No Game, No Gain: Realizing the ROI of Your Facebook Fans

[#1]

Michael Wu, Ph.D. is Lithium'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.

---

Hello and welcome back. I must apologize that I’ve been a little behind on my blogging. I was at SugarCon last week and didn’t have much time to blog in the last week or two while dealing with my engineering responsibilities. By the way, my SugarCon presentation on The Physics of Influence has been uploaded here. Feel free to view and share.

In my last post, I finished the theoretical analysis of gamification via Fogg’s Behavior Model (FBM), examining the three factors in FBM: Motivation, Ability, and Trigger, and showing how they manifest themselves in various game mechanics and game dynamics. So, what’s the big deal? The deal is that well designed games are able to solve complementary (and relatively much harder) problems than social networks do. If used properly, gamification is able to drive long term engagement and persistent actions reliably.

Connection vs. Interaction

Social networks connect people. In other words, social networks determine who can potentially interact with whom. However, social networks do not determine the level of interaction (how deep, how frequent, etc.) between any pair of connected individuals. Clearly, if two persons are not connected, there won’t be any interaction; but simply because two people are connected, it doesn’t automatically imply they will interact. A connection on the social network can only guarantee the potential to interact, but not the actual level of interaction.

Take a look at your own social network (e.g. Facebook, LinkedIn, etc.); do you interact with every one of your friends (connections) with an equally high frequency? My guess is probably not (I know I certainly don’t). The question is why not, even though you are connected? The reason is, although interaction implies connection, the reverse is not true. A connection between two people tells you nothing about their level of interaction. If two persons are connected, they may have very high level of interactions, or have absolutely no interaction at all.

It is easy to create a connection, since it only required a simple initial action to connect (or accept your friend’s invitation to connect). Therefore, according to the FBM, people will create connections. The social networking platform then automatically maintains these connections for us indefinitely, no further action is required. A high level of interaction is much harder to maintain, because it requires many subsequent actions to persist over time, long after the initial action of connecting.

This is where gamification comes in. A good game not only drives the initial action, it can reliably and predictably drive persistent actions (or interaction) in the future. That is something that connections on social networks cannot offer. Why is this important for enterprises? Primarily because the value comes from subsequent actions and interactions are much greater than the value initial connection.

Realizing the Value of Your Connections (Fans)

When you acquire fans on your fan page, they simply establish a connection with your brand. This connection has a lot of latent value waiting to be realized. Otherwise, investors and analysts wouldn’t peg Facebook’s valuation at $40 billion. So what precisely are these latent values, and how can they be realized?

Think about all the different reasons why you would maintain a connection with someone, say Bob, who you don’t communicate or interact with regularly. Maybe Bob can help you get a job in the future, or perhaps he could accommodate you when you are stuck in his home town due to a canceled flight. The potential value is limitless. However, none of these values can be realized until you start communicating and interacting with Bob. In other words, if there is no interaction, then the latent value of the connection cannot and will not be realized.

Value through Influence

The same can be said for enterprises. A fan can potentially bring many values to an enterprise. For example, he can influence other fans, customers, or consumers through word of mouth (WOM). If you browsed through my SugarCon slides, you would find that our research showed that WOM influence can bring value to an organization through two different mechanisms:

1. Acquisition of new customers: This is what Dr. V. Kumar calls customer referral value.
2. Acceleration of adoption (or, early adoption): The value of money is discounted over time, so early adoption can very profitable to an enterprise over the lifetime of the customer. Businesses often realize the value of early adoption by faster time to market, and WOM influence can accelerate this even further.

The caveat is that without interaction, there can be no influence. If you get a fan that never interacts with others, then there is no way he can influence others. Although interaction does not guarantee influence, it is required. This is manifested in my influence model as one of the six necessary factors - high bandwidth. Whether the interaction is direct (e.g. communication through various social channels) or indirect (e.g. through recommendations, or other user generated contents), it requires your fan to take some actions after connecting to your brand by clicking the “Like” button. A fan that never takes any action to interact is actually useless to you, because the latent value in the connection can never be realized.

