Thought Leader Interview
Michael Wu is the Principal Scientist of Analytics at Lithium. He received his Ph.D. from UC Berkeley’s Biophysics graduate program, where he modeled visual processing within the human brain using math, physics, and machine learning. Michael is currently applying similar data-driven methodologies to investigate and understand the complex dynamics of the social web. He has developed the Facebook Engagement Index (FEI), Community Health Index (CHI), and many predictive social analytics with actionable insights. His R&D work at Lithium has won him the recognition as an Influential Leader by CRM Magazine.
Denis Pombriant: Can you describe some of the opportunities that you see coming out of the intersection of social, mobile, and analytics as they converge?
Michael Wu: I think that's a big area. An important point that I don't think many people understand about mobile is that mobile is not really a computing device. It's more of a user experience, or information retrieval device. One of the consequence with that is most of the analytics, the heavy number crunching, have to be done on the cloud somewhere, and the result gets pushed back to the mobile device.
I think that's kind of interesting because once you collect a lot of data from a lot of users, over a long history on the cloud, you could do very interesting stuff. For example, you can do what last.fm (which powers spottify’s music recommendation engine) did with their music listening behavior. You could compare musical taste, predict songs that a user might likes, and do a lot of very interesting analytics there.
Back to the convergence of social, mobile, and analytics. What I think is interesting there is what I called “the inference of context.” Right now, a lot of analytics are, in some way, kind of out of context. If you walk by a restaurant, your mobile device may send you some ads about deals in nearby restaurants. But you may have already eaten just an hour ago, and whatever device you use and analytics it is using, have no memory of your history, so it doesn’t know that you just ate. But, once you do all these analytics on the cloud, you have essentially unlimited storage, and you can store all your history, as long as you opt into it, then, the system will have a history of when you typically eat, what you like to eat around the specific location you are in, where you've been to.
Then they can essentially target you with a much more relevant ad at the time when you need it. So the system knows that during that time, it's just about two hours past his dinner time, and he has just spend 45 min in a restaurant an hour ago, so this customer is probably not looking for dinner. He's probably just out for a stroll and happened to walk by a restaurant. The accurate inference of this context is very important. I think that's a great opportunity. I don't think many people have explored that yet.
DP: I think you're probably right. I find it interesting that you talk about a mobile device as not really being a computer, and I agree with you. Mobile devices are consumption devices. They're not input devices. Nonetheless, there are large parts of the world where that is the only computer that people have.
MW: Yeah. That is really an illusion though. Even though mobile devices are the only computer people have in many parts of the world, the actual computation may not be done on the device itself. They simply get a result back transparently, as if the device is computing and generating the result.
For the longest time, most of the heavy computation is done on the computer, but I think now with the cloud, all the distributed computing infrastructure, and high bandwidth communication, it's much easier to push those heavy computing, number-crunching, tasks onto the cloud; onto the big servers, and just get the results back. I think the reason that we still have quite a bit of those on our personal computers — maybe a laptop; maybe a desktop — is probably because of legacy. Because inter-device communication used to be very inefficient, so we have to do all the computation locally. Software is written to do computations on your device. But communication is much more efficient now, so all that will change.
DP: Okay. Now, I know gamification is both a hot topic everywhere these days it seems like, and it's part of something that's very interesting to you. Can you describe it so that my readers can understand it?
MW: To understand gamification is really simple. People love to play games — most people do — but there are things that they don't like to do. Like, for example, go to school, or go to work, or using a certain really complex Enterprise software. The whole point of gamification is, can we put some elements, some attributes, of games that people like and infuse them into activities that people don't like to do, and make those activities more enjoyable.
If you want a definition, “gamification is really just the use of game attributes to drive game-like player behavior in a non-game context.” So, the first part of the definition is that you have to use some game attribute. Game attribute could be anything from game mechanics and game dynamics that people often talk about, or they could be game design principals, gaming psychology, narratives — anything that people use to make a games fun and entertaining.
Now, the second part of the definition is that it has to drive game-like behaviors. These behaviors can be engagement, interaction, competition, collaboration, and maybe even addiction, if you want to go that far.
