Michael Wu is an influential and inspirational data scientist. The projection of his work reaches and moves the global community of analysts, data professionals, and data enthusiasts. He is also an influence for many global C-level decision-makers that understand and embrace the data-driven world we live in.
He is a leader and a vital member of the visionary team behind the innovative technology of Lithium Technologies, a company pushing the boundaries of digital customer experience. His work at Lithium touches directly many points that are extremely relevant for us in this book—social media marketing, analytics, data management, customer relationship management, and the creation of true communities around brands.
Beyond his work at Lithium, Michael influences the global discussion by speaking in many events and publishing articles and books on many topics related to data science, data analytics, artificial intelligence, machine learning, and on the “human side,” which he mentions as the top-down approach, considering the connection of social sciences and human behavior to the development and use of technology.
I first came across his work through one of his books, The Science of Social (Lithium Technologies, 2013), years ago, during a peculiar moment in my life. I had just been given a Kindle e-reader and was excited that I could easily continue reading everything as I commuted through the subways of Prague to work and home, so I was on the hunt for new reads. During that time, as it happens to many of us riding in public transportation, my mind while catching those trains and watching all those people would flutter over ideas, observations, and assumptions of what could motivate each person to do what they did.
I would try to relate what I saw to what I was learning in my work with marketing and social media, asking myself, What shaped their habits? Their attitude? Were they influenced by the advertising around them? Why were they using a certain brand of headphones? Why wear the sports team jersey? Were they comfortable with themselves or trying to become a new version of themselves? Were they experimenting in life? Would they take risks? And the key questions to us in this book, would social media influence any of this? How? How could I measure it? How could I collect such data? How could someone improve that human connection through social channels? How could the voice of people be heard by brands? Where was the connection between people as a community more than an audience? How to use data to reveal all of that?
Then came along The Science of Social with a holistic take into the universe of social media. I was immediately hooked. That content was speaking to all my questions and giving me possibilities that I had not considered before. It spoke of data, analytics, but also of community, true engagement, gamification tactics, and more. I read book one and two in three days with my new e-reader and bumping into everyone in the subway. Then I began to hunt for more content by Dr. Michael Wu.
I reached out to him back then just to be in touch, basically as a fan and admirer of his work, which I am even more today. When the time came to write this book, I immediately reached out to Michael again, to capture some of his views and thoughts and broaden the range of our studies here. In this interview, I then tried my best to capture a little bit of everything that he studies which can be related to social and to this book. It is a conversational material which will likely open our minds to thinking “outside the box” when it comes to social media analytics strategy. Michael also inspired me in adding prescriptive analytics into this book, explaining to us—as you will see in this interview—that the technology for doing so is already here.
Question 1
One area of your work that I feel is very interesting can greatly add to our need for broader knowledge when working with social media analytics and making sense of the data. This is the concept of building a true community. This understanding can help analysts better interpret their results and better suggest solutions for improvement.
Exploring your view on the concept of community related to social networks, how can we differentiate a true community, understand the differences to a network, and eventually explore more details into elements that compose the concept of a community?
To summarize the difference between community and network, we can start by truly separating social media into these two main categories—community or social network—and know that sometimes they can eventually be a hybrid. We can see this as two ends of the spectrum, where in the middle are the hybrids, and some are more community-like, others are more social network-like.
The main difference that distinguishes one from the other can be seen when we look at what actually holds them together. What holds the group together? Both social networks and communities are groups of people. What holds a social network is essentially interpersonal relationships. And what holds a community together is basically a common interest.
Many social media technologies are seen as networks, when they’re actually a community, with a common interest. The common interest for YouTube is basically videos. The common interest for Instagram is photos. Blogs are basically the written content. The more specific the common interest is, the more niche the community becomes. There are communities around, for example, HDR photography. That’s a very specific kind of community.
The social network, on the other hand, is held together by interpersonal relationship. For example, Facebook is a social network. It is held together by interpersonal friendship. There’s also LinkedIn, also a social network, held together by interpersonal relationship in the professional realm. There are a lot fewer of these social networks, and the kinds of relationships that people care about there are then friendship, kinship, the professional relationship, and so forth. Everything else in between will be a hybrid.
