In today’s smartphone-enabled, post-PC era, computing giant IBM is hedging its bets on Watson, the cognitive computing program best known for winning Jeopardy!. Marketing News spoke with Marcus Hearne, worldwide marketing director for Watson Analytics, about the program’s amped up predictive analytics capabilities and its effects on today’s data-driven marketing landscape.
Watson, owned by Armonk, N.Y.-based International Business Machines Corp.—better known as IBM—is perhaps the most famous computer program in the world. In January 2011, Watson competed against two contestants on one episode of the TV game show Jeopardy!—and won. It may seem like a computer could easily beat two mere humans in a trivia game, but to win Jeopardy!, Watson had to do more than look up answers to questions: It had to understand and use English to win the game. After much reengineering, the team at IBM had honed the program’s language skills enough to allow it to understand what Jeopardy! host Alex Trebek was saying, and to answer his questions in understandable speech.
Since then, the team at IBM has been working to put Watson’s data- and language-processing skills to broader use. Enter Watson Analytics, a cognitive software program designed specifically for marketing analytics. Cognitive, or predictive, programs are gaining momentum in the business world as marketers struggle to keep up with an ever-increasing amount of data, and they need programs to help analyze that information and enable them to act on the resulting insights.
Marketing News sat down with Watson Analytics’ worldwide director of marketing, Marcus Hearne, to find out how marketing and analytics dovetail today, and how cognitive technology programs like Watson are helping shape marketing going forward.
Q: Talk about your analytics experience and your career in marketing. How did they come together?
A: I started my career on the implementation and support side of analytics software. I got a business degree but very quickly learned that what was getting me out of bed in the morning was working directly with technology. I joined SPSS [a predictive analytics software company] in 1999, and at that point, rather than being a practitioner, I moved into the product management side and then quickly moved to the marketing side. The ability to communicate the value of this type of technology to people who had only clicking data [but didn’t have actionable insights] was something that I found challenging and rewarding.
SPSS got acquired by IBM, and I moved to eventually lead the marketing for predictive analytics. Meanwhile, IBM was developing Watson technology; it was the ingestion of a huge corpus of information to then give very straightforward answers to people’s questions in, eventually, a natural interface, in anyone’s native language. It became my job to communicate to people the value of this, how it works, and how it had really advanced analytics capabilities in addition to natural language processing and Big Data capabilities. It allows people without advanced degrees or 20 years’ experience in analytics to access this type of resource. A lot of people are feeling a need because there’s a huge gap in the marketplace for applied analytics knowledge.
Q: Watson made its way into pop culture with its appearance on Jeopardy!. Does that help or hinder your goal of communicating how Watson works?
A: It gives us great leverage and much more media recognition in the marketplace than we would have if we had to do it from the ground up. Watson really is the underpinning of cognitive [analytics], and IBM is about cognitive business now. We want to take our clients to that space where technology is able to see data in context, understand and learn from it, and interact with people in their natural language. … Now we have a multitude of capabilities—modules, if you will—in the Watson DNA. There are new solutions being spun off all the time.
There are points of hindrance because there will always be something that’s anomalous to the norm, but the norm is [that Watson] drives great conversations with clients. Clients are a mix, from, ‘Explain cognitive to me,’ to ‘We have an idea of how we think this could actually help.’ … They [might] have a ton of data, but they find themselves repeating the same solutions over and over. A cognitive technology will learn and start to bring that learning forward so you’re not solving the same problems in a different format every time. The Watson brand is opening up people’s imaginations, so it’s giving us a great entryway to have conversations with clients.
Q: How has the design of predictive analytics changed to accommodate the ubiquity of data and marketers using it?
A: It’s bringing it to the level where it’s accessible to more than just the data scientists, to more than just the person who has the Ph.D. It allows people to access advanced analytics very readily. That’s one big thing that’s important: take [the data] down to an individual level to give very ready access to it as is possible.
Have marketers always been the primary customers of predictive analytics programs?
It varies by role as to who is most ready to adopt predictive analytics. Marketers in retail absolutely understand customers and are able to scale interactions. Here, the Holy Grail for marketers is this ability to do a one-on-one, personalized interaction with you to maximize your interactions with a customer. From the point of discovery to becoming a prospect to becoming a customer to having an ongoing relationship, how is that done most effectively and efficiently? What’s the best ROI on the dollar that maintains that high level of customer satisfaction? Predictive analytics is extremely powerful in this area because if you can start determining what factors are predictive of an outcome, then you can start toying with those factors to optimize that outcome and avoid negative outcomes.
It’s something very simple that we inherently know as marketers: If someone has to try multiple times to get a resolution to an issue, their satisfaction is going to go down. They’re less likely to do repeat business with you or recommend you to friends and colleagues. We know that inherently, but now … there’s a lot of data that precedes you being able to control that relationship, so you have to gather that, and Big Data makes that possible now. …
Big Data flipped the scenario for marketers. We used to get those insights from samples: We would take surveys or we would do a sample or focus group or something like that, then try to understand our entire population, either our target market or our existing customers, whatever the case may be. After Big Data, and especially with the digitization that’s available through the Web and the mobile platforms through Android and iOS, all of a sudden you can get the data on the entire population. You’re getting the data on the population to understand and predict what the next entrant is going to do. What’s your next prospect going to look like? How should you handle them based on everything you’ve seen across all of the customers and prospects you’ve dealt with so far? That’s where predictive analytics becomes really powerful. It can come through that existing population. Look for all the predictors and help you understand how best to deal with clients as they come to you in the future.
