My Chapter on Influencers

[#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.

---

Recently, there has been a lot of buzz around the term "influencer." This is partly due to the launch of Fast Company's Influence Project. Beside this project, there is another unrelated project with a very similar name: The Influencer Project, which was also launched around the same time. And yesterday, I gave a webinar at WOMMA on the topic of Influencers and WOM marketing. My presentation is on SlideShare (which fail to convert the animations in the deck); a fully animated version of the slide show is linked at the end of this article as attachment.

Since I’ve done quite a bit of research and written many blog posts on topics related to influencers, I thought it would be nice to collect these posts together in a single spot. This would facilitate the sharing and distribution of these articles, which I am asked about frequently these days. I’ve created a word cloud (via Wordle) for these posts, so if the topics in the word cloud look interesting to you, then you will enjoy this collection. Aside from listing the posts, I will add some commentary to these articles along the way.

The first set of four articles introduces a simple model of the influence process (or simply influence model). This model emphasizes the importance of the target when identifying influencers. It also gives us six necessary factors (i.e. credibility, bandwidth, relevance, timing, alignment, and confidence), which must all be accounted for in order to achieve true influence.

1. The 6 Factors of Social Media Influence (introducing the model)
2. Finding the Influencers (covers credibility and bandwidth)
3. The Right Content at the Right Time (covers relevance and timing)
4. Hitting Your Targets (covers alignment and confidence. There are several excellent and in-depth dialogs in the comment section)

The next set of three posts is on the application of the influence model. Following the principles of this model, I’ve devised a step by step procedure for identifying influencers in interest-oriented online communities. The implementation of the influencer identification algorithm involves heavy use of social network analysis (SNA) and various SNA metrics. Using SNA and social graph visualization, I was able to discover some interesting insights within the communities from our client base.

1. Community Influencers Step by Step
2. Social Graphs: The Art and the Insights
3. Social Network Insights from Unconventional Graph Metrics

If you are unfamiliar with SNA, the following two posts offer a very basic introduction and an application of several SNA metrics for characterizing different types of influencers.

1. Social Network Analysis 101
2. Are all Influencers Created Equal?

Finally, the last two articles are on Fast Company’s Influence Project. Since this project has stirred up much controversy in the industry, there are a lot of very provocative discussions and debates in the comment section. The comments in these two posts are probably much longer and more interesting than the post itself.

1. The Fast Influencer Myth
2. Social Fame, Social Shame, and the Accounting of a Game: Fixing Fast Company’s Influence Project

Alright, that is what I’ve written on influencers so far. This topic is so deep and important that people start companies (e.g. Klout) just to find the influencers. There is certainly a lot more research that needs to be done, and it is very likely that there will be another chapter on this topic later. But for now, enjoy this self-contain collection of article on influencers. Again, comments, questions, discussion, as well as criticisms are all welcomed as usual. You may comment here or on the comment section of the respective posts. See you next week.

[#2]

Good idea to consolidate this topic...

I was thinking about Social Shame as an anti-scamming tool the other day, and also how cheating is problematic on 'checkin' applications like FourSquare. I know companies are working on technology-based solutions to prevent cheating/false check-ins to earn badges, become mayor, etc. - e.g. more precise GPS controls.

I wonder if there could also be an effective crowd-based solution. For example, when I check-in at a venue, others at the venue (maybe employees, as well as customers) get notified and can validate or refute my check-in. It could get messy, but it struck me as having potential.

Any thoughts on this?

Thanks,
Mike

[#3]

Your slideshare link does not work. It says the document has been removed.

[#4]

Hello DavidG,

Thank you for notifying me of this problem. It appears that my slides have too many animations in it. Althought it was uploaded properly, SlideShare failed to convert and transcode it into their format. I tried many different methods of uploading, but SlideShare really just can't handle it. Finally, I resort to converting my presentation into a PDF file, which will, however, remove all the animations. This is the only way that seem to worked for me right now. So I apologize if you don't get to see the nice animations. The link is now Fixed, but the slide won't play as nicely.

Thanks again for point out this problem, and I hope to see you around Lithospehre.

[#5]

Hello MikeD,

Thank you for the comment again.