Twitter, for example, is a hybrid. You could follow someone because you know someone, so there is interpersonal relationship, and you could also simply follow someone because you find what they say interesting, the common interest that holds these groups of people together.
Author’s note: The conversation on this topic continues throughout the interview, but if you have further interest, the article at http://lith.tc/2u1M8ZO can give you a lot more information.
Question 2
Going further into that line of thought, how do you see these concepts fitting into a marketing strategy?
We can approach it from a more scientific perspective first and then look at the application in marketing.
In our life, in everybody’s physical life, we actually have a single social network that consists of pretty much all the people that have any relationship with us. That is our social network. It includes our friends, our colleagues, our relatives, as well as anybody who has any relationship with us. That’s our physical kind of social network in the real world.
Technology, or a technological platform like Facebook, will typically highlight a part of our social network through a social graph that captures some relationships. For example, Facebook captures our “friendship graph”; the relationship that it captures is obviously friendship. Whereas LinkedIn tries to capture a different kind of relationship in our social network, that can be called the “coworkers graph” or “colleagues graph.”
We all have our unique social network consisting of many different social graphs, but at the same time, we are part of many different communities. Some people, who are connected to us on social network, will also be connected to us through common interest. I am a photographer, for example, so I like photography and have some photography friends. Other people may have friends from grade school, university, or companies where they worked, and still keep in touch with these friends and colleagues. This is also my case; I am still in touch with my friends from when I was pursuing my PhD in neuroscience.
That’s how it works in our real life. Anthropologically, that is how community complements social network. There are actually functions for each one of them. The functional difference between community and social network is that community is where you build relationships and social network is where you keep those relationships even when you moved on from those communities.
If you take a look at your own social network, and you look at all your Facebook friends, for example, you actually share some common community with them in the past—whether they were in the same kindergarten as you or same college or same company. It is actually within those physical communities that you have built relationships.
I think that community and social network could actually complement each other a lot more when it comes to marketing. Many marketers will overlook the importance of communities and only try to find the big numbers, always driven by the idea that anything big is good. Consequently, they tend to focus only on the big social networks and miss out on the more niched communities. Social networks tend to be bigger because they’re more generic. Everyone has relationships. Communities tend to be niche, because people have all kinds of different interests that can be very specific, which tends to result in smaller groups for each. As a result, the marketer who is only looking for the big numbers will gravitate toward platforms like Facebook or LinkedIn and forget that where you actually build a relationship with your customers is within a community.
That’s actually where Lithium comes in to help. We can provide a platform that will help marketers and companies engage with their customers so that they can build a strong relationship with their customers. Once they build these relationships, they will see the positive results.
A strong relationship with a customer is typically manifested in loyalty. If you’ve built a strong relationship with your customer, then your customer is going to be a lot more loyal to you because relationships are pretty long term. You don’t just get rid of your friends. Once you develop this friendship, once you develop a weak tie into a strong tie, and they become a stable part of your social network, those relationships tend to last quite a long time. Before they become a strong tie, of course, things could still be fluid, but once that relationship is built, then it becomes stable. And if that strong relationship is with customers, then these customers will tend to stay with the brand much longer. That’s what I mean by loyalty.
The manifestation of loyalty is then essentially repeated business. They will do business with you for a much longer period of time and hence have much greater customer lifetime value.
Question 3
A main keyword for us in this book is strategy. With this in mind, and part of the reason why I’m so curious about topics apparently unrelated to analytics, is because people arrive at analytics with some kind of goal in mind, which in many cases may not be yet well defined. They will be asking, and sometimes looking for a direct answer from analytics, on questions such as, “How can analytics help me have my customers buy more?” Analytics will help, but many times the ultimate answer will also relate to a broader strategic approach, such as the better use of community. With this in mind, how do you see the use of analytics related to goals and broader strategic options?
I think it is important to have a goal in mind. What are you trying to accomplish? That is actually one thing that we advise. Even if you’re trying to start a community, what are you trying to accomplish with that community? Are you going to use a community to reduce your support costs through peer support, or are you trying to use it to drive more awareness of your product and your brand? Are you trying to get the community to talk about your product so that you can improve upon it and build better products? What is the use case? What are you trying to do with the community? After you have that, then you can formulate a strategy around, “Okay, I’m going to use this community to do a certain thing.” I think it is important to have the end goal in mind from the start.
