Hitting Your Targets: Influence Analytics 4

[#1]

Michael Wu, Ph.D. is Lithium's Principal Scientist of Analytics, digging into the complex dynamics of social interaction and online communities.

He's a regular blogger on the Lithosphere and previously wrote in the Analytic Science blog.

You can follow him on Twitter at mich8elwu.

---

This is the last article of the blog miniseries on Influence Analytics. Previous articles in this series are here, and I recommend that you read them if you haven’t already done so.

1. The 6 Factors of Social Media Influence: Influence Analytics 1
2. Finding the Influencers: Influence Analytics 2
3. The Right Content at the Right Time: Influence Analytics 3

If you follow the procedures described in the above articles, you should be able to find the most powerful and relevant influencers. However, finding the influencers will only get you so far, because eventually your target has to make the final decision. This post examines the last two factors that affect a target’s likelihood to be influenced (e.g. where are they located and who do they trust most).

Alignment: Locate Your Targets

Last time, I showed you how to process social graph data so that the subsequent social network analysis can extract the most relevant influencer to your target. This requires that your influencer and target to be on the same social network. However, the target is a very large and heterogeneous group, so that some of your targets may not be on the same social media channels as your influencers. I call this misalignment (following the basic model that I presented in part 1 of this miniseries). Moreover, individual targets often use multiple channels, and their preferred channels may not be totally aligned with your influencers. With the long tail economy of the internet and so many different choices of social channels, it is likely that even though the largest group of targets are aligned with your influencers, the smaller misaligned groups may still sum up to a very substantial amount.

So it is important to know where your targets are. But traditional demographic data of your target audience is insufficient because it does not tell you what channels your targets use. Fortunately, Josh Bernoff and Charlene Li at Forrester Research have created the Social Technographic Profile, which translates your standard demographic data into the social technographic data. Although it does not give you the precise channels where your targets dwell, it tells you how your targets use social technologies.

This is often sufficient, at least as a starting point, to infer where your targets spend most of their time. For example, if most of your targets are spectators, they are probably passive members of a community and the blogosphere. If they are joiners, they are probably on Facebook and Twitter. If your targets are collectors, they probably spend a lot of time on digg and delicious. If they are critics, they probably spend much time on review sites, like Yelp, Trip Advisor, etc. Finally if they are creators, then they are probably active members of a community, the blogosphere, YouTube, Flickr, or other content sharing sites.

If you are a B2B company, don’t worry. Forrester also has a Social Technographic Profiler for tech companies. However, if the social technographic profiles are not sufficient, then you will need to do a little more market research.

Confidence: Find Out Who Your Targets Trust

If we achieved channel alignment and have powerful and relevant influencers, then the last mile of social influence is the targets’ confidence. Confidence is the factor that is ultimately going to tip the balance and convert your target. So the important question is who and what do your targets trust? In terms of social media influence, there are three sources that are worth mentioning.

1. People Trust Friends: Data from many research labs have shown that by far the most trusted sources of information are friends and peers. Most of these are people who are directly connected to the target on a social graph. This is precisely the reason why marketers are interested in social networks and the social graph data. But in order to leverage these data properly, marketers must process these data in ways that give high temporal relevance and content relevance (see part 3 of this miniseries).
2. People Trust Known Authority: Next, people also tend to trust authority figures they already know. These are usually reputable individuals in a very niche domain. However, they are usually unknown to outsiders, so they can only be identified through a domain specific social graph. Since the target as a group often consists of very diverse individuals with very different interests, finding authority figures in each and every interest group is not practical. Therefore, even though authority figures are the next most trusted source of information, the market has not figured out how to leverage this fact yet.
3. People Trust Social Proof: Social proof is actually a combination of two factors.
       (a) Independent sources.
       (b) Many of them.
   People will trust voices of strangers if they are independent third parties, and there are many of them giving a consistent message. Now, don’t be tempted to release any fabricated reviews or stories. First, it will not work. You cannot possibly keep up with the overwhelming voices of the crowd that is out there speaking the truth. Second, people will find out. If you know anything about crowdsourcing, you will know that the crowd is usually very smart and very efficient. Because it only takes one out of possibly millions of user to spot the wolf among the sheep, you can very easily squander all of the good will you have built up. Third, it is not worth it, because the cost of repairing your good will is far greater than the benefit you can reap from the temporary confidence of your target.

