Predictive Analytics has been used for some time in exploiting patterns among data, and Big Data has now removed the limits on what can be analyzed. But there’s a difference between prediction and persuasion. If Predictive Analytics can’t lead to something actionable on a business level, what good are they? So how can we define what makes Predictive Analytics effective, persuasive to business leaders and capable of helping the business achieve a positive ROI?
Contributors

• • Eric Siegel, Founder, Author, “Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die”, Predictive Analytics World
  • Jack Levis, Senior Director of Process Management, UPS

Transcript (How to Make Predictive Analytics Effective)

Speaker A [00:00:00]:
CIO Talk Radio is brought to you by HP and Bosch Software Innovations. Welcome to CIO Talk Radio with your host, Sanjog Aul. All. All comments, views and opinions expressed on this show are strictly those of the host, guests and callers. Now here’s Sanjog Aul.

Sanjog Aul [00:00:26]:
Hello and welcome to CIO Talk Radio. To learn more about the show, please visit www.ciotalkradio.com and as always, we invite you to join the discussion on Twitter Ctrlive and look for this show as Predictive Analytics. Today’s topic is How To Make Predictive Analytics Effective and our guests for today’s show are Eric Siegel. He’s the founder of leading cross vendor conference series Predictive Analytics World which takes place 10 times a year across industry vertical including PAW government next month in Washington D.C. PA Boston and PA Healthcare in October as well as Paul London and Berlin events this fall. Eric is also the author of best selling Predictive Analytics The Power To Predict Who Will Click Lie Or Die. Hello Eric, how are you doing?

Eric Siegel. [00:01:18]:
I’m doing well, Sanjog. Thanks for having me.

Sanjog Aul [00:01:22]:
Definitely. And I see with all the different events and the book writing and speaking, I’m sure you must be keeping crazy busy.

Eric Siegel. [00:01:29]:
There’s been a lot of growth in the industry and so over the last few years, playing this role with the conference production has been fun because it sort of puts us in a central position across industry to see how much is going on. It’s really been growing quickly, so it is a lot of fun.

Sanjog Aul [00:01:48]:
Great. And we also have Jack Levis, who is the Senior Director of Process Management with UPS and he’s also giving a keynote address at the Predictive Analytics World Conference in Boston. Hi Jack, how are you?

Jack Levis [00:02:00]:
Fantastic. How are you, Sanjog?

Sanjog Aul [00:02:03]:
Very good, very good. Thank you for joining us. And in your world, where you come from, I’m sure there are a lot of moving parts and this predictive analytics must be playing a key role in the way you serve customers and the kind of experience you deliver, right?

Jack Levis [00:02:19]:
Absolutely. You know, UPS, we say we love logistics and it’s really hard to be an incredible logistics provider without having analytics built into your processes. So analytics plays a huge role everywhere at UPS.

Sanjog Aul [00:02:35]:
Definitely. Now, predictive analytics, along with big data is the hot new term in which every company feels they have to be involved and work together. Insights. But the ultimate question everyone has is how this will all lead to value that comes down to how well can predictive analytics not just predict, but persuade decision makers and business leaders that all this data crunching will produce ROI so today we wanted to explore just how to make this predictive analytics effective and what are the nuances to doing so. So, Jack, I’ll start with you. When you look at the different applications that may be very natural for us to use predictive analytics for, and of course, we could be like a kid in a candy store, which ones would you think, which areas, which applications would you think be a good ones to go after where predictive analytics is applied and it will give us the most effective results?

Jack Levis [00:03:36]:
Well, you know, I certainly love the topic because as you said, the key is not the methodology. The key is the benefit, the ROI that an organization is going to achieve so the truth is, you know, people talk about big data all the time. I care much less about big data, but what I care about is big insight and big impact. I think you need to understand that analytics is about making better decisions. So where I start is what are the big decisions we were going to make where analytics could have given us better information to make those decisions better? Because it’s the decisions that matter. It’s not the tool set, it’s the decisions themselves, which then leads to the insight and the impact.

