client discovered a particular meme growing quite slowly within Twitter until it reached two specific individuals who managed to rapidly 'sling-shot' the meme to a much wider proportion of the network than it might otherwise have reached over a much longer period of time.  I'm pretty sure Lithium could provide similar evidence to demonstrate conclusively that key social nodes exist and that they have unusual power to increase the speed, reach and credibility of a message's propagation.


Best wishes,


Jason Brownlee, Dollywagon Ltd





Hi Mike,


This is a great and much needed clarification on the W&D research. I specifically like the Q & A in the comments, more specifically the A on Miriam Notten's question and the Q&A with Esteban Kolsky because it touches two improtant factors:

- effectiveness vs costs in terms of time needed to achieve the desired (end)state (also referred to as ROI Smiley Wink and,

- the importance of interconnectedness between people (and influencers) in the social network


Like Esteban states, I'm over my head here. Nevertheless I think these two elements are highly connected and they lead me to conclude that it does matter to understand who are the influencers in your target's social networks because this can influence the rate of adoption over time. More so it matters to work on the characteristics of your target's social network, in a sense that people actually need to be connected (in an emotial manner even..) AND interacting for the influence to happen. Thus the main challenge is not to find the influencer (that's probably the easy part, at least in online social networks) but to establish and maintain a community where people actually interact.


Now, in terms of ROI, where you need to invest your time (and money) most depends on the context & goals for which you are working. E.g. if you have a highly engaged community around your "social object" but somehow it doesn't seem to grow where you want it to grow, you may want to find out who is and influencer in your target community and try to engage him/her in your own community, or you may want to find out who is the connector between your community and your target community and engage him/her to expand yours..


In the real world there are really a lot of factors that influence how successful your approach is. My take here is: don't just rely on influencers in isolation, focus on the entire network dynamics, analyse in there what's needed to achieve your goals.. your success (and the time needed to get there) will highly depend on it.


Thx for a very insighful post.. it helped advance my thinking a lot.



This discussion is interesting. As a sociologist I've been interested in Duncan Watts work for some time. For those who want to read more of his thinking you can find most of his published works at his Yahoo research page. 

Community Management
Software Development
Customer Service
Social Media



Taking your point one small step further, there is no reason to assume that an ordinary person's network and the network of an influencer contain an equal distribution of ordinary people and influencers. There may well be reason to assume that an influencer will have more influencers in their network. Even one additional influencer will have an impact on both the speed and perhaps the persuasiveness of the information distribution.



Data Science

Hello Jason,


Glad to see you on Lithosphere. I still remember some of our early interactions on LinkedIn. Thank you for stopping by and commenting.


And thank you for telling me this very interesting debate from the archaeological anthropology community. This piece of history would have never enter my radar if you didn't mention it.


After hearing it from you , I do see why this debate would arise. The time scale that archaeology and anthropology deal with are usually very long. And if you wait long enough, then the effect is like waiting for the system to reach steady state (In fact, physicist's formulation for steady state is usually to take the limit as time approaches infinity). So at such long time scale that is of concern to archaeologist and anthropologist, maybe it is valid to conclude that the "impact of so-called 'key nodes' is irrelevant and their very existent is questionable." But at shorter time scale, that may not be the case.


I assuming you are analyzing SNA periodically to get the temporal information, because typical SNA only takes a snap shot in time and analyzes the network structure at a particular point in time. But doing this over a series of time points, you get the dynamics. That is the approach I've implemented here at Lithium. And yes, we definitely see influencers who can drive WOM propagation with extraordinary speed as in my reply to Kammerait.


If you don't mind (assuming that you are not bound by any NDA etc.), would you share a little more detail on the story of your twitter network analysis for your airline clients? I think your result on meme propagation is a great story that many others can learn from.


Either way, thank you again for the very interesting history, and thank you for coming to Lithosphere. Hope I will see you next time.


Data Science

Hello Wim,


Good to see you back. I'm glad that you like the post and the Q&A in the comments. We all learn from each other through discussion.


You made very good observations. I've been telling people over and over again that finding influencers is actually not that hard. Once you have the necessary data and follow the 6-factor framework, you will find them. Most people are having a hard time finding influencers because they simply don't have all the necessary data, or they may have the data but don't know how to use them.  Smiley Happy  But really, identification of influencers is pretty mechanical, that is why I can write an algorithm to find them. However, working with them and keep them engaged and passionate is way much harder. Although we have strategies, but it is a process that is not easily put into an algorithm for automation (so to me, that is much harder).