In terms of analytics, looking at descriptive analytics, it typically tells you how you perform related to your goal. This can start from zero, when you don’t have anyone in the community yet, no content, nothing. Then the goal comes in: What are you trying to get? Are you trying to get to a million members by the end of year one or year two? Are you trying to get to a community with lots of content, say, a million member posts by the end of year one? What are you trying to do?
Then you can go into further goals. What are those members and posts supposed to drive? Are they supposed to drive more discussion about your product? Or recommendations about your product? Or reviews about your product? Or other business outcomes? What do you want the community to do for you? The descriptive analytics will tell you how you progress along this trajectory.
It can happen, for example, that your goal is to have a million posts by the end of the year, and by month eight you’re not even halfway there, so you know you probably will miss the target. You know you need to do something about it. Maybe the answer is then to use gamification to motivate people to drive more activity, or run a contest, or launch a superusers program. There will be many different tactics that you can employ once you know where you want to go and your progress toward this goal.
Question 4
Within this descriptive field, how do you see, for example, things like competitive benchmarking, thinking about the benchmarking of communities and social networks? When can it be interesting to bring such elements into analytics and strategy?
That is very interesting. We also have a benchmarking service in Lithium. The beauty of being a SaaS platform is that we basically host all the communities on our platform. What that means is that not only do we have access to one community’s data, but we also have access to all of them. We can actually benchmark, to some degree, on whatever similarity or criteria that’s relevant to you.
For example, if you are a retail brand manager and you say, “I want to benchmark against all the retail communities that are as old as mine,” you can. We will filter out all the other irrelevant communities that are not retail focused and that are too young or too old. Say you are a community that is three years old, we can then look at communities that have between two and four years. We could take a look at their data and then look at how they progress from day one, the launch, all the way to their current stage.
There are privacy restrictions, however, so typically we would do this benchmarking service if there’s actually more than five communities in the benchmarking group. The averages over communities within the benchmark group will then hide details of who is in the group and what exactly they are doing. If there are not enough communities in the benchmark group, either we will have to relax the similarity criteria or we will not offer benchmarking.
Going beyond communities, competitive benchmarking on social media will follow the same concept as doing any competitive study. Why do you want to know anything about your competitor? It will help you focus on what you do best and not compete head-to-head on what your competitor does best. Maybe instead you can focus on improving something that they’re doing poorly. Having the customer’s voice in this process is actually very important. This is basically the realm of, I would say, social media monitoring services.
Social media monitoring can listen to a social media conversation and tell you how many people are talking about a competing brand compared to yours. It can also tell you the sentiment around their conversation about the brands. How many mentions have positive sentiment? How many mentions have a negative sentiment? All that will tell you something about where should you focus, and therefore help you formulate a strategy to win in this hypercompetitive market.
Question 5
How does the idea of community and social network come into play when a brand is trying to use channels such as Facebook and Twitter?
I think it is important for marketers to also engage on these social networks precisely because they’re big. These channels become amplification platforms for the content created by the community. If you only have content that stays within the brand community, and you don’t have any medium of propagation, then the community doesn’t actually get anywhere. You could certainly build a community that’s pretty big, but then typically communities don’t get too big because people don’t just have only one interest, and those interests go beyond your brand.
The key is then really to use community and social network together, to have them complement each other. You use the community to build relationships to create relevant and trusted contents about your brand, and you use the social network as a propagation medium to help you spread these contents and reach a wider audience.
For example, if someone wrote an excellent review about your product within the community, you should try to propagate that out onto social networks and let them create what is sometimes called the “viral loop.” Bringing analytics into that, thinking on the example with social media monitoring, you can then monitor the propagation of this content and see if it will increase the positive sentiment about the brand. That will help you formulate your strategy and decide what to do next. If you see the positive effect growing dramatically and it’s actually improving sales, you know that this specific propagation strategy is something that you should continue to do. If there is growth in activity but no correlation to sales, then maybe it’s ultimately not that useful. Maybe you should focus on doing something else.