It is true that check-in apps like FourSquare are susceptible to gaming and various abuses. But I don't think the problem is very serious now, because the incentive for cheating is not huge right now, and they are purely psychological incentive of being mayor or getting badges.

However, I can see that if venues start to use check-in data to reward people, then that is a whole different gaming psychology there. For example, if a restaurant or a bar is offering a free drink when you get 10 point on their restaurant/bar, then there is a strong economic incentive for people to cheat.

As for GPS location control, I think I can be done completely from an algorithmic level. That means it should be possible to do it without the validation of a second party, such the venue workers or other visitors of the venue. The solution that I am thinking involves 2 subtle changes to the current system:

A1. Using the GPS location data from other user who claimed to have check-in at the same venue

A2. Require a check-in and a checkout, and the points you get is proportional to the amount of time you spent at the venue.

These are similar to the normalized accounting that I talked about in the previous post: Social Fame, Social Shame, and the Accounting of a Game: Fixing Fast Company’s Influence Project

A1. Assuming that most people are honest (we can do some preliminary analysis to verify the validity of this assumption) and that there is only a small fraction of people who cheat. Then we can use data from all the people who claim that they've checked into this venue to get a very good idea of the proper distribution of long-lat coordinate of any particular venue. Then, when you check in, they just have to record your current coordinate and see if it is an outlier of the proper distribution of the coordinate for the venue that you checked in. This will ensure that you are in the correct location.

A2. Right now, the player gets points with only a single check-in, and the point value may differ depends if that was your first check-in, and number of times you've checked in already, etc. But the system can incentivize people to check-out by multiplying the points by the minutes that people spent at a venue. Basically, if you only check-in and don't check-out, the system only assume you spend 1 minute there and you get 1 point. But if you check-out 30 minutes later, then the system will know that you spend 30 minutes there and reward you with 30 points. If that was your first check-in where you should get 3 points, then you would have gotten 90 points if you checked-out.

There should also be an expiration time or max time spend at a particular venue that can be set for particular types of venue. For restaurant, maybe the max you get is 120 minute of multiplicative factor, but not beyond that. And if you don't check out within 3 hour of your check-in, then it assumes that you never check-out, and you only get 1 minute (point). This prevents you from coming back tomorrow to check out and get 24 hours of time spent at the venue. The precise numerical value for these thresholds can change after we analyze the data to establish the normative values. Of course the check-out coordinate must be verify via the same procedure as A1 to make sure that people are indeed check-out at the same venue they check-in.

The time-spent compnent make it very hard for people to game the system, because they have to actually spend time there to get the point. So, if they just stop by, check-in, then drive to the next location and repeat, they can only accumulate very little points for that venue.

These are couple of things that I can think of that could significantly reduce cheating when the incentive to cheat becomes strong. There are probably a lots of ways to implement this. For example, they can create a cheater badge and accumulate point secretly whenever people check in a location coordinate that do not fall within the proper distribution of the particular venue they check-in. This would be leveraging the social shame idea. But there might be some adverse effects to using negativity. This would need to be thought through carefully.

Anyway, hope this can get you started with this problem. Thank you for asking the question.

[#6]

I definitely agree with Michael W. that this topic has to be  thought through carefully before a solution gets executed. You definitely want to reduced false positives here -- meaning, you want to be sure someone is gaming the system before you put that label on them.

To that point, I am in favor of a triangulation approach. Cross-check and time-spent as Michael suggested are definitely two ways of blocking gaming. One might also develop heuristics based on historical data. Yet another approach could be to have a "co-signer" - so if you are checking at a restaurant, the restaurant also have to 'do something' before points are awarded.  Or one could pre-establish a check-in at a location and once they arrive there, confirm it through some GPS location control (just shooting off my thoughts here without much pondering).

I also agree that the "social shame" concept should be one of the last in execution. Maybe there can be a tiered approach. If someone is found to be gaming, you send a warning for them to either clarify or defend. If it still goes on, then you go to the next stage ...etc...and the last one is the social-shame badge.

Anyway, interesting topic.