And I definitely agree that you should have a holistic perspective on any social media strategy. Overly focused on just one thing, whether it is influence, empowerment, internal collaboration, cultural transformation, etc. is likely not a good idea. In the real world, there is rarely one thing that fixes all. So it is crucial to have a clear objective, because base on the different objectives, you may need to focus on different things.


Finally, I must say that I've learn a lot from our discussion about influence and empowerment too. I've written a short summary reply to Thierry with a new picture of where influence and empowerment can operate. Of course this is not a complete picture; it is only on influence and empowerment, since that is the subject that started all these discussion.


Thank you again for the comment and see you next time.


Hi Michael,

Excellent post. I don't have much to say as we discussed this exact topic a couple of weeks back when I was in SF :-). I agree with you whole-heartedly and just want to reiterate some of the points you made that I strongly believe in too.



1. Influencers do exists and they do matter in business.  And that both influence and empowerment can add value to a firm's strategy.



2. Simulations and modeling (whether a process or customer behavior) are always good as it provides insights you would not otherwise glean from the real world (or not easy to glean). However, when it comes to applying the learning from those simulations/models one has to very very careful of the assumptions and context under which those were conducted. As is very clear from your writing, one of the drawbacks of simulations is that the validity of results are bound by the constraints.



3. The temportal component is very key to business (btw, loved your tortoise and hare example).  And I am not just speaking from an accounting perspective. As an example, a firm might want to use influencers to improve sales of a product. Since influencers can reach their "steady state" faster, they might be able to send qualified leads at a larger rate and quantity. By not using influencers, the firm might eventually get the same number of leads but a time will come when the original product is no longer the focus and a new product has replaced it because of innovations, tech advances etc. and so the 'law of diminishing return' comes into play. Of course, there are cost/benefit considerations on the worth of engaging an influencer...



4. Theory is quite different than practice. As you underline, there are lots of insights we can get from the Watts & Dodds and other literature but creating an influencer strategy is an exercise in itself. There is the distance between influencers as you mention in one of the replies. Other factors such as number of core influencers in relation to the universe, span of the influencers, the nature of the network & product (selling an Airbus vs trying to sell deodorant) etc. can all play a role in the selection and engagement of influencers for your purposes.



Again, really enjoyed the read. Great work.





First, I must thank you for being able to distill this research down into "plain English" that non-mathematicians like me can get our arms around.  This is critical to advancing the applied value of the concepts. Hugely helpful!


So with that distillation in place, I head to that applied value.  First, I do see the value in the time dimension of influence.  I differ from Esteban a bit on that point in that time to market has value.  Yes, you could argue that 100 customers is a 100 customers.  It doesn't matter ultimately when you get them.  But this doesn't account for such things as competition.  I've been out of B-School for a looong time, but I think first mover is still a valid competitive strategy.  The longer it takes to "acquire" those customers, the more you expose your self to competitive infiltration.  If your business processes and organization can support a faster growth curve, that has advantages. 


Marketing messages, while not universally, also have a shelf life.  The longer your message ages in the market, the longer it takes for your message to penetrate the market, the less effective it becomes- frequency and reach.  And I'm not convinced that speed of acquisition, when we're talking about delivering messaging through social media, necessarily translates into higher cost of customer acquisition.  Isn't that a core tenant of  of SM/WOM marketing?


Where I do whole-heartedly agree with Esteban is that credibility is also a variable when evaluating the monetary value of leveraging influentials.  


Question: Do W&D assume credibility as a given function of higher influence?  In other words, is higher influence not possible without higher credibility? So, therefore it's not an independent variable that needs to be measured? 


With all that, I come back to the application of this in quantifying the value of influencers to the bottom line.  Perhaps you address this in your while paper (which I will read) but I'll ask it anyway.  When you state "Compared to random people, influencers can bring 50% more values to a firm through their WOM influence."  Is the value you refer to revenue dollars? EPS? Valuation?  So, could you take a number like CLV and use this 50% or some other multiplier to determine the monetary impact to the business?