Sanjog Aul [00:04:24]:
So when you look at your world, Eric, and see all the different people talking about predictive analytics and saying that we have tried a number of things, some have worked, some were less than stellar, and some were totally phenomenal in terms of the value that created, do you see a pattern?

Eric Siegel. [00:04:43]:
Oh, yeah, absolutely. In business, main application areas are in marketing, financial risk, kind of credit scoring and fraud detection and there’s plenty of other applications on the web and across sectors, including government and healthcare, etc. But within business, those are the main areas. Those are where for most organizations, and I think Jack’s story is a bit different at UPS, about where their main operational decisions are driven with data. But for most organizations, it’s those three marketing, risk and fraud and those are the big decisions.

Eric Siegel. [00:05:22]:
So Jack likes to say that he loves this topic and that’s what really excites him is how this technology is actionable, where the value takes place, not just the core sort of rocket science technology of the analysis and how the analytics work. I actually feel a little differently. I kind of feel like I have two personalities and there’s one that gets very excited about that and I’m also a big fan, separately from that of the core science. Let me just mention how we define predictive analytics, what that core science is. It’s to learn from data how to make predictions for each individual and oftentimes it’s individual consumer or healthcare patient. Sometimes it could be an individual store or corporate client, this kind of thing. In order to drive operational decisions.

Eric Siegel. [00:06:11]:
So it’s very much part and parcel to the definition that it must be valuable in terms of being action oriented.

Sanjog Aul [00:06:22]:
So Jack, when you look in this predictive analytics world, do you think that a company truly goes in an R and D or sandbox mode when they say let’s apply predictive analytics or would you say it all starts in a very planned, goal oriented fashion and that’s what is the key to us getting some concrete value out of this endeavor?

Jack Levis [00:06:46]:
From my perspective, it’s both. You need to start and have some kind of idea where you can make better decisions because it has to become actionable, as you said at some point. That being said, from my perspective, there’s research here. You don’t know what you don’t know. There’s an exploratory period where you have to find out what is possible and then you have to look to what decisions you can impact. I agree with Eric totally and I love his concept of predictive analytics. Gets down to the individual, which is different than a forecast. I love that thought process.

Jack Levis [00:07:26]:
But if it’s not actionable and you end up with some result where it just becomes a did you know? With nothing you can do about it, then predictive analytics will turn just into trivia, which isn’t what we want. So I do think there’s an exploratory period. There’s a period where you learn what you didn’t know and then you need to move from research to development with a plan on how to actually achieve gains.

Sanjog Aul [00:07:51]:
So Eric, at the top, when you go and take this proposal to the folks at the top, do you think you can get away with saying we are going to have some exploratory phase and that is going to be sifted through to see which specific areas we will dig further and then we’ll be able to show you value. Please give us million dollars upfront.

Eric Siegel. [00:08:13]:
No, you don’t get a million dollars for the exploratory. It’s a lot cheaper than that but if there is a little exploratory, certainly there’s some analysis before you can be entirely sure that it’s worth pulling the full trigger. But all along in the entire process you do have your eye on the goal. There’s that carrot at the end of the stick which is the action ability. How exactly are these predictions going to be used? Which operational decisions are they going to be driven, rendered mass scale decisions more effectively in those areas, such as I mentioned. Let me make those a little bit more concrete as far as the actionability. So in marketing, it would be targeting marketing, who’s it worth expending, the cost of contact, who’s most likely to respond or make a purchase.

Eric Siegel. [00:08:53]:
Also in marketing, who’s likely to leave, defect, attrite, quit? So that retention offer is going to be targeted. Who should we approve an application for a credit card that’s in terms of predicting credit risk. So there’s always an action or operational decision that’s in mind. It can be really cool and nifty and neat to create a predictive model, but we don’t do that just for the heck of it or just for the fun. There has to be the value, there has to be a particular way in which it’s going to be used. So I like to say that within all the realm of big data, which could be any and all kind of analysis and ways of making use of data, that predictive analytics is by definition the most actionable, valuable form of analysis because all those millions of per person decisions directly inform operational decisions on that level of detail.