Regards,
Ned

[#7]

Hello Ned,

I see that you've finally registered and become a Lithospherian. That is great. Now you can take full advantage of the features of the Lithium's platform, such as kudo, reputation engine, private messaging, and much more.

There are a few points to your comment that I like to dig a little deeper. When you said triangulation, do you literally mean triangulate to figure out the location of the person checking in? Because the GPS signal is already a result of such triangulation, and it gives you the long-lat coordinate (or location) of the person checking in, I don't see the point of doing that again. And I am not quite sure what you mean by cross-checking.

The method of a "co-signer" at the venue is the method suggested by MikeD originally. I did not suggest that route because that involves adding an extra step that has to be validated by another human. This would make the whole process and user experience less autonomous, less entertaining, and less scalable. But it is definitely a sure way to prevent gaming.

The use of historical data can actually be make much more rigorous than just heuristics. And that is the method that I mentioned under A1: Using data from others users who checked in to the venue to construct the proper distribution for a particular venue. I believe this is the primary method that people at Four Square are working on.

Finally, I agree 100% that social shame should be a last step. You definitely don't want to mis-label a user as cheater unless you are really sure. I guess "social shame" has a strong negative connotation just from the terms, a better term in this case would be "social accountability."

Anyway, Thanks for the comment.

PS: Re your private message: On the request to view the animated version of my WOMMA presentation on Influencer and WOM, I will figure out a way to make my presentation available. I totally didn't realize SlideShare's transcoding capability is so limited. Besides, several people have already asked me for it, so I believe there is a demand out there. However, I will need to discuss this with our in house counsel first, because the presentation contains some client data, even though they are fully anonymized. But I'm certain that we can figure out a way to make the presentation available.

[#8]

Hello Everyone,

To those of you who want to get an animated version of my WOMMA presentation on Influencers and WOM, I got good news! I'm able to provide the PowerPoint file as a slide show to you. You may pass on the file as long as the content stay intact as a complete unit. In fact it is protected so it would stay that way.

I have to get our operations team to increase the attachment size limit on our blogs to 10MB due to the size of my PowerPoint file. Not to be advertising for our product, one of the cools thing about being a SaaS solution provider is that we can deploy this change pretty much on the fly transparently. Once that change is deployed, I just upload my presentation as an attachment to this blog. That is it! This way of delievering the file is much more scalable than sending it via email. Here is how you get it.

Scroll back up to the body of this blog article. At the end of my blog post you will find a link to my ppsx file like this.

   Attachments:

   2010-07-21 WOMMA Webinar Influencer & WOM v03ss.ppsx 6356 KB

Just click on that link to download my presentation.

Let me know if you run into troubles. I'd be glad to hear your comments on the presentation too.

[#9]

Hello Michael,

Lithospherian!! - sounds like I am from another planet :-).

First of all, thank you for putting the presentation out - much appreciated.

On triangulation, no I did not mean "traingulation to figure out the location" but more along the lines of your A1. If a person says they checked into a location at a certain time and for a certain period, then validate that with some independent information from other sources (cross-check).

On the co-signer stuff, you are right that MikeD did suggest using crowd-based solution. I was thinking more along the lines of "loyalty cards" or "QR codes"  type of action. A person checks into a restaurant the first time, they get some kind of venue-specific card which is scanned to confirm their check-in and from that point on they check-in and  get their cards scanned to get points. Or have some kind of location specfic QR codes (containing the address) which they have to read and upload using their mobile device.

Best,
Ned

[#10]

Hello Ned,

Yeah, didn't you know that we are all alien species?

You are welcome. Glad we can work out a way to provide the presentation without compromisng our anonymized client data.

Interesting point on QR code. Since you mentioned it, I will expand on an idea that I had a bit further. This idea would only work for venues that provide some kinds of receipts upon monetary payment, such as restaurants, bars, hotels, etc. It wouldn't work for places like public parks, shopping malls, work places, etc.

The idea that I was thinking is that venues can print out a date-time specific QR code on the receipts (i.e. the date and time are encoded in the QR code as well as the venue's information). Then people can simple scan the QR code on the receipt to check-in (or check-out, if we implement the duration of stay for scoring) to the venue. This would essentially implement a co-signer without adding the extra step of co-signing by an actual person. With the QR code provided on the receipt, which the venue need to provide anyway, Four Square user can again have an autonomous user experience with this game even if there is a cosigner.