The Final Touch Up

Now that we know how to locate the targets and we also know a bit more about whom they trust, how can we leverage this information? Clearly, if you already have a list of influencers and list of targets, you want to match the target with the influencer that has the greatest potential of influence. And these are either people that are close to the target on the friendship social graph or known authority figures for your target. Another simple way to align your influencers with the targets is by promoting your influencers, so they become more visible to your targets across different channels. This will also boost your targets’ confidence by enhancing the effect of social proof. Moreover, if you have enough influencers, the network effect of their influence will amplify and spread.

You can be creative in promoting your influencers, but this must be done with great care. If your target feels that the influencer somehow has a vested interest in promoting a point of view and that he is no longer an independent third party, the influencer’s credibility will be greatly discounted.

A simple way to promote your influencer is to make his credibility (e.g. his reciprocity data or reputation data) visible to your target. You can also help your influencers in a subtle ways that make their content more searchable, and therefore more influential. For example, tagging and labeling their works with SEO terms. Cross post your influencers' work (with permission of course) and link back to the sources. Tweet and social-bookmark your influencers’ reviews, blogs, videos, etc., and show others how many people have benefitted from them. These simple acts will help search engines direct the targets to your influencers more effectively without turning your customers into shills.

Conclusion

This concludes my miniseries on Influence Analytics. We have come a long way. We started with a simple model of social media influence and outline six critical factors for the influence process:

1. Domain credibility
2. High bandwidth
3. Content relevance
4. Timing (temporal) relevance
5. Channel alignment
6. Target confidence

Subsequently, I’ve shown you the data and analysis that must be done in order to cover these six factors. I must emphasize that each factor is a necessary condition for influence, but none of them are sufficient! Therefore all six factors must be met to achieve true influence. Any missing factor is like a missing link that breaks the whole chain, and it will prevent the influencer from ever reaching your target.

Finally, you are now ready to embark on the journey of social media marketing. Next time I will show you how we have applied this knowledge in the community setting.

[#2]

Hi Mike,

First I'd like to compliment you on the influencer series. Job well done..

I do have some questions though, since I'm currently doing some research myself to better understand influencer theory & practice. The theme has been "lurking" in my brain for quite some time now

Here are some questions:

1st: I would be interested to understand if you also researched the effectiveness of influencer programms, and if so, what were the results? How do you measure effectiveness and how do you distinct the (relative) influence of the influencer from other influences on the influencee?

2dn: I'm also interested to understand if you researched the difference in impact of influencer programms on potential new customers and existing customers? Moreover I'm interested to understand how influencer programms have an effect on the value perception of the product/service by Customers already using the product/service?

3rd: how do you distinct the "targets" from the infuencers? Do you consider anyone who's not an influencer a "target"? Do you think the influencer is actually also a target?

Last but not least: To what extent do you take strong and weak ties into consideration when it comes to assessing who is an influencer?

These are just some questions for now.. Look forward to hearing from you.

Thx

Wim Rampen

Wim Rampen's Blog

@wimrampen on twitter

[#3]

Hello Wim,

Thank you for the compliment and for leaving me a comment. Influencer is an interesting topic, but I feel that rather than digging deep into the problem, many people are just give the clients what they want, telling them what they already know and what they want to hear.

Alright, to your questions:

Question 1. I have work with academician to measure value of WOM. I believe that the most accurate measure is the most direct measure -- sales increase. Any indirect measure: GPR, TPR, leads generated, or metrics along the marketing funnel, such as reach, frequency, impressions, etc., are less accurate. Since there is already a lot of uncertainties in the model, introducing a less accurate measure can often render the model variance so large that they are essentially useless (ie have no predictive power. Note, the model will always predict something, but you shouldn’t trust the prediction).

The result were publish in a whitepaper that we publish. In essence, targeting influencers can increase the profit up to about 50%. Specifically, a random seeding program in a competing market can achieve profit gain of 13% where as an influencer program can achieve 20%. But when there is no competing program, random seeding can achieve 80% profit gain, where as an influencer program can achieve 107%.

We must make a distinction between social media influence vs. offline influence here. Although research have shown that most of the influence occur offline, people still want to believe that social media influence works. Don’t get me wrong, they do contribute, just that the contribution is rather small compare to offline influence. If we try to measure offline influence, then the problem become intractable. So even though social media contributes a small amount of the total influence, it is the part that we can measure and leverage. And that is what make is valuable.