Sanjog Aul [00:09:48]:
Jack, whenever we talk about crystal ball, which is when we want to sit down and we are relaxing and saying, what if I knew what the future is going to hold and my life would be different and our business would be different. Are we trying to go that route in terms of what we don’t know and we’re going to predict and we’re going to bet our paycheck on it? Or is this going to be a small percentage of what we can do by investing in a smaller fashion and then see what the results are and then kind of find our way through? Is it more navigating versus going for the kill?

Jack Levis [00:10:26]:
So let me, take a step back and Eric mentioned that UPS is a little different than others. Analytics is generally understood as descriptive analytics. Where am I today then? Predictive analytics, where am I headed? And as Eric says, down to the individual. And then prescriptive analytics is where you optimize and UPS has gone through that path of descriptive to predictive to prescriptive analytics. So I don’t know that we could have done the predictive stuff without having really cleaned up our descriptive models, get our data, get our data correct. Let me tell you what happened when we moved from descriptive to predictive analytics and this was in 2003. We already thought that we were the best in the industry.

Jack Levis [00:11:13]:
We’ve had 70 years of industrial engineering behind us and we’re known as the most, productive company in the world when it comes to efficiency. When we put our predictive models in place, we saw a reduction of 85 million miles driven a year. That’s eight and a half million gallons of fuel that we’re not purchasing just from using the predictive models and the associated methods and processes around it. So I don’t know that I knew that when we started that we would get that amount of gain. We knew it was an area where we could gain, but we did it a step at a time. And from 2003 to today, 85 million miles reduced a year is pretty good and by the way, there’s more coming.

Sanjog Aul [00:11:59]:
So it’s very encouraging based on Jack, your response and the example that you gave, that now you can feel proud that you went this route and essentially enjoyed the savings. It’s very encouraging for the rest of the world. But would you say that this was just you got plain lucky and the fact that predictive analytics did have the power in the first place to produce results, or there was some magic in the process, in the way you carried out a process of the way you carried out predictive analytics, this whole process where people, process technology all came together and made something happen.

Jack Levis [00:12:33]:
Well, it’s interesting, two of my divisions are advanced analytics divisions. I have quite a few PHDs and mathematicians that work in my group. But look at my title. My title is Director of Process Management. That’s because the analytics, the process, the procedures, the methods all have to be the same thing. So I don’t think we got lucky. It was a lot of hard work and we skinned our knees a lot along the way. The analytics did not just produce results day one.

Jack Levis [00:13:05]:
To be honest, it took a number of years before we started seeing those kind of results, before we saw the change management where our frontline people truly knew how to turn the dials and make the best out of the analytics. Now, looking backwards, it’s obvious. From my perspective, these type of projects go to three major steps. The first one is proving the technology that’s where the research comes in. The second is building the technology inside of processes themselves so the systems, the support, the IT systems. The third and the hardest project of them all is implementing it and truly getting the gains

Jack Levis [00:13:43]:
and we monitor that we have new metrics, etc. So it was honestly a lot of hard work using analytics and we really try to hide analytics. The front line doesn’t even know they’re using it they’re just doing their job.

Sanjog Aul [00:13:56]:
So, Eric, when you look at the way the landscape is now, would you think that organizations, while they understand the term predictive analytics, but they have been able to build the maturity in their organization so that they, in a, modeled fashion or in a predictable fashion, are able to skin their knees so that they’re not just, you know, oozing blood all over, but still with no results, effective results?