Even though this wouldn't work for places that don't issue receipts, it is probably OK, because there is less economic incentive to cheat in these places. But of course, all the venues (that issue receipts) would need to be able to provide the data-time specific QR code for this to happen. To get all the venues on board with this process is not a trivial process.

Thanks again for the discussion. See you next time.

[#11]

@MikeW - printing out QR codes for each checkin - yeah, that'd be really "green".  ;-) 

You can spend as much time as you want trying to stop people from gaming the system, but in the end, they'll always beat you, since the "opportunity cost" is so small, no one will invest much in any technology or time to stop cheaters...

[#12]

Hello Muerte,

Thank you for the comment.

It's great that you are thinking "green." But venues are going to print receipts for people to sign anyway. It's no like that people are printing the QR code just for the purpose of checking in. Like I say earlier, maybe venues that do not issue receipts (e.g. Central Park) will not require a QR code to check-in because the incentive for them to cheat is lower, because Central Park probably won't come back an offer you free stay when you become mayor. So the player probably won't get anything besides the entertainment value of being the mayor of Central Park.

I agree that there is absolutely no fool proof way to completely stop the cheaters. Likewise there is absolutely no way to completely stop crimes, deceptions, or evil doings. But is that a reason for not putting anything in place to prevent them? We just have to put in some measure to make it harder for people to cheat. The people who really want to break the rules will still find a way, but it would stop most of the trouble makers.

I believe that once the problem is serious enough that it affect the user experience of other players, then people will spend the time and money to develop system to fight cheaters. If not, we wouldn't have any spam filters in our email systems now.

Anyway, thanks for the comment and hope to see you again.

[#13]

The powerpoint link goes to a more modern version of PowerPoint than most of us actually have.  Your presentation could be more functionally available were is uploaded in the older .PPT format.

[#14]

Hello RichReader,

Yeah, I use Office 2007. I will try to convert it to the older .ppt format to see if all the animation still works. If all goes well, I will upload another version of it as an attachment to this blog. Stay tune!

Thanks again for the suggestion.

[#15]

Hello everyone,

You ask for it. You get it. I've just uploaded .pps version of my deck on Influencers and WOM & ROI. Just scroll up to the body of this blog, and at the end of the blog you will find links to the slide shows are as follow:

Attachment:

2010-07-21 WOMMA Webinar Influencer & WOM v03ss.ppsx 6356 KB <-- PowerPoint 2007 version.

2010-07-21 WOMMA Webinar Influencer & WOM v03ss.pps 13613 KB <-- PowerPoint 2003 version.

I'd be interested to hear your thought on the presentation. So drop me a line if you can. And I hope to see you on Lithosphere some days.

[#16]

Hi All,

I was on vacation and just now getting back to this discussion. The first thing I would like to add, is that vendors already are giving freebies and discounts to four square mayors. I've seen this at a few locations in my area - independant coffee shop, italian ice shop, etc. Granted, these are somewhat small economic incentives, but we all love that free coffee or ice cream right?

I really like the QR idea, since it removes the human step. More importantly, I think it better aligns with vendor intent - incentivise repeat purchasing, not just checking-in and buying nothing. And who says the QR idea can't be green, down the road?

Thanks!

Mike

[#17]

Hello Mike,

Welcome back from vacation.

Well, that is good to know. So the economic incentive is already there. Please excuse my slowness in adopting all the social technologies.

I'm glad you like the QR code idea. But like I said, the algorithm there is pretty simple, but getting it rolled out and getting all the business to use it is the bigger challenge there.

Speaking of Geen QR... Since the QR reader is completely visual, it does not have to be on paper, venues can simply display the QR code on a monitor, or project it  on a white wall; as long as our eyes can see it, the QR reader can read it. There are plenty of ways to make QR greener.

I'm glad to have your validation and thanks for commenting.

Speaking of vacation. I feel like I need one soon too.