Q: What does it mean to interact with data in a ‘natural’ way?
A: As marketers, we understand that it’s more than just a transactional demographic when it comes to a person. It’s more than just how much they spend and what they spend it on, or what’s their gender and postal code and how many children they have. There are the attitudes of opinions, there are things that change, how the weather affects how people behave. This is the Big Data part of it. There’s the volume side, which is, for example, how you track millions of transactions. Then there’s a variety of others: How do I gather and connect the pieces that I have, such as the demographic? There’s social media data, which is incredibly unstructured. How do I bring all of these things together to get a holistic view of a customer?
Marketers know that they need all of those pieces to better understand and interact with customers to find new ones. That technology is now available. With cognitive [programs], there’s the ability to pick up the structured and unstructured [data], bring in social information, connect it to weather. … If two customers have the same transactions and they’re the same demographically, they’re probably quite different based on a number of other things: things they say, the weather where they live, whatever the case may be. The flip side is then for technology to understand the context of the data it’s dealing with.
For marketers querying some information, technology like Watson can look at a date and can put it in context and understand if the date is the date of the latest transaction, or if it’s someone’s birthday or the day the warranty expires. Even simple things like that, being able to prepare the data and understand exactly what this data means is really important. Marketers are not trained to be able to analyze this kind of information, though now it’s an imperative. That’s where predictive analytics comes in. Watson will learn new things, and it will never forget. As trends change and people’s behaviors change, it will learn that over time.
The marketer now gets the entire 360-degree view of the customer that they want, with all of the data—whether it’s structured or unstructured, internal or external—made available to them through the advent of Big Data technology, and then the Watson technology itself is able to put it together, put it in context, understand it, learn from it, and then deal with the marketer and the language that they understand.
Q: What will data analysis, and specifically predictive analytics, look like in the next year, and in the next 10 years? Will marketers even need to touch data?
A: There are two answers to this. The first one is that it becomes embedded and invisible in the systems. Today, when people are looking at dashboards of information, no one sits there and thinks, I’m looking at a report that’s built on prescriptive infrequency statistics. They just say, ‘There’s my report. What’s my average deal?’ That was one of the great evolutions of business intelligence: It made generalized analytics available to everyone through the use of dashboards. That’s where predictive is going to go. It will become so inherent to the system that no one would think twice that they have something like predictive analytics under the covers helping them.
Then there’s the issue of how people will be interacting with predictive analytics in 10 years. Today, everyone can be very diagnostic. I can tell you through a predictor what size coffee I’m going to buy in the morning is pretty much solely [based on] what time I had to get up. If I’m up at 5 a.m., it’s almost 99% guaranteed I’ll buy a large coffee. If I’m up after 5, probably a medium. We’re already very diagnostic, but where it will go is very prescriptive—prescriptive towards optimization.
Today, a marketer would sit down and simply interact with a system to diagnose a situation and then think about potential next steps or things to avoid, or how can I go out with a better marketing campaign. Prescriptive optimization would let a marketer know that given their parameters, their desired outcomes are XYZ, and they have a budget of D. Then, prescriptively speaking, they should be doing ABC. Marketers will always have room for creativity and that’s absolutely required, but it will become invisible. It basically will run models of future scenarios and say, ‘With your budget, given the segment you’re after, your geography and all the other predictive [information] available to me, here’s a recommended campaign or action or command. … And here’s what it is and what you should be doing.’
Q: What advice would you give marketers when they’re looking for an analytics tool?
A: The key for anyone shopping for an analytics solution is to look for one that doesn’t make the mistake of letting their bias drive the outcome. If you see revenue going down in certain territories and you gave the marketing and sales data to a marketer and to a sales person and asked them to find out why revenue is going down, they’ll both come with bias. A sales person will look at head count, covering closure, [length of] a deal, where the leads are coming from, how old they are, etc. Those are all the right things to ask. But a marketer comes in completely differently: How many campaigns are being run? How many leads? Which sellers are picking them up? Are they letting them age? There’s always an inherent bias. When looking for a tool, remember that bias is such a big part of our daily lives, so you need to find a tool that is unbiased in its analysis and presents the cold, hard reality for you. … Of course, there’s always the human element of judgment against what is done, but at least the information that’s being presented, the analysis, and the results therein are unbiased. That’s the first step, and the second step is looking for tools that other customers are making use of and succeeding with. Who’s got a good story, not just of adoption but of results?
The appeal of this cognitive thing is that it removes that requirement for expertise in areas beyond the one that you should have expertise in.
A marketer should have expertise in marketing: their offering and customers. They certainly, by and large, shouldn’t be required to be experts in analytics. What we’ll see in the very near future—and we’re seeing it already—is that more and more of this technology is becoming automated, intelligent and directed very specifically at what that role’s problems or issues or challenges are, and solving for them very quickly. Cognitive data analytics provides everything from the data preparation to the analysis to the output in the most appropriate format so that ideas, insights, decisions and everything else can be communicated amongst marketers very quickly.