So if we limit ourselves to strictly social media influence, then we can tease apart the relative influence contributed by influencers from those contributed by other influencees. Since influence contributed by other influencees are happening online all the time, a controlled experiment before and after we introduce the influencers would be sufficient to tease apart their influence. The result we obtain already took this into account.

Question 2. I haven’t done any research that directly addresses the differential effect of influencer programs on new customers vs. existing customers. But we do have a few data point that is worth mentioning. We identify different mechanisms at work for existing vs new customers. These program increases profit by acquisition of new customers and acceleration of adoption from existing customers. The difference is most dramatic when there is competition. If there is no competing program, the sole influencer program can achieve 71.69% profit gain by acquisition of new customers and 35.31% from acceleration of adoption from existing customers, which sums to 107% profit gain. If there is competing programs, then there is no acquisition of new customers, so the 20% profit gain is purely derived from acceleration of adoption from existing customers.

From these result, it is clear that influencer programs certainly have an effect on the value perception of existing customers, because the company can still have a profit gain from acceleration of adoption from existing customers. 35.31% where there is no competing program vs. 20% when there is competing programs.

Question 3. In daily social contact, people influence each other. So the short answer is that influencers can themselves be targets and the targets can also be influencers. But as mentioned earlier, things are a little simpler when we consider strictly the influence from social media. In this case, the influence of the targets amount themselves are pretty small, and most of the influence are contributed by the influencers. But this is domain and temporal specific. For example, I may be an influencer for Blackberry because I have been using it for a long time and I am very happy about it, but I may well be a target when it comes to HP laptops, because my Dell laptop broke and I am considering getting an HP laptop which I have never use before. The same can be said about the effect of timing. I may be an influencer today, but a target tomorrow; and the domain can be with different brand or with the same brand.

Question 4. First, I must clarify what is weak tie and what is strong tie. Many people speak of the power of weak ties and 6-degree of separation; in reality, they meant distant strong ties.

I consider 2 persons connected with a strong tie if there is trust. If you just added a friend to your Facebook just because you knew each other, but there is no trust, I would consider that a weak tie. For example, suppose you are about to make a purchase decision. You sent out a message to all of your FB friends, and this friend that you’ve just added to your FB voices his opinion. If you still like to seek other opinion, then he’s just a weak tie. If you trust him and make the purchase accordingly, then he would be a strong tie.

Keep in mind that whether a tie is strong or weak also depends on the context. Suppose I just added some random dude I met at the party to my FB. I know nothing about this person, except that he is a camera expert. Then I would trust him when it comes to buying cameras. So when constructing the social graph about camera, he would show up as a strong tie to me. But when it comes to buying a smart phone, I would not trust him, so he would show up as a weak tie in the social graph about smart phones.

When I construct social graphs for all my research, I compute the tie strength for all connections based on the amount of conversation between the two persons, and how much other conversation there is. I will talk about this in a later blog. So I won’t describe how I construct the graph here. So in short, we do take into account of strong tie and weak ties. In fact we do much better than that. We compute the tie strength, so we can have any level of granularity we like in tie strength. We can have strong, medium, and weak ties (3 levels), or 10 levels of tie strength, or even 1000 levels if we need to. This is because the tie strengths I compute are mathematically a real number.

Alright, this reply is getting long. I hope this address all your questions. Thanks for asking them. These are good questions that I think will help many others. Let me know if you have more. I'm always happy to discuss and share my knowledge.

[#4]

In my tweet I've focused on one extractable golden nugget from your post: [Boost influencers' cred with your users by making their content more searchable, & promote it via tweets & bookmarking]. Thanks again.

[#5]

Hi Mike,

Can you explain this more?

"We must make a distinction between social media influence vs. offline influence here. Although research have shown that most of the influence occur offline, people still want to believe that social media influence works. Don’t get me wrong, they do contribute, just that the contribution is rather small compare to offline influence."

I thought that social media greatly amplified WOM and therefore would be more effective than "offline".

Walter Adamson @g2m

http://xeesm.com/walter

[#6]

Hello Stuart,

Thanks for tweeting my post and leaving me a comment here to let me know.

Yeah, promoting your influencers is a win-win strategy if people don't over do it. And it is an immediately actionable item that anyone can do once they've found who there influencers are. Great observation

[#7]

Hello Walter,

Thanks for asking this great question! I think my statement that you quoted definitely deserves some clarification.