Thank you again! (P.S.. if Wim and Esteban are over their heads, I'm at the bottom of the sea Smiley Happy 


Data Science

Hello Larry,


Thank your for the comment. It's so nice to have a professional sociologist stopping by.


Thank you for the link. I actually read a lot of Duncan's work there. I am most impressed with his early work with Steven Strogatz and when he was at Columbia. He is a great sociologist and have a deep understanding on networks. Definitely a prolific writer who has published a lot of peer reviewed papers. Great resources.


BTW, thank you for the retweets to spread the word about my work. Hope to see you around next time.


Data Science

Hello Robb,


The distribution of users around ordinary people and influencers can certainly play a role.


This point, however, has been address by W&D. The users in their network constitutde a continuous range of influence value, and they took a cut arbitrarily, and choose the top 10% to be the influencers, and set the remaining are ordinary people. So the distribution of around influencers and ordinary people are not uniform. But the problem, as I mentioned in my post, is that they are connected randomly. So influential people do not have a higher affinity to connect with other influenctial people, they are just connected at random. So when they repeat the simulation over and over again, it essentially wipe out these local distributional differences.


But the conclusion they are trying to get to is the average case scenario, so I believe their approach is valid. In real life, the connections between users are not random. It is definitely plausible that influential users have a higher affinity for each other. And that can certainly play a role in making the any WOM message more persuasive.


Thank you again for the comment. Hope to see you next time.


Data Science

Hello Ned,


Thank you again for stopping by and commenting. It's great meeting your couple weeks ago IRL when you visit SF. And thank you for summarizing some of the points I covered.


The temporal dynamics seem to be something that is often overlooked in business for some reason. For example, the timing (temporal relevance) factor, one of the 6 factors of social media influence, is often overlooked. They are ignored even by companies that solely does influence scoring. Clearly, this W&D study is the same. In physics and math time is usually the first thing that we think of.


For the tortoise and hare example, you have to thank our community admin, PaulGi. My original article didn't have that paragraph, but he suggest that I should add some real life example as analogy in there. Thank you Paul, people love your suggestion.


To your point #4. The proper distance and number of core influencer in a community to keep a healthy and diverse dialog to keep everyone engaged is a very challenging problem. I've look at the data a bit, but so far I haven't been able to identify any consistent patterns yet. Definitely will need to do more data analysis.


You sum up everything so well that I don't have anything to add anymore. Hope to see you again next time.


Data Science

Hello Barry,


Thank you for coming back and commenting. I’m glad that I was able to clarify some confusions. I just wish whoever that is interpreting W&D’s study initially were a little more scientifically savvy. 


Totally agree with your point about the value of time-to-market. I often told my academic friends that in industry we optimize under a very different set of constraints then in academia. Academics usually optimized for quality and rigor, because time is virtually infinite. But in business, time become very important, and in an attempt to have a better time-to-market, we are often forced to make the tradeoff to sacrifice the quality a bit.


Time definitely has value. Even when all things are well controlled, monetary value is constantly discounted in time due to inflation. This is probably more a finance concept. I like your business argument about market competitiveness though. I will definitely use these concepts when I explain this in the future.


I think that if people are treating celebrities as influencers, which I don’t think they should be (see my earlier post The Fast Influencer Myth), then it may cost more to engage these PR influencers. But if we identify influencers using the 6 factors, then they will usually be narrow down sufficiently that the highest bandwidth users (usually celebrities) are filtered out. In that case the cost probably won’t differ significantly.


Credibility definitely plays a role in influence. It is the first factor in my 6-factor model of social media influence. But W&D did not model credibility in their simulation at all. In their simulation, they assumed that the only difference in individuals’ influence is due to his connectivity within the network (specifically his degree centrality).


Concerning the value gain from WOM seeding program. The 50% increase in value gain from seeding with influencers is the final ROI gain. In fact, we simulated the customer lifetime value (CLV) that a companies can gain when seeding randomly or seeding with influencer. Seeding with random users can already give you some ROI, say 60% (that is you gain 60%; if you spend $10,000, you get $16,000 back). But seeding with influencers can get you 50% more ROI than seeding with random users. That means instead of getting 60%, you get 90% ROI (If you spend 10,000, you get 19,000 back).


As you mentioned, there is also a huge first mover advantage. I will let you read about this yourself in the whitepaper.