Eric Siegel. [00:14:27]:
Sure. I mean, there’s certain costs, there’s certain things that there are certain pitfalls in general, deployment goes well and there’s direct ways to mitigate risk and obviously the place where there would be the biggest risk is when you go to that third phase that Jack just mentioned, which is implementing another word for that is deployment. So you are deploying the predictive model. You have created this analytical model that was discovered from the data, learned from the data that makes the predictions. So now you want to use those predictions, you want to improve the operational decision making process so that it integrates the actual output of the model so those little predictive scores actually make a difference. That’s deployment, pulling that in and making a change to the way operations and business as usual is conducted

Eric Siegel. [00:15:20]:
and at that phase there could be risk. What if the models are predicting poorly? What if there’s some other unanticipated consequence? What if we make this mass scale change by integrating the predictive scores, the output of the models and something goes wrong? Well, the fact is that risk can be directly mitigated and actually controlled like turning a knob, simply by controlling the degree to which you implement or deploy. So you can start with an incremental step. You can say we’re going to use this 10% of the time and the other 90% going to continue in the legacy manner so that you can sort of do this head to head test and slowly increase the extent to which you’re relying on the models. So risk is something that can be directly mitigated.

Sanjog Aul [00:16:14]:
Let’s take a quick break listeners. When we come back, let’s talk about the people side. So we did talk a little bit about the process, the way you carried out. And Eric had suggestion regarding just doing some sort of an AB testing where 10% of the time you apply predictive analytics and see how that bears results versus the regular fashion the way you carry out. And so all of that while being done, what mindset and the type of people who need to be by your side for this to actually be a great journey to begin with. Please stay tuned listeners. We’ll be right back.

Speaker A [00:16:54]:
HP is proud to sponsor this program. Find out how the HP as a Service solution for SAP HANA can help you gain instant, impactful business results without capital investment. By logging on to hp.com transform information into intelligence and a competitive advantage with a full spectrum of SAP HANA products and services from hp, a global SAP hosting partner.

Speaker E [00:17:25]:
Bosch Software Innovations is proud to sponsor this program. Visit www.bosch-si.com Connected Manufacturing to find out how Bosch can help you improve your operational performance and become a manufacturing industry leader in a connected world. Change the way you predict, manage and produce outcomes. Bosch Connected Manufacturing.

Speaker A [00:17:57]:
HP is proud to sponsor this program. Tap into our expertise, innovation and services to bring your most important workloads to the cloud. You are listening to CIO Talk Radio with. To learn more about the show, please visit www.ciotalkradio.com. if you have a question or comment, call toll free at 1-866-472-5790. That number again is 1-866-472-5710. Now back to the show. Here’s Sonjog All.

Sanjog Aul [00:18:42]:
Welcome back. So we were to talk about the skills and mindsets necessary to make data predictive and deriving insights that would actually create value and that is the people side but before we get into it, Jack, I’d like to get your perspective on that approach where you can put 10% based on what Eric said like a 10% of the time when you run a process, you apply predictive analytics. It’s like an AB testing that you do in marketing and perhaps in that area you could do it but what happens in your space? Can you literally slice off your same process and say I’m going to apply predictive analytics on a portion of it.

Jack Levis [00:19:19]:
So I think what Eric says makes perfect sense. Unfortunately for ups, that’s much harder to do. I can’t create a driver’s route that says 10% of it is using a predictive model and 90% of your route is using traditional methods. However, that being said, as we move down our path of analytics, there’s a point in time where we say we’ll use the analytics as an assistant so the human is still in the loop. The analytics may say, here’s what I suggest, but we allow the humans to still override it because there’s things that they know, often subjective things that we allow the planner to override and as we move further and further and further, we measure what percent is overridden and what percent are they using the analytics by themselves and then we look to constantly update our analytics to improve it.

Sanjog Aul [00:20:12]:
So Eric, coming to the people side what have you seen the world doing? Do they kind of take the same people and perhaps add a few PHDs who can do real math, but then whatever results that they create, they have the regular folks, the people who had been already having the business knowledge and experience, do the interpretation of that and that’s what brings out the real predictive analytics, which is usable.