There have been at least 4 reputable studies that independently confirmed that most of the WOM occur in the offline world, I do not doubt this result. If you are interested on how they arrive at these results, I’ve listed some reference here.

1. Carl, W. J. (2006). What’s all the buzz about? Everyday communication and the relational basis of word-of-mouth and buzz.... Published in Management Communication Quarterly, 19(4), 601-634. This is copyrighted pay-subscription material, so I cannot provide it here.

This is one of the earliest study that I know on the topic of online vs. offline WOM. Although I speculated the result, it is published in a peer-reviewed journal with solid methodology. This report claims that 80% of the WOM episodes takes place offline face-to-face.

2. There are two studies by Keller Fay Group (KFG), who specialize on WOM research and consulting. Since WOM has the greatest effect on products or services that are expensive, KFG study how business executives are influence by WOM. They also claimed that “vast majority of executive word of mouth occurs offline, with over 75% happening in-person” Later on KFG release another report TalkTrack™ Press Release May 15, 2007 on consumer WOM, which claims that 92% of the WOM occur offline.

3. Even online marketing company Razorfish reported that people trust offline friends (66% over no trust) way more than their online friends (just 9% over no trust).

4. Forrester Principal Analyst, Nate Elliott, also admit in his recent blog article that “the huge majority of users influence each other face to face rather than through social online channels like blogs and social networks.” He did not present any data in his blog, but I suspect that he have data in his report “The Analog Groundswell: Using Social Media To Create And Amplify Offline Influence” Unfortunately, I do not have the full report, so I cannot say much more about his methodology or data with regard to online vs. offline WOM influence.

But, do not be dismay. Although most WOM influence occurs offline and are have a strong experiential component to it, online WOM do play a significant role.

1. Online WOM do amplify the WOM offline. The KFG report, TalkTrack™ Press Release May 15, 2007, claim that internet is the top driver for offline conversation. Offline conversation derived 12% of their content from internet, but only 7% from TV and 5% from newspapers.

2. Given the above data, online social media is probably most effective at driving awareness and consideration (the earlier phases along the marketing funnel), even though the final purchase decision is influenced most strongly offline. But as social media is becoming a indispensible part of our lives, I expect their effect to increase.

Finally, as a scientist, I must point out some observations that I made about the above report that might confound the conclusions of these reports.

1. None of these studies normalize for time spent online. So we cannot be sure whether the overwhelming large amount of offline WOM influence is due to the fact that we spend most of our day offline. Although we are spending more and more time online, it is still a small amount of time. Note: online time spent is not the same as time spent on a computer. It may very well be that Online WOM is more efficient, If we look at the amount of WOM episodes per unit time. But none of the study addresses this issue. I don’t blame them, these data are not easy to obtain. A person may report that they spend 1 hour surfing the net, but how much of that time is truly immersed in pure surfing. These endogenous and exogenous effects are difficult to separate.

2. None of the above result account for the efficiency of information transmission online vs. offline. Have you try to convince someone to buy your product on twitter? Or try to sell a product or service using only blogs, video. As mention in several reports above, much of the WOM recommendation and purchase decision are made from direct experience with the product or service. Even though social media made communication more efficient, but at the same time it is limiting in some other ways. Given these limitations of the social media, maybe 15-25% of online WOM influence is very good already. Unless technology can bring social media/interaction to an experiential level, maybe we should not expect this number to change very much. But since experiential social technology does not exist yet, I can only speculate.

3. The social media landscape is changing very rapidly the way information spreads. None of these studies addresses the real time aspect of online vs. offline WOM. Although the most recent report that I cited above are published 2009. The data are probably analyzed before that, and collected way before that. Things could change a lot in a couple of years in the time scale of social media. We just need more studies and more time to observe.

Conclusion

Finally, I must reiterate, regardless of how small or large the effect of social media is, it is a part that we can measure, leverage and act upon. That alone is very valuable. If I tell you that 99% of all social influence exerted through subliminal sensory perception, but there is absolutely nothing that you can measure about it; and you cannot manipulate it, control it, or leverage it in anyway. Then it is useless. This effect will just happen in the background, and those who can leverage that final 1% will have an advantage over the rest who do not.