Question 6
Going a little further into communities, and taking reference from social media channels where there is a member of the brand as a moderator, or participating in the discussion to answer questions, and so forth, how does the flow of interactions happen within communities?
Within the communities that we host for the brands, typically all the members are actually customers of the brand. This is another important difference between off-domain social channels (like Facebook or Twitter) vs. on-domain communities. Social channels like Facebook and Twitter are often perceived as the voice of the brand, whereas the communities we host are perceived as voices of the peer (i.e., other customers). One of the consequences of this difference in customer perception is that customers are much more critical and negative on social channels than they are in communities. We have many customers that have both community and off-domain social channels, which we encourage. They often see many complaints on the off-domain social channels and are surprised to find their members so cooperative and helpful within the community.
Even though there’s nothing stopping the employee of the brand to participate in the community, they are a very small fraction of the community’s population. These customers within the community are the ones who will essentially influence others, and they’re just like you or any other customer. They’re not experts in the product nor do they get paid by the brand. If they’re helpful to other people, then that actually creates value and people appreciate it. They slowly grow their influence in the community and eventually become recognized by the company or the brand as an expert. The company can then reward them positively for what they have done. Maybe not necessarily monetarily, but they could certainly reward them by other nonmonetary means to encourage them to continue their positive behaviors.
For example, “we invite you to beta test our new product,” or “maybe you could come help us design the next version of this feature?” Or “would you like to come over and meet our senior leader team or have dinner with our CEO?” Or “I’ll give you a ticket to come to our conference” or offers like that. All these are ways that the company could reward these influencers, which sometimes we call the “superfans” and the “superusers” within the community.
Question 7
Going into a different topic and looking to the future, how do you see the arrival of technologies such as machine learning and artificial intelligence? What challenges and limitations are we facing? Why isn’t everyone using these technologies yet? Is the technology truly available, or are we not quite there yet?
The current limitation is that most companies don’t collect enough data to train the AI model. That is the issue. Companies that have enough data are already able to produce predictions that are very good. They could even prescribe actions that are somewhat optimal or better than a human could. So why isn’t everyone using these predictive and prescriptive analytics yet? Because most people don’t have all the data that they need to build this type of AI model. It’s not that the technology is not there. The technology is actually already here.
Examples that go from algorithmic trading, automatic loan origination, or even self-driving cars, which I wrote in many of my articles, show that the technology is already here. The reason why they are here now, however, is because there are efforts, huge efforts, invested in collecting lots and lots of data.
Google is spending lots of money to have cars driving around collecting data about the road conditions, traffic patterns, and different kinds of weather conditions. They will then have lots of data, right? That’s why there are self-driving cars and the artificial intelligence that tells the car how to drive from point A to B. The AI is able to drive the car very well, just as well as a human.
In investment, algorithmic trading already exists, although not fully automated. But capital-trading companies have years and years of data. It is not huge data, though. You could load the past 20 years of the entire stock market into your laptop. It’s not that big; it’s all just numbers. Today, there are actually a lot of other data that we could collect and add to the stock market data, such as the market conditions, news, public sentiments, political events, even weather and disasters like earthquakes, and so forth. Having enough data, the investment firms can then build something like a robo-advisor that can automatically trade for us. I actually signed up for one of these robo-advisor types of services myself. There are several such companies out there doing that, like Wealthfront and Betterment. So the technology is here. I think a lot of companies simply don’t have the data that is needed to fully utilize these technologies yet.
In the case of marketing, we can then question ourselves, “Can I actually make the AI smart enough to help me send, for example, the right emails to potential buyers so that I don’t have to do marketing?” You could, but you have to train your AI so that it knows what to do. You have to show your AI what to do with data, for example, “Okay, this is what should happen when a customer exhibits this behavior, and these are all the different behaviors of a potential buyer. So first, you send them this, then you do that,” and so forth.
You need to have all this data so that your AI can learn from it first, before you could actually have an AI that is smart enough. Without enough training, the AI will be very “dumb.” Your AI can only be as good as the data that you have. If you don’t have the data, then your AI is just a “stupid” or “dumb” computer.
Question 8
That is very interesting. It makes me curious to ask you about the need for data integration in this process. How do you see the integration factor coming into play within this challenge of collecting enough good data to be able to train the AI? Is data integration a challenge?