Eric Siegel. [00:20:36]:
That’s a great question. And the way in which an organization will go about with their first implementation. So we’re assuming here this is sort of first endeavors. The models vary greatly and it really depends on the organization and the predictive, the particular application, whether it’s marketing or other kind of operations like in the case of UPS and you know, the scale is going to be of numerous endeavors across the organization. So to take it to one extreme, if you’re doing a direct mail campaign and that’s a central part of at least a business unit and it’s periodic, improving the targeting with a predictive model does not require a few PHDs. It probably does not even require one PHD. It probably requires a staff member who has experience with predictive analytics initiatives

Eric Siegel. [00:21:32]:
and then the core modeling component, the actual analytics where you’re taking the data and learning from it using predictive analytics tools, that’s actually a very small part of the hands on hours of the overall project. When you consider for example deployment and integration, that’s often something that’s outsourced. So that’s taking it to one extreme. On the other extreme, if you get a large enterprise and lot, you do want to be building a team and there’s a great demand, it’s good to start early in terms of that search for the right employees.

Sanjog Aul [00:22:11]:
So Jack, when you look at the way you, you did mention that you have a number of PHDs and folks with maths. What is the mix overall of that team which eventually churns out the right type of analytics which is usable.

Jack Levis [00:22:25]:
Besides those people, I fully agree with Eric on things like marketing plans, send marketing campaigns to. For us, we need to make our analytics truly actionable inside of a process. So on an analytics team generally I’ve got a mix of skill sets. Usually there’s the analyst. More often than not, a PHD on the team itself is also a software engineer who can take the math and build a prototype or make it work faster and more robust. But I also put business people on the team. So an industrial engineer who can help engineer the process as well as a skilled frontline person who can explain the business process. So the team is really cross functional from that perspective trying to make sure that the analytics not only makes sense mathematically, but also makes sense operationally.

Sanjog Aul [00:23:26]:
If you were to take the people side, and if that is fully cooked, that means you’ve put in the right type of folks. Do you think their learning has to change continually in order for you to grow progressively in this direction? So, Jack, if you had to build a team and you put some people in place, do you think this whole area is morphing so that they have to learn beyond what they already knew for them to be effective on a regular basis?

Jack Levis [00:23:53]:
From the success of my group, absolutely. I could tell you that if you sat in a meeting with my folks, I think you’d have a hard time initially figuring out who’s the PHD in the team, who’s the business person, because they all start talking the same language. It’s almost like being a test pilot. They’re listening for all the little problems that are going on and each one of them will increase their skill set. So the business people person starts talking about operations research and the operations researcher starts talking about business problems. So they truly become a business team, just with complementary skills to get the analytics built.

Sanjog Aul [00:24:36]:
Eric, have you seen people having specific needs of the type of education that needs to be imparted or the training that they have to go through, and do you think that training itself is morphing, that what you are learning about in order to be effective with predictive analytics, that that learning has to morph to something new and different or the core fundamentals are going to be the same all along?

Eric Siegel. [00:25:00]:
Both. I mean, the core fundamentals are pretty,static, but there’s so much else that where there’s continuous learning around best practices in the industry and to some degree in terms of the core technology and what the existing tools are that, you know, software tools that embody that core methodology. But you know, when you’re coming in new and you’re looking at this whole new area, you know, I think the focus should be on that initial ramp up. In the longer run, once you’re in it, it’s not, I wouldn’t consider it burdensome or overwhelming. The fact that there’s continuous education, I see that as something that emerges organically. There’s a natural inclination towards it because it’s always directly tied to that value oriented actionability and how you’re deploying and how you’re pulling the data together in the first place to make the best model possible. So it all has to do with the value and there’s always an excitement around it. I wouldn’t say that it’s financially burdensome.