Finally, thank you for the wonderful discussion. As I said, this is how we learn from each other. Thank you for asking the question. I’m certain it clear up some confusions for others. Hope to see you next time.


Hi Michael,

On seeding with influencers, have you/Lithium done any research (or know of any) on the impact of different seeding strategies?


Here is my gut take on this. Say you have ten influencers in your space. Depending on your goal, needs, product, and other organizational and marketplace contextual factors my thoughts are that the firm might want to have a clear strategy on how you want to use these influencers.


On one hand you can 'seed' with all the ten at the same time so that the cascades are compounded (of course, there will be span overlaps where someone is 'influenced' by more than one influencer). Or one can decide to seed a smaller set and then as with the case with product life cycles, seed another set as the 'influence' from the first set gets to a steady state (maturity).


Anyway, was wondering if you are aware of any resutls for these kind of variations.




This article does a good job in breaking down and analysing the original one which suggested that the concept of influencers is toast, but in my opinion it does little to show the proof that influencers exist, or how the author had proved that they exist.  Michael has kindly pointed me to more of his articles on the subject (, which I will carefully read through later. It would have been nice - and saved a misunderstanding! if a reference had been placed in this article (if there is, then, Michael, my apologies).

My personal feeling on the subject is that everyone influences everyone else; it's just a matter of degree, and depends on the subject. I'll study more.

Data Science

Hello Ned,


Thank you for coming back and asking this great question.


Concerning seeding, we have done some preliminary studies on them, but not to the extent and rigor that I like. But I’m happy to share some results on this.


We found that seeding less than 5% of the potential customers is sufficient to get 80% of the potential value of a seeding program. However increasing the seed population has a diminishing return due to the overlap of the seeds’ social network. That was how much we know a about seeding empirically.


Concerning your question. The following answer is just something that I would try. It is not based on any data or empirical result.


If you have 10 iphone to give away, and you have found 100 influencers that are equally influential (this is a big assumption), then you can probably maximize reach by picking 10 out of those 100 that has the smallest overlap among their social networks. However, if the product is expensive, and your targets typically need opinion from several persons, then maybe it is worth it to allow some more overlap.


If your WOM campaign is not time critical, and you can seed over a long period of time, then maybe you can even seed 1 and watch how their WOM spread on twitter or FB, then pick another influencer that near the periphery of the reach of their WOM. This allows some back propagation of WOM for reinforcement, but also pushes the WOM further. But this will depend on product. Certain products requires a lot of reinforcement, where as others do not.


Again, these are not proven strategies, so please don’t quote me on this. It is just something that I would try and study when I get a chance.


Thank you for the question and see you next time.


Data Science

Hello Nigel,


Thank you for continuing our twitter conversation here. Doesn't it feels great that we don't have to be limited by the 140 characters.


You are right, it's my bad for not providing a link to my earlier posts on influencers in this article (although there are many links in the comment discussions), sorry about that. I shouldn't assume that most reader follow my work so closely.


However, I do like to point out one thing. Most people were not questioning the existence of influencers initially. It matches their intuition that influencers should exist. And it is natural that some people are more influential (in fact much more) than others.


It was only after the idea that influencers don't matter has been popularized that people start to question the existence of influencer. This is because the definition of an influencer depends on their ability to influence. The thought was that if what people thought were influential are actually not reaching and influencing any more people then the ordinary, then may be they shouldn't be call influencers the first place.


This is what led to the idea that maybe there is no such thing as influencers, maybe they don't exist at all. But the real confusion is whether influencers matter or not. If influencer matters, then they exist, because the definition of an influencer is based on the fact that they matters. If they don't matter, then they are not different from ordinary people, and therefore they don't exist.


That is why I focus my discussion on the more fundamental problem of whether influencers matter or not.


Hope this address your concern. Thank you for the comment, and hope to see you again.


I am chairing a panel discussion on this topic -- defining and measuring influence -- at the Institute for Public Relations Summit on Measurement, Oct. 6 - 8, in Portsmouth, New Hampshire (it's only about an hour from Boston), with panel participants from Yahoo!, GE, and Forrester Researc. See the link below. It is always a great conference in a beautiful location.