I hope this clarifies my statement, and also addresses some of your questions. So when you say that social media greatly amplifies WOM. That is true! It greatly amplifies over traditional media. And it does drive much brand awareness and product consideration. But the bottleneck is still offline face-to-face communication. And there are also those three speculative observations that I made about these research reports. Until we have more data and better research methodology, I can only take it with a grain of salt.

Man, this reply is long. I feel like it is a separate blog all by itself. But I think this is an excellent question, and many people can benefit from it. So thanks for asking.

[#8]

Thank you, Mike, for sharing this incredible resource of research on influence. There is an aspect of social media influence that is not directly measurable and identifiable: hidden influencers.  As a network weaver, I utilize what I call the "underground economy of social influence."  With DM and email to my social networks contacts, I make connections between people and ideas that I feel would provide value.  Would this be classified as "offline" influence? I don't think so, but that is probably where it is being measured.  I recently did an interview with Valdis Krebs on hidden influence here: http://bit.ly/aI8bNA and discussed Valdis and June Holley's work on  "network weaving" for my work blog @odomlewis here:  http://bit.ly/aGGJkP.

Would be very interested in your thoughts on this.  Thank you!  

Angela Dunn

@blogbrevity

[#9]

Hello Angela,

Thanks for leaving me a comment and the compliment.

I just looked at your video interview and description of these so called hidden influencers. Very interesting. They are what social network analysis identified as connectors in a social network -- they are not connected to a lot of people themselves, but those who are connected to him have great influence and reach.

We can definitely identify these people reliably and measure their online reach and influence with social network analysis (SNA), specifically using a metric call betweenness centrality and the clustering coefficient. So they are just another type of ONLINE influencer. I described a few types in an earlier article that I published "Are all Influencers  Created Equal?" The type with high betweenness centrality is similar to what you call the hidden influencers.

It is an very old idea from SNA, dates back to the 1970s, but marketers are just realizing its potential. I strongly recommend you take a look at this Wharton research article "The Buzz Starts Here: Finding the First Mouth for Word-of-Mouth Marketing." You might need to register to see the content, but registration is free. Physician 184 that was mentioned in the report is precisely the hidden influencer that you talked about, or what social network analyst called connectors.

If you come back in a few weeks I will show you some examples of how we identify these network connectors with real data from our social CRM platform. In fact, I already have the post written, just waiting on the publication schedule to release it. Due to our Lithium Network Conference (LiNC2010), it will be posted couple weeks after that.

Hope this address your question about the hidden influencers.

[#10]

Mike, thanks for your comprehensive reply to my question re online versus offline WOM. You're no doubt aware of the recent McKinsey paper A New Way Measure WOM Marketing:

https://www.mckinseyquarterly.com/A_new_way_to_measure_word-of-mouth_marketing_2567

As I read it they don't mention online versus offline, is that how you read it?

You've made a pretty compelling case for the effectiveness of offline versus online. Where I still ponder whether it has current value is in the whole leverage effect. We know that social media is to WOM as the steam engine was to the lever - it created a whole new hugely more efficient way to create new things. 

McKinsey say, in my words, that WOM power is a function of:

who X what X where X why X volume = impact

The volume is the lever for social media of course. In offline, for sure, a single referral by a single trusted person in a bar would be effective, and this is what all your references are telling us. In this case the "volume" is 1. If online is only 10% as effective but the volume is 1000 or 10,000 times more, then ...?

That's where I'm at with this at the moment.

Regards, Walter Adamson @g2m

http://xeesm.com/walter

[#11]

Hello Walter,

Thanks for coming back and keeping the discussion going.

I've read that McKinsey paper a few weeks ago. Personally, I don't think they explicitly state any new way to measure WOM. What they have is merely a bare bone framework for WOM, as you quoted "who X what X where X why X volume = impact." They never tell you exactly how to measure the who, what, where, why and volume. Pretty fluffy. The report is more for managing WOM campaigns.

I think what you mention is true about volume; it is essentially the "Social Proof" that I've mentioned in this blog. It will take a lot of social proof to override a trusted friend's comment in a bar. It depends on how much you trust your friends vs the online voices that you researched. That is why trust is the last mile of influence (The 6 Factors of Social Media Influence). I don't think that anyone knows the precise number. Would be interesting to do a study on this... i.e. find the tipping point between online vs offline influence.

There are still a lot that we must learn about the human decision process. But that is what make it fascinating!