Value through Loyalty

Another way that fans can be valuable to an enterprise is by being loyal to the brand. Aside from having greater customer lifetime value (CLV) through persistent consumption, loyal customers also offer greater opportunities for up-selling and cross-selling. But does being a fan on your Facebook fan page make him loyal? Maybe, maybe not.

To ensure loyalty, enterprises need to convert a connection (a fan on your Facebook fan page) into a real fan. Facebook fans are not necessarily true fans in real life, because it is too easy to become a Facebook fan. Since all it takes is a single click to be a Facebook fan, virtually anyone can be your Facebook fan. So fandom on Facebook loses its meaning and no longer has any reliable association to the fans’ actual level of loyalty.

To turn your Facebook fans into real fans (or superfans), enterprises need to build deeper and stronger relationships with them. This requires interaction between your fans and your brand. It would be very difficult to build any relationship with someone who never interacts with you.

Conclusion

So what did we learn today?

1. Social networks build connections that give people the potential to interact, but these connections are not indicative of subsequent actions and interactions.

2. The latent value of a connection is the potential to interact, and when people actually interact they can realize this value.

     a). Interacting with other consumers enables WOM influence

            - Increases the acquisition of new customers

            - Accelerates adoption of new products for existing customers

     b). Interacting with brands builds deeper and stronger customer relationship

            - Increase customer loyalty and CLV

            - Offer greater opportunities for up-selling and cross-selling

3. If you think the connections established by social networks (e.g. fans on your Facebook fan page) are valuable, then gamification is even more valuable due to its ability to drive subsequent actions/interactions.

Finally, I must emphasize again that a connection does have intrinsic value, since it is a prerequisite for subsequent interactions. Interactions simply create greater value. The social gaming industry (e.g. Zynga) would not have gained such a huge market if it weren’t for the existing connections on Facebook. However, just having the connections on Facebook is by themselves is insufficient to drive the level of interactions you see on social gaming platforms.

Connections are easy to establish, because it only requires an initial action to connect. Subsequent interactions are much harder to facilitate. Many fan pages are collecting thousands of dead fans who never take any action to interact, either with the brands or with other consumers. However, well designed games are able to drive interactions reliably on the social graph. So you should start embracing gamification. Without games, the latent value of your connections (e.g. Facebook fans) may never be realized to its full potential.

Alright, next time let’s look at another application of gamification in the e2.0 and productivity context. In the mean time, I welcome any discussion as usual. Stay tuned for more applications of gamification.

[#2]

What a great distinction. It is very easy to make a facebook page or a twitter account and make connections; though, what really sets an enterprise apart is the level of interaction they take with these opportunities. This next-level interaction that "gamification" brings is and has been utilized by leading enterprises and it is really great to get your thoughtful and well-writen background on why it is so powerful. Thank you for all your hard work.

[#3]

Hi Mike!

I was reading CLV article from Wiki. And you know, these 2 situations with "nature of the customer relationship". 

Pure mathematics depends on this real customers behaivor .. in future. Classically it s calculated as function from preexisting dynamics. But games are some turbulent for these "order" of things (everyday customer actions).

How strong can we predict given gamification outputs for CLV migration? Throught Retention rate?

[#4]

Hi Michael,

Interesting post as always. Agree with you totally on connection vs interaction.

The one thing I would like to add is that the traditional methods of modeling CLV and loyalty will not work well unless modified. Traditionally, CLV models have been transactional in nature - and the value derived and based off hard numbers. In the social world, value can be derived from soft or hard areas - as you said from being a fan (one level), interaction (another level), buying a product/service (yet another level) etc.

Some day I would definitely love to hear more of your thoughts on how gamification can "reliably and predictably drive persistent actions".

Enjoyed the read.

Regards,
Ned

[#5]

Hello Hunter,

You got it.

Facebook fans, twitter followers, Linkedin connections, or any connections on social network are only valuable because of their latent value. However, none of those latent values can be realized without interaction.

The key is that connections will saturate as people make more and more connections, because these connections persistent. Once you are connected, you are connected indefinitely until you disconnect with your connection. But interaction is not cumulative. If 2 persons stop interacting today, the level of interaction between them will start to drop. So maintaining interactions is much harder than maintaining connections.