[#18]

You totally deserve a vacation, man! Hope you got to enjoy our 3-day weekend and stay away from the keyboard for a while. But I know how you 'mad' scientists are - it's a labor of love

[#19]

Thanks Mike,

You are definitely right! when I get passionate about something, it's almost like driving without a break. Could be quite dangerous sometimes. 

I think I should take a vacation soon just to refuel. It's probably better in the long run. So I probably will. We are already talking and making some plans for it. So I may disappear for a while in late September, early October for couple of weeks. In the mean time, it's something to look forward to.

[#20]

Great stuff here as usual. Says the .pps file needs a password?

Can you provide?

[#21]

Hello Dean,

You should be able to view and play the slide as a read-only slide show by not entering a password. I've mention this before (see here), that I can only provide the slide as a complete unit in read-only form due to the fact that it contain anonymously aggregated derived data from our client.

[#22]

Your articles about influence are very interesting. I am a PhD student from Finland - Aalto University School of Economics (former Helsinki School of Economics), from the department of information systems science. I am studying how companies could use social media and first approached this them from the perspective of consumer values. Now I would like to enlarge my perspective to influencers. Do you happen to know any academic articles/theories/frameworks related to the influencing or influencers?

Best regards,

Miia Äkkinen

[#23]

Hello Miia,

Glad to have a PhD stopping by.

There are too many to list here. There are also a lot of paper on information diffusion and word-of-mouth, which can be interpreted as a form of influence. A small list of literature that I can find in my computer at the moment is listed here:

1. Oestreicher-Singer, G., & Sundararajan, A. (2009). Recommendation Networks and the Long Tail of Electronic Commerce.
2. Cialdini, R. B. (2000). Influence: Science and Practice (4th ed.). Allyn & Bacon.
3. Dhar, V., & Chang, E. (2007). Does Chatter Matter? the Impact of User-Generated Content on Music Sales.
4. Dholakia, U. M., Bagozzi, R. P., & Pearo, L. K. (2004). A Social Influence Model of Consumer Participation in Network- and Small-Group-Based Virtual Communities.
5. Carl, W. J. (2006). What's All The Buzz about?: Everyday Communication and the Relational Basis of Word-of-Mouth and Buzz Marketing Practices. Management Communication Quarterly, 19(4), 601-634.
6. Nitin Agarwal and Huan Liu and Lei Tang and Philip S. Yu. Identifying the Influential Bloggers in a Community. WSDM, 2008.
7. Kiss, C., & Bichler, M. (2008). Identification of influencers - Measuring influence in customer networks. Decis. Support Syst., 46(1), 233-253.
8. Rice, R., Grant, A., Schmitz, J., & Torobin, J. (1990). Individual and network influences on the adoption and perceived outcomes of electronic messaging. Social Networks, 12(1), 27-55.
9. Romero, D. M., Galuba, W., Asur, S., & Huberman, B. A. (2010). Influence and Passivity in Social Media.
10. Stephen, A. T., & Berger, J. A. (2009). Creating Contagious: How Social Networks and Item Characteristics Combine to Spur Ongoing Consumption and Reinforce Social Epidemics.
11. Watts, D. J., & Dodds, P. S. (2007). Influentials, Networks, and Public Opinion Formation. Journal of Consumer Research, 34(4), 441-458.
12. Hogan, J. E., Lemon, K. N., & Libai, B. (2004). Quantifying the Ripple: Word-of-Mouth and Advertising Effectiveness. Journal of Advertising Research, 44(03), 271-280.

Despite the numerous referrence out there. The framework I've talked about in The 6 Factors of Social Media Influence are developed by me, and I haven't publish it in any peer review journal yet.

Thank you for the inquiry. Come back again.

[#24]

Wow, thanks for the list! I'll be back after scanning those articles. A very valuable list for me. -- Miia

[#25]

Hello Miia,

I hope that by scanning you meant reading them.

If you mean to digitize them, then I want to let you know that I have these articles in PDF. If you send me a private message with your email, I'd be happy to email you the PDF files? However, you have to register and be a member of Lithosphere first to use that feature. Alternatively, you can follow me (mich8elwu) on Twitter, then DM me your email address. Either way, There is no point wasting time to re-digitize them.

See you next time.