Yes, I think that is a real issue. Many companies have lots of data, but the data is dispersed. That is not very useful because the AI has to be able to see everything in order to learn from it. If you only feed the AI financial data, for example, and build a model, the model will be able to forecast the financial aspect very well, but may not be able to respond to the other conditions, such as marketing.
It is important to have what we call a “fairly complete coverage” of how your business operates. That’s why one of the themes that is very heavily discussed nowadays is the idea of digital transformation. Why do you want to digitally transform your company? When you digitally transform your company, eventually everything, every part of your business operation, becomes digitized.
When that happens, it means you can have data on every part of the business operation, whether it’s marketing, sales, support, or just a day-to-day operation; everything can be digitized and become data. Those data can be integrated. You could then actually build a very massive model about how your business operates by feeding it all these integrated data.
By providing all the business operation data, you essentially teach the AI how to operate your business, so that eventually it will be smart enough. Then it could help executives make decisions instead of the executives having to go through all the data themselves, which is actually what we are doing at this point. We look at all the data from different parts of the business and then we make a decision. Once you feed all that data to the AI, and you train it with your strategy, for example, “under this situation, I make this decision, and under that situation, then I make that decision,” the AI will learn to mimic how you make decisions. It eventually makes decisions like you would have, because you used your data to train it.
Question 9
Stepping into the role of the marketer, to create an even better bridge with the concept of this book, and taking our conversation so far as a basis, how do you see the marketer navigating through this data-driven world? Would a marketer need to code, for example? How far would a marketer need to go to fully participate in the digital world?
I don’t think it’s necessary for marketers to do coding. That’s not what they’re necessarily best at doing. They need to know enough about coding to work with a data scientist who can code. They need to be able to, for example, tell the data scientist that, “Okay, I think these are important variables that we should consider, this is the context.”
They definitely need to know how to collect data. That’s an important skill for future marketers. They also need to be able to explain to the data scientist how to collect it. Are there any systematic bias? How did they sample their audience? Those are concepts around data sampling that marketers need to understand well.
If a non-savvy marketer goes out and collects some data, say, from smartphones, and tells the data scientist, “Okay, I built this app and I collected the data.” The data scientist then asks, “So is this biased or not biased?” The marketer answers, “Oh, I collected from everyone. It’s not biased.” That would be an incorrect assessment of the bases within the data. The market for the brand will likely consist of a subset of the entire population. When it comes to smart devices, not everyone has them, so you are actually sampling the data, collecting the data from only a biased group among this entire population.
These are the things that marketers need to learn and understand better at the conceptual level. They need to understand statistics and data sampling and biases so that they can actually talk to the data scientist. The data scientist would then write the code and maybe try to adjust for the bias when they know what is causing the bias.
In a certain region, for example, where not everyone has a smartphone, it could be known that only about 70% of the people have smartphones, and the poorest 30% don’t. Results can then be adjusted based on the fraction of the population that actually has a smartphone.
For marketers who come from a creative type of background, the data collection aspect is very important. Without good data, everything is off. You need to collect good data and understand how your instruments do that.
If there is any bias, you need to understand what the bias might be and what created such bias. Sometimes you may intend to collect data from everyone, but maybe there is some bias that crept in there just from the design of your app. Maybe the design of the app, or whatever data collection device you have, is not completely free of bias. You designed it in a certain way. Inadvertently, you created some bias by designing it a certain way and not choosing to design it in 100 other different ways, and so forth. Not only you need to understand what those biases are, you must understand what is creating those biases also.
Question 10
Going into the last question and taking such broader aspects around the understanding of data and the business into account, how important can it be for a social media marketer to step out of the social media “bubble” and into further aspects of the business and of the technology landscape?
Stepping out of social is actually the first step toward proving the ROI of social. If you only stay in social, all the data and all the analytics and metrics that you’re going to get are what we call social operation metrics. You’re never going to get financial data or business KPI data. You’re never going to get that from your social media platforms.
You have to step outside of it to look at other parts of your business and see how social operation is related, or how it is correlated, to the other parts of the business and their metrics. That’s how you can actually prove ROI and prove that your work is actually making a difference in the business.