Finally the last part of the definition is that you need to apply that in a non-game context. So, that means at work, in education, health and fitness, sales, marketing, and all that. Basically, anything but a game. That's what gamification is.
DP: Okay. Now, an associated idea, I think, is influence, and I'm not sure how they're associated, but perhaps you can tell me.
MW: I have done a lot of research on influence before, and I define it as an ability to cause a change of mind or behavior. And it also matters how you cause the change. People could put a gun at your head; that would certainly change your behavior or your mind really quickly. But that's not influence. That's coercion.
So, I have some criteria. To produce a change in mind and behavior is the minimum requirement. But, it matters how you do it. And the criteria is that you can't use money, so, bribes, and no force (coercion). Also, no frustration and no deception. So, I call this "No carrot, no stick. No annoyance, and no tricks." "No annoyance" means you can't frustrate them. The person must change their mind or behavior willingly. And "No tricks" means you can't deceive them into doing that. So, that's what influence is.
Now, gamification is all about driving actions, getting people to do something that they normally would do. In short, gamification changes people’s behavior, so, it is related to influence in a way that gamification is actually one way to influence people. It's one mechanism of influence.
Influence is actually a broader concept. Influence involves changing people's minds or behaviors, under the four criteria of no carrot, no stick, no annoyance, no trick, and games clearly don’t use these four things..
When you play a game that is genuinely fun and entertaining, you usually don't have to be bribe to play it. You would play it without anyone forcing you, frustrating you, or deceiving you. So, all these four conditions are met in gamification. But then, gamification also drives action; so it drives the change in behavior.
So, in that respect, gamification is a form of influence. It can be viewed as a one vehicle for influencing people. And it's the kind of influence that changes behavior rather than mind.
DP: That's very, very interesting. Now, all of this has some meaning for social CRM, Enterprise 2.0 and things like that. Could you knit all this together for me?
MW: I think a lot of people have kind of dabbled in Enterprise 2.0 and social CRM face some difficulties. I think the same thing happen in the CRM space — CRM 1.0, is that what they call it?. It has essentially failed because management focused too much on the technology and not on the people.
CRM systems helped you automate many things, but,nothing's really, truly, 100% automatic. People have to still do it. Even if it's as easy as pushing a button or a simple mouse click, a person still has to do that. Until you get to a point where the computers can make their own decisions, and make the choice of doing something by themselves, then humans still have to make the decisions and tell the system what to do — So that requires some action, some behavior, on the user side.
So when the technologies fails to deliver what it promised, it could be one of two reasons. One is that the software simply does not work. The other one is, okay, maybe the software actually works, but maybe it is a human problem. That is people are not using the software. The failure is not on the software side; it's human.
DP: I would say there is maybe a third option, and I think maybe you just said it. It's a user adoption issue. It's, I think, rather typical in the early phases of a new technology that we focus on the technology and not necessarily how best to use it. And best use only comes about after we stub our toes a lot.
MW: Precisely. The user adoption problem is human problem. But beyond adoption, there is the best practice. How do you use it effectively? This is perhaps the third option you are thinking. That is, we don't know how best to use it. But, this third option comes after the fact that people have to use it first. If nobody actually uses it, you would never find out how to use it most effectively.
So, that is where I think gamification could really help, because gamification could actually drive all these behaviors. People may not want to use the software, but if you make it fun, then maybe people will start wanting to use it, and if they start deriving value out of it, then they will continue to use it. Gamification can also drive experimentation and exploration, so enterprise can develop their best practice. Enterprise software often fails at the adoption step due to the steep learning curve.
DP: Exactly. I would imagine that analytics plays a very important role here.
MW: That's right, let me just give you one example. If you play a game, like World of Warcraft, it keeps track of everything you do. All the tools or the weapons you collected; all the pieces of gold that you found; all the monsters that you kill, etc. And if you kill a hundred monsters, you level up to the next level; you gain experience and all the goodies. The gaming platform keeps track of every single user action.