Eric Siegel. [00:26:02]:
It’s sort of a natural part of any career path but certainly there’s a continuous evolution as far as the initial education that’s required. That’s actually quite, that varies a lot. People come into this field of predictive analytics from all kinds of disciplines, oftentimes quantitative as far as the hands on practitioners themselves, sometimes PHD, but there’s a whole bunch of different routes. Anyone out there, if you’d like to email me, I have an informal long email I can reply to you with a long list of all the articles I’ve seen about what kind of education you need to become a practitioner in predictive analytics.

Jack Levis [00:26:44]:
So I think Sanjog, there’s another part of the education, at least for us, especially as we start to deploy prescriptive models, you know, optimizations and that’s the front line because those now become the decision makers and we found that we’ve had to educate that front line as well and the education is over and over and over again. You know, a model may have a lot of dials that need to be turned and for us, deploying to 55,000 drivers means really one model that can handle 55,000 different situations and we have to educate the frontline planner in how to twist the dials to make a result of an optimization, for instance, actually implementable. You know, it’s one thing to have a result that’s feasible, meaning it works mathematically, it’s different to have a result that is implementable. And that’s where some of that tweaking of the dials and the education of the front line is important.

Sanjog Aul [00:27:44]:
Jack, what would you say about the persuasiveness of any analytics related findings? If it’s not persuasive but it is factual, do you think we are still in that mode where anything that we find out it has to be done by the top brass and if they give the blessing only then we will get the dollars and then we will be able to implement. Are we still in that kind of constant blessing seeking mode or do you think that’s a non issue?

Jack Levis [00:28:16]:
Well, you know, businesses still need to have prioritization and for us we have a very well defined project prioritization and choosing process. So we can’t just go and as you said, get a million dollars and just run C level executives. You’re not selling to them like you would sell to a consumer. Analytics is not a pet rock. So you come in and you say here’s an analytics project, here are some benefits that we think we’re going to get the next six months we will evaluate it and come back to you with those results. I found them to be extremely open to this concept and not just at ups. I talk to lots of companies and everybody is trying to get gains from their data. Everybody is trying to turn their data into information and then knowledge.

Jack Levis [00:29:13]:
But they are all struggling with how do I know it is going to work. I think that you have got a risk if you promise internally that analytics is going to work and produce gains and two years later you have nothing. The C level suite will say, well the analytics didn’t work so you really have to shepherd it through a process. You have to ensure that gains are achieved and then they love it once they see some. The next question is where else can we apply analytics to our business? It turns out to be a very positive experience when you’ve gotten some ROI behind it.

Sanjog Aul [00:29:50]:
Let’s take a quick break listeners. We’ll be right back and then look at further on what is the way an organization needs to make a business case. So yes, the top brass doesn’t look at it like this and perhaps they have the faith in the leaders who have been given the responsibility to run predictive analytics. but at the same time there has to be some accountability and some visibility into what’s going on. So as you go about this rather fuzzy journey once in a while where we do not know what’s going to come out but we still have to keep investing, how do you keep everyone in the loop? How do you make sure that the journey that we are going on, everybody is on board and there are no unnecessary stumbling blocks. Please stay tuned listeners. We’ll be right back and explore.

Speaker A [00:30:38]:
HP is proud to sponsor this program. Tap into our expertise, innovation and services to bring your most workloads to the cloud.

Speaker E [00:30:50]:
Bosch Software Innovations is proud to sponsor this program. Visit www.bosch-si.com Connected Manufacturing to find out how Bosch can help you improve your operational performance and become a manufacturing industry leader in a connected world. Change the way you predict, manage and produce outcomes. Bosch Connected Manufacturing.

Speaker A [00:31:22]:
HP is proud to sponsor this program. Find out how the HP as a Service solution for SAP HANA can help you gain instant impactful business results without capital investment by logging on to hp.com transform information into intelligence and a competitive advantage with a full spectrum of SAP HANA products and services from hp, a global SAP hosting partner. You are listening to CIO Talk Radio with Son Jogue All. To learn more about the show, please visit www.ciotalkradio.com if you have a question or comment, call toll free at 1-866-472-5790. That number again is 1-866-472-5710. Now back to the show. Here’s Sonjog All.