Michael and all commentators -- This is an extremely interesting discussion ... is there influence occurring? Michael, thanks for initiating the post. Could you encourage Watts and/or Dodds to respond? Let me enter the discussion from the business end. As a CVMO, for example, I would define influence as the ability to provoke a change in awareness, attitude, and ultimately behavior. Your influence defined by the propagation of blog posts or tweets is all very interesting, but is this leading to behaviors that I, the CMO, care about? Where should we be measuring influence? Is it a property of the influencer, the message, or the recipient. W&D clearly believe that it lies in the latter. What we need to do as a professional group is move to the next step and link the propagation of information across a network to action.


Thanks everyone for the interesting discussion.

Data Science

Hello David,


Thank you for the comment


Your definition of influence is actually very much align with mine. I defined it simply as the ability to induce a change in thought or behavior. So if someone casused a change in awareness, sentiment, opinion, action, purchase, etc., he has influenced someone. So I believe this is well aligned with your definition.


Change in behavior is often easier to measure than change in thought, because behaviors are observable, whereas thoguhts are usually not observable. If you really want to get technical and academic about this topic, this can be whole separate discussion. Neuropsychologist can argue that the minute that you've become consciously aware of something, whether by reading it, looking at it, or any other form of perceptions, your mind has changed. So simply being exposed to a piece of information can influence someone. If I made a loud sound, or flash a bright light at your eyes, your behavior change by cover your ear or blick. But these are probably not the influence that CMO cares about. So to make the influence process more relevant to business, I've developed a 6-factor model that carries influence in the form of information from the influence to the target.


So most of my work on influence is based on this simple model. And unlike most models out there that either focuses on the influencer (e.g. Malcolm Gladwell) or the target (e.g. W&D), I believe it is important to consider both. So to really achieve true influence, you need to measure it along this chain of information propagation that links the influence and the target together. That is, the measurement must take into account of both the influencer and the target. Focusing on either one will inadvertently bias the conclusion.


I hope this address your question. Thanks again for the nice comment.

Thank you for this excellent article, and for resurrecting attention to the original Watts & Dodds study.


At the peril of revealing my naivete in the study of networks, I was hoping you could clarify one point.


You wrote: "Each person may have a different threshold, but when that threshold is reached, adoption always occurs. Clearly this influence process is unrealistic, because the target doesn’t seem to have a choice. Regardless of everything else, a target is automatically influenced when a sufficient fraction of their neighbors are influenced."


However, if an activation threshold is not meaningful (it does seem to be an over-simplification--I would expect influence to be qualitatively different among network participants), then how do you define "influence" in the first place, if it's not something that amounts to causing action? While it might be poignant to say that the person "doesn't seem to have a choice," this appears to be simply an inversion of the very concept of "influence" (that the target acts as a result of it).


Perhaps you are suggesting that, unlike the W&D assumption, real-life influence can't be measured simply in size of the fraction of network neighbors adopting a certain preference. This certainly makes sense intuitively, since different people realistically influence others to a different degree. However, I wonder if this is where we get into some circular reasoning--if our very assumption is that some people have more influence than others.

Data Science

Hello Boris,


Thank you for the comment. I'm glad you find this post interesting.


Let me try to clarify the issue you pointed out. The activation threshold in W&D is meaningful, but it is not the only factor at work as W&D assumed. For a simple definition of influence and how we measure it, you can take a look at my reply to David's comment above.


Adding to my earlier response is that the amount of influence a person has is a macroscopic property of the network. An individual (i.e. a target) is either influenced or not influenced, but the amount of influence a person has depends on how many can he influence. So the activation threshold that W&D proposed is meaningful, but it is a view from the target's perspective. As you said, "influence is something that amounts to cause" a change in thought or behavior. But that amount may not come from a single influencer or a single source. In reality, it probably came from many sources: multiple influencers, media exposure, etc. And these sources interact nonlinearly in complex ways that lead to the target’s mental or behavior change.


Since I can't possibly repeat everything I wrote about influence here, I will have to direct you to my earlier works on this topic.


What you said is definitely a part of it. Influence certainly can't be measured simply by the fraction of network neighbors adopting a certain preference. The influencer’s persuasiveness, credibility, his local network structure may all be different. And these factors interact with the target’s confidence, urgency, etc for the needed information in very complex ways that probably cannot be accurately described with a single activation threshold.


I hope I’ve address your question. Thanks again for the comment and hope to see you again later.



Lithy Awards 2017

The winners in digital CX have been crowned!

See the winners!!