So I just want to point out that interactions can bring much greater value than connections, and gamification can help us do that by the FBM (i.e. by creating a convergence of motivation, ability and trigger).

Thank you for the comment. Glad to see you back and hope to see you again next time.

[#6]

Hello Andrei,

That is true. Ideally, CLV should depend on the future customer behavior, but classically, CLV is computed based on pre-existing customer dynamics. The assumption that you have to make is that the future behavior is similar to the past. That is the assumption that is built-in to the CLV calculation, because there is no better way to predict future customer behavior.

Other factors, such as discount rate, retention cost, etc., can be model more accurately and extrapolate into the future. So we do not have to make such a strong assumption. There, we can let the model tell us what their future value should be.

As for gamification, it will probably alter the churn rate if the gamification process is design to increase brand loyalty. This could only give us a change in the mean and confidence level of the future churn rate. Beyond that, we still need to make the assumption that future churn rate is either similar to present, or follows a conservative trend determined by gamification.

To your question of how strong can gamification alters the CLV. Well, that depends on how good you designed your gamification process. You just have to implement it, collect some data, and watch how the churn rate change over time and get and estimate of the trend in churn rate. There is no magic about it. We still need to gather data, do the standard statistics, and iterate to get the answer.

I hope this address your question. Thanks for the comment and see you next time.

[#7]

Hello Ned,

There are actually many CLV models out there. Which one you want to use depends on how opportunistic you are. Personally, I think that the actual CLV should really depend on the hard numbers. Otherwise we are just fooling ourselves. You can assign some small amount of value to being a fan, but what if this fan never interact and never realize his value? We had seen the effect of people being too opportunistic with the housing/credit bubble already, there is no need to artificially inflate the social value in CLV that may or may not be realized.

Let me clarify that doesn’t mean that you should only use the most conservative model. We should include a customer referral value term in the CLV if we have customer referral data, how many of the referred actually became customers, and how much they purchased. The model can include many terms that are not traditionally included in the CLV calculation, but only if we have data for those terms.

With the example you gave about fans, if you don’t have data on the fraction of your fans who actually purchased, and how much difference they make in revenue compare to non-fan customers, then you shouldn’t include them in your model. If you have those data, then sure... use them by all means.

About gamification’s ability to drive persistent actions, that is basically what I’ve been talking about in the last 5 post on the science behind why gamification is able to drive action so effectively. Basically, it is the FBM (i.e the convergence of Motivation, Ability and Trigger).

The key to driving persistent actions long term is to understand the psychology of flow. No fixed gamification scheme can work forever. People change, and your gamification scheme must adapt. And good gamification includes this adaptive process. I will try to expand on this idea a bit, but I don’t want to just list a series of gaming dynamics that you can find elsewhere. Let's see if I can turn this idea into a post.

Alright, I hope this address your question. Thanks again for coming back and for the discussion. See you next time.

[#8]

Hi Michael,

Couple of clarifications. My intention was not to say that hard numbers should be replaced in the CLV modeling.  A good CLV model can be a good proxy for customer value evaluation - but for that there has to be measurable and quantifiable metrics incorporated into it. Neither did I mean we should artificially inflate the social value. 

However, I absolutely do believe (this is of course my own personal bias) that we can build better and more accurate CLV models by going beyond the transactional piece. There are many other great measures that can be incorporated - you mention referral value and that is definitely a good one; you can even put a score to social aspects & interactions (of course, the big assumption here is that you have a process set up to identify and link this information with the data from other channels). 

But forget social for a moment. I think the CLV can be strengthened by looking at customer interactions with even your traditional channels.  I don't have empirical proof, but I am pretty sure not many folks look at customer interaction with sales, customer support, in-store etc. as components of CLV. 

The same goes for loyalty. In a relationship, the true measure of loyalty is when things go bad. And that is why I think measuring the "bad" events and incorporating them in your analysis is as important as the good ones.  (If a customer keeps coming back even after an "issue", that to me gives a different signal than a customer who never had issues -- all other things being equal).

Anyway, these are some of my thoughts...:-).

On the gamification point, I should have worded it differently, You are absolutely right -- your past articles provided an excellent summary on the concepts and the why/how etc. I was just thinking out loud on how best one can incorporate the results from gamification into customer value modeling & predictive analytics. 