That's typically not done in an Enterprise software. For example, in most Enterprise software all they track is that, you write a document, and you just save the document. That's all. The software never know all the details, all the edits, and every single action that you did to get to the final result of the document. It turns out these usage information are actually is very important, because that tells you something about the user and how they use the software.
If the user had created a case all the way from the beginning to end, gone through all the work flow swiftly, and used every single feature and function perfectly, we know that he's an expert. He should be acknowledged publicly. And the software should refer people to him if people have trouble using the features and function that he used so well.
On the other hand, if another user is trying to do the same thing, and he still achieves his goal. But he stumbles around the workflow and tries one function after another. He has to click through twenty other menu items before he finds the right function. And then, eventually he got the job done. Then we know that this user is not an expert.
And maybe next time he encounters this problem, there's a notification that says, “You did this last time when you created this case.” to kind of give him some positive feedback to encourage him to learn more and retain the knowledge better. And if he really can't figure out, the system can offer suggestions and corrective actions that guide him to the right workflow. The system may even suggest an expert whom he can consult: “Maybe you want to try this function, and this person could actually show you how to use it. He's an expert and he had done this twenty times.”
This is an examples of social facilitation coming out of tracking how the user use a software. There can be social competition too, from tracking all the user behavior while they use the software. That could give us a lot of insights into the human side of the puzzle that we often ignore, and I think that can really drive adoption of enterprise software.
DP: So, is that where we're heading with all of this social technology?
MW: I think that's only one area that people will put a lot more energy into. I know Microsoft had created, literally, a game called “The Ribbon Hero 2” — it's really just a training tool for Office Suite — but they made it fun to play and while people play the game they learn new features and functions of the Office Suite. It essentially challenges the user to explore and learn to use new functions around their own usage patterns for the suite.
The worst way to teach people is to teach them something that they never have to used. But when you are playing this game, the game is listening and watching your every actions and predicting what you're going to do. And it can suggest features and functions for you based on what you’ve done and what you might want to do. Moreover, it’s fun and engaging, so people want to play the game. So this becomes a really good kind of a training tool for people, because it is relevant, so it’s useful, and it’s fun, so people want to learn more.
That was just one example of enterprise software that use gamification really well. And I think that other enterprise software vendors will catch up in that.
DP: That's fascinating. So what are the disconnects, the things that you think people in the marketplace don't see or don't get in relation to social, mobile and analytics?
MW: As I’ve mentioned earlier, people tend to think of a mobile device as a miniaturized computer, and many apps are written as miniaturized version of desktop or laptop softwares. I think there should be a mind shift there, to think of mobile as a consumption device, and really, a user experience device; more of an entertaining device rather than a computing device.
With respect to social, I think what a lot of people understand the importance of personalized experience. Everyone knows that customers want to be treated as a person rather than as a customer I.D. But, what they don't realize is that personalized experience, such as a conversation, like what we are doing right now, talking to each other personally, one-to-one, these conversations are really hard to scale. You can do that maybe, when you have a few thousand customers, but what happens when you have ten thousand, a hundred thousand or a million customers? How do you scale that?
I call this a “social scaling problem.” It’s almost impossible, I would say, for organizations to scale with social by themselves. That is, you can never hire enough employees to scale with social. So, the best way I think for organizations to scale with social is through social. That means they have to start collaborating with their customers who are outside the company. For example, crowd source for ideas; crowd source for support, and eventually they co-create value with the customers.
So that's one area I think a lot of people miss with respect to social.
There are certainly many aspects of analytics that people don’t get. One of them is the usage of data. Companies often said they want big data, and I think it is important, but what's more important is smart data.
You can have all the data in the world, but, if it doesn't solve your business problem, then it won't do you any good. If you have the right data, the very precise data that is just what you needed to solve your business problem, then even a few data points are sufficient. Big data may help, but more important is the ability to figure out the right data, the ability to pick them out and apply them in smart ways that address your business problem. I call this smart data.
DP: Okay. Wow. That'll give us plenty to think about. Thanks for taking the time to speak with us.
MW: My pleasure.
