IBM acquired predictive analytics vendor SPSS in October 2009. Erick Brethenoux, predictive analytics strategist for SPSS, an IBM Company, talks about the growing importance of the technology in helping enterprises address customer needs, what is driving the demand for it now, and how it fits into IBM’s idea for a Smarter Planet.
DBTA: Can you explain the difference between predictive analytics and business intelligence?
Brethenoux: If you are using business intelligence, you can ask, as a sales manager, for example, who are my best customers by salespeople, that have been answering the promotion that we set up 2 months ago. You will get a list of 10, 20, 30 - depending on what you are asking for - people, answering the question. Now, why these people bought and what they are likely to buy next, you don't know - because BI doesn't do that. That's what predictive analytics does. That's the big difference between the two. Business intelligence will give you what happened in the past and what is going on right now, giving you an answer, provided that you have the right question. Predictive analytics will give you a view of what is going to happen next, a look into the future, and also be able to uncover some very special trends, very special behaviors that you might not have thought about, because you didn't know they existed.
DBTA: Such as?
Brethenoux: If you are telco company, and this is a real case example by the way, and you get a report of female customers that are 70 and above, that have been churning quite substantially over the past two years, you cannot fathom why - and you can search and ask all the questions you want - unless you use predictive analytics, which will tell you that in fact the trend among those women is that they live with or very close to their grandchildren, who are often granddaughters between the ages of 12 and 17. And those young ladies give their grandmothers advice on features in new phones that they should be getting - such as bigger numbers and larger screens. Unless you have a technology that is going to draw all these different trends that you were not expecting outside of your data, BI will never be able to tell you.
DBTA: Do you need both?
Brethenoux: They are very complementary in nature, but you don't need one to do the other, meaning that you can do predictive analytics even if you don't have BI tools in your organization or vice versa. If you have both it is better, obviously. And now that you know about all those female customers, you can ask, is it true for males as well, or is it true in all regions of the U.S. - so now I know what other questions I may need to ask in order to direct campaigns, so BI becomes interesting there. And then I may find another correlation, and think, that doesn't make any sense, and go back into predictive analytics, and so on.
DBTA: So predictive analytics can help organizations determine the products they should be bring to market and what their customers will need?
Brethenoux: Yes. I think there is also another value proposition that is actually more fundamental to this one, although this one is the outcome. As people get to know each other well, they make inferences about each other, and one can make a joke and anticipate whether the other will be able to laugh about it or be offended, making predictions like that.
DBTA: How so?
Brethenoux: Our technology provides a way to engage that kind conversation, that kind of intimacy. Our technology is able to build a profile, an understanding of the customers, the constituents, the patients, the students, the citizens - between that and the organization - and replicate that on a very large scale. For example, if one company had a million customers and a million products, we could match them one on one. Our technology can do that. It is kind of scalable intimacy. That is what the technology provides at a fundamental level.
DBTA: Who uses predictive analytics?
Brethenoux: We tend to categorize the problem that we solve in five main categories. We help organizations to attract new customers, to sell more to the existing customer base, to retain customers that are profitable and that the company wants to retain, detect fraud, and manage risk. These five business problems are inter-related.
We have mostly been talking to the retail, banking, insurance and telecommunications industries and the public sector. But you find us everywhere. We have 250,000 clients worldwide, unique organizations to whom we are selling our software, so you also find us in manufacturing, pharma, healthcare and other industries. I don't mention academia, because it is kind of obvious, given that it is where we started and where we are still well- known.
DBTA: What are the trends moving this forward?
Brethenoux: There are three main external trends that have been pushing the market in this direction. One is an increasing reliance on mathematics in making decisions; and you see that in television shows like Numb3rs, and CSI, best-selling books like Moneyball, and the movie 21. Mathematics is increasingly trusted and important in society, and it makes its way up of course inside organizations, as well. The second thing is that right now the rate of producing data is exponential on a daily basis. It is amazing. And, it is only the beginning; wait until RFID, and all these devices start to produce data. Organizations are starting to really struggle now, they have been gathering a lot, but they don't know what to do with it.
DBTA: And the third?
Brethenoux: The last trend that has been important as well has been the fact that there is a real return on that investment right now. We have actually measured it consistently in the organization and it comes to a point where not doing it becomes a competitive disadvantage, so you are really being left behind. In telecommunications, for example, where most of what is going on is a zero-sum game, not using that technology means that you are losing customers on a daily basis, and reacquiring those customers is a very expensive proposition, so it really impacts the bottom line quickly.
DBTA: So companies would have a blind spot if they are not using predictive analytics?