Regards,

Ned 

[#9]

Hello Ned,

Thank you for coming back and continuing the discussion.

I guess we have different ideas about CLV models. You seem to think of traditional CLV model, where as I think about all the new agent based model developed recently and contains many terms such as referral, loyalty, etc. And I think we both agree that CLV models should only include terms to the extend where data is available. Without real data that is tied to real transactions, we can artificially inflate the CLV of a customer.

For example, if there is a referral term in your CLV model. And you have referral data, say you know I referred 100 friends. However, you don't have the transaction piece of this referral. That is, you don't know how many of my 100 friends actually buy, or how much they actually spent. Then there is a danger to assign some value to my referral term. Just because we have referral data, doesn't mean we know their value. If we assume 1% of my referred friends buy, then that is an assumption. It better be backed up by hard transaction data in the past. The value component of each of the CLV terms should still be based on real transaction.

I hope this clarifies what I meant. CLV should include terms that takes into account of social interactions (ie referral etc). And many CLV models do. The problem is people may not have all the data to use those newer models. And by putting in numbers that are not based on transaction history, they risk inflating the CLV of the customers.

To see the effect of gamification on CLV, we really just have to implement it and see how it change the slope of each of the CLV terms in the model over time. I briefly described that at the end of my reply to Andrei (above).

OK, thanks for the discussion again. Definitely appreciate it.

[#10]

Michael,

Excellent, as always.  I completely agree on the distinction between connection v. interaction.  I liken it to potential v. kinetic energy.

Gamification is certainly on my mind and in my strategies for the firms I work with, but other than companies that engage foursquare, fb, and other shopkick type applications I don't see much in the way of companies instituting gaming practices to increase individual interactions.

I like what Mashable is doing with their new "follow" program but they don't need to monetize individual CLV.

Do you have some specific examples that pay off how gamification can "reliably and predictably drive persistent actions"?

Thanks for the compelling content--love it that you've tied in the "flow" concept from Csikszentmihalyi!

Killian

[#11]

Hi Michael,

You are right - I was referring to the traditional CLV models as my assumption was that a majority of the readers will be having that mental model (unless one is really into analytics & modeling in which case as you said they will be aware of the newer variations and inputs that can be modeled in).

Irrespective, I think we are almost on the same page. Like you, I also strongly support inputs that are quantifiable and measurable. The only slight deviation we have in our approach is the weight placed on latent constructs as denoted by info like referal data. I think depending on the goals and intent behind creating the model (driving awareness, retention, cross-sell/upsell etc.), I still think referral data & social data can be put to good use even without real transactions behind it  :-)

Thanks for a good discussion.

Ned 

[#12]

Hello Killian,

Thank you for the comment and the nice note.

Being a Physics undergrad and a Biophysics PhD I can certainly appreciate you making the connection to potential energy and kinetic energy. The only difference is that potential energy is a true form of energy, but the latent value in connections do not necessarily have real value. When we make analogy, we always have to be careful where the similarity ends.

The application of gamification is only limited by your creativity. Gamification.org maintains a pretty good list of gamification stories by industry. If you are looking for specific example, I suggest you look there. However, to give you an idea how gamification can be use creatively, take a look at this application for the Speed Camera Lottery.

Yeah, Mihaly Csikszentmihalyi is definitely one of my favorite psychologist in the 20th century. His work on flow has profound implications in both gamification and non-game settings. I actually had the honor of meeting him in person at a conference.

Thanks again for stopping by. See you again next time.

[#13]

Hello Ned,

Thank you for coming back again and glad to hear that we are on the same page.

You are absolutely right. If you want CLV to model the monetary value of customers, then you must tie it back to the transactions. But if you only want awareness, or something fluffier, then sure, you can make some assumptions on their value. Personally, I don’t believe people can quantify true awareness perfectly anyway (unless people are all wear some kind of neural sensory devices), so it doesn’t hurt you much to make wilder assumptions there.

But ROI is a clear and defined quantity and it means money, not awareness, referral, loyalty, or even fans or interactions for that matters. So we must be careful when we say CLV, and we should be clear what kind of value we are talking about.