Brethenoux: Absolutely. You're taking a risk by not using it now. Before, it was an advanced technology. It was an emerging technology. People were taking some risk in using it-so the usual suspects like telco and banking companies, who try everything, of course were there. But now we are getting to retail where the margins are fairly small and people are saying, we can't market to all these people, we have too many channels. We have to decide which channel, who we are talking to, what we can expect, and what we market to them. In order to know these things, you have to know who you are talking to. You have to understand those customers better and try to anticipate what they are going to do next.
DBTA: Were IBM's acquisitions SPSS and, before that, of Cognos, seen as compatible?
Brethenoux: There is another company that IBM acquired that is actually as critical as Cognos has been - ILOG, a business rules and optimization vendor. The intelligence that you can provide is enabled by two things, rules and optimization, and predictive analytics. Those are the intelligent technologies but you have to report on it, you have to let people know what is going on in systems and let people react to that as well - line-of-business managers or simply managers themselves - and that is where Cognos dashboarding and reporting comes along. Cognos is a monitoring and measuring device that comes then to promote the data and promote the action that has been taken by ILOG and SPSS in the background. That is where the intelligence comes in.
DBTA: Data integration is fundamental to these technologies.
Brethenoux: It is, but you don't have to integrate all of the data. Predictive analytics will tell you for example that in the 1,000 variables that you have about a customer, only 15 or 20 will be predictive, the rest are nice noise in the background but won't allow you to predict anything. You need to make sure you integrate the data that is relevant. Predictive analytics gives you a way to identify the relevant variables and the relevant data that you need to integrate so that it is not an insurmountable effort. It just guides you to what is important to your business or the business problem you are trying to solve.
DBTA: What are the applications for predictive analytics in social networks?
Brethenoux: For us, social networks are becoming an important source of data for the analytics. We are not unfamiliar with that because SPSS originally also had a piece of software for surveys. We have been saying for a long time that knowing the demographics, where people live, knowing the transactions that they have had done with an organization and interactions that they have had with a call center, was not enough for the organization to predict what is going on.
You need to get their opinion, their mood, their sentiment, what they want to do. Before - even 5 years ago - the only way to do that was to put surveys out, to ask them. Today you don't have to do that. You can go out and look at where they have been contributing in collective places. We have been used to look at information to complement what we do in predicting what people are doing and we also have text analytics capabilities that help us understand the opinion, the sentiment, the negativity or the positive feelings that are being expressed online.
DBTA: What are some of the internal forces within IBM driving the emergence of predictive analytics?
Brethenoux: The internal forces are linked to the Smarter Planet idea. There are three components to the Smarter Planet idea. One is instrumentalization, with data coming from many different things-supply chains with RFID produce data, scanners produce data, GPS in cars is producing data, it is coming from all over the place. The amount of data is overwhelming. The second tenet is about interconnectivity with everything connected to anything, with Wi-Fi, with cell devices, with computers being hooked up anywhere. The data flowing as fast as it is produced freely everywhere. And the third tenet is intelligence - how do you make sense of the amount of data that is being produced and the way it is flowing, and where do you send it to make sure the flows are productive. The intelligence capacity with ILOG and SPSS and then Cognos is actually the reason IBM has been making these acquisitions, to instrumentalize and make sure they can actually formalize the intelligence that is needed to see what is the good data and the bad data, and what you need to do to solve problems.
DBTA: How does predictive analytics relate to the IBM concept of the Smarter Planet?
Brethenoux: At the beginning when I joined IBM, I was very surprised to see how deep that concept goes. We talked about instrumentalization, interconnectivity and intelligence but how is that possible in cities? Well you can get data from taxis, from the police forces, education systems, the city and service systems, the electric grid of a city and the water grids as well, and then communications. You can go deeper and deeper and identify problems very quickly before they even happen, linked on the demand from people and linked on the infrastructure that you have to monitor as well. That is where predictive analytics is going. It is going to be embedded into absolutely every single decision that people are making. It is going to result sometimes in a recommendation and sometimes it will be automatic because we know that there is a sign of bigger problem that has to be fixed right away. Even now, in everyday life, whether we are aware of it or not, predictive analytics plays a role.
DBTA: If the future goes according to plan, how will things be different 5 years from now?
Brethenoux: If the scenario plays out, there is going to be a lot less waste in many different kinds of systems. You will be paying less taxes, getting better services, we will be able to have much more reliable networks, whether they are electric, or telco or water, and part of what we are looking at as well is that we can stop being inundated by ads, without infringing on our privacy. There is a delicate balance between the two and we are working hard to preserve that balance. That is where we are going.