Thank you again for the excellent discussion. See you again later.

[#14]

Hi Mike!

I see your great discussion with Ned!

Let me ask 2 questions about. 

1. How strong does newest CLV models be moderated with only transactions (purchases) logic?

Can we assign, for ex., monetary values for some types of interactions between customers - such as social support (high expertising or simply "helping" Q&A) or R&D ("good ideas" that works). It requires to see customer as not only buying being, but also collaborative with a company.

2. And I think it is some extend of one logic CLV, CRV or СNV to calculate ROI. 

[#15]

Hello Andrei,

Thanks for the great question.

However, the answer is that this is still a research question which we do not know the answers. We have the methodology, but we still need to get the data. Our platform does not track transactions. If you have transaction data from companies, please let me know. I'm happy to analyze it.

Modern CLV includes many terms, referral, network, co-innovation, even lurking. But how much do each one of these contribute? We'll have to wait for the data to tell us. Without data, we are just guessing, and I don't like to speculate, so I am not going to. Maybe the data will tell us some of these terms don't even matter. That is not to say they have no value. They do! Just that they may be only 0.0001% of the total ROI, so we don't even need to care about it. But we just don't knonw the answer yet.

BTW, I'm curious do you use Google translate or other machine translation tools to help you post your questions? I can't help noticing some of the machine like colloquialism in all your comments.

Anyway, thank you for coming back with questions. And thank you for your interest and determination that overcome the language barrier. See you next time.

[#16]

Hi Mike!

I am not using Google Translate. It is simple my current language skills level. 

I understand your not-speculative and empirical spirit.

Really I am going to make analysis of social support efficiency of the "freewill" support activity by one auto brand customers (on small communities of brand|product specific forums). And I will to demonstrate financial impact of this activity by attributing nominal $ to some typs of helpers acivity (solving some valueable problems in requests of another customers - if it is impactfull for the company). 

And I think that it is a long "causal road/chain" from this support for the customer (given it is impactfull for "loyalty") to transaction by the customer (purchase new auto or service). The problem is not to track (really in my case it is impossible ) transactions and understand ROI for the given customer, but to attribute the value on the base of valid causal linking support and transaction. 

I am interested in the vision of separated ROI for transactions (purshases) and support/service as both valuable actions for company bottom line. In another case my analysis is not 1$ worth.

It is simply my problem situation and I will only understand whether you can help me to solve it.  

[#17]

Hello Andrei,

I see. What is your native language? Just curious.

Support ROI is different from CLV. We have a pretty good model of support ROI through our call deflection model. And we have shown that even under very conservative assumptions our platform is able to generate demonstrable ROI. We have even built this into our platform as a ROI calculator. In the support ROI, it is a cost saving. So most companies know the precise amount they spend on maintaining other support channel. So we can easily demonstrate the cost saving ROI from social support.

I would just take analyses without hard data with a grain of salt. Anyone can come up with plausible causal links. And these links may be real and valid. But to make a claim that they are worth $xxx is a leap of faith that I am not willing to take without some hard data. That is not to say you shouldn’t do it. Different people have different tolerance for risk and error. I just like my analyses clean and with as few assumptions.  :-)

So if you are interested in support ROI, maybe I can help you. But if it is something fluffier, than you’ll have to wait till I get some more data.

Thank you for coming back and see you next time.

[#18]

Thx, Mike!

My native language is Russian. 

Yes, I am interesred in support ROI.

And I will not make something without hard data. 

How can you help me, Mike? 

[#19]

Hello Andrei,

I see. I'd be happy to share with you our Support ROI calculation. Keep in mind that this is a call deflection model and rather specific to our platform, which keeps track of the participation data in our community. I will Twitter DM you my skype ID and/or other communication channels, then we can speak over skype or something with higher bandwidth. Sound good?

I'll have two travel/speaking engagement starting tomorrow for the next 2 week. So let's plan something after that.

[#20]

Hello Mike!

It sounds good!

And deflection model is a start point to think about my value attribution problems.

My English is my shame, but not flame )

Let see whether we can really speak or chatting. 

I've added you in Skype!

See you later!