Sanjog Aul [00:32:25]:
Welcome back. So, Eric, when you look at the different organizations, the way the analytics leaders, the ones who have been given the responsibility to run predictive analytics and make it valuable, what is the communication approach between them, the business unit leaders, the executive management and the people below so that everybody’s in the know, everybody’s even learning about what’s working, what’s not working so that it is basically we learn from the collective intelligence and collective feedback.

Eric Siegel. [00:32:57]:
Yeah, that’s a great question. You know, there does need to be a lot of communication. It needs to be quite iterative. There’s an industry standard best practice model that includes that part and parcel to it, that it’s really iterative meetings across different business functions and in that, you know, the biggest component of explaining it and keeping everybody’s eye on the value is not with regard, is not with regard to that core analytical technology, but rather is with regard to the business case with regard to the message that made this persuadable in the first place, where the value comes and it’s not rocket science to understand that. It’s simply a matter of what exactly decision are we improving with predictions? So for example, should I send a brochure to this person or not? Should I approve this application for credit? And then exactly what’s being predicted? Are they going to make a purchase? Whatever the prediction is that’s driving that decision. So you put those two pieces together and the rest of the details fall out from that with regard to what data do we need to analyze to make that predictive model and which operations need to integrate the predictive scores.

Sanjog Aul [00:34:20]:
So one is the owner so Jack, this is a question for you. One is the owner of this predictive analytics project or initiative and then there are sponsors. So how do you make sure that the owners are identified so that they have enough, not just the authority, but also the empowerment that they are able to go through this and then the sponsors, they are also basically while then they are sponsoring, they also are able to somehow tie it back to the value that they are creating because at the same, at all times, everyone should be accountable and answerable to the people who they serve in this predictive analysis. Because this is a lot of fuzziness, right? And what will happen? How will we go about etc.

Jack Levis [00:35:10]:
Sure and you know, there’s research involved often here. I’m fortunate that for me, I’m the, I have the analytics teams that report to me, but I’m also the owner and the sponsor so when I take a project, I’m end to end again. My job is process management. However, there’s often projects we’ll do where maybe there’s another sponsor. I’m working for another group and I found that one of the important areas is education of the sponsors. I think it’s important that the sponsors know what analytics can do, but it’s also important to know what it can’t do if they think there’s going to be magic coming out of something that’s not going to be helpful as they go thinking what can be produced.

Jack Levis [00:35:53]:
I really believe that forums like Predictive Analytics World isn’t just for the geeks, it’s for those business owners. They can come in, take a look and find out what other people are doing. They can find out what analytics has done and I think that helps them in their own job, their own education. So we spend time educating our people. Whether it’s Predictive Analytics World or the Institute for Operations Research and Management Science Informs, we spend a lot of time there and I think it’s helpful for those business people to understand the environment of analytics so that they can point out places where they can actually use it to get gains in their own operations.

Sanjog Aul [00:36:36]:
Eric, do you think when you do start or at least even work towards engaging in this predictive analytics, there are some pitfalls starting from who becomes the sponsor and who becomes the owner?

Eric Siegel. [00:36:53]:
I think the biggest pitfalls is when there may be a lack of clarity on what that value is, exactly how this model is going to be integrated and deployed. So that’s really the leading pitfall that potentially comes up if there hasn’t really been agreement and buy in full understanding across the right people, including internal clients, that hey, not only are we going to make this model that learns from data, but once we have it, the predictions are now going to actually directly affect these current operations. So that is these operations, the way you are doing it today, that is going to change and there has to be buy in from that, that you don’t want that to stall out later when you are actually starting to go to the deployment phase.

Jack Levis [00:37:44]:
See, I couldn’t agree more. I think that change management analytics is about making better decisions, which means you want to change things you’re doing today. Descriptive analytics honestly is relatively straightforward because once you have the insight, the descriptive analytics can help you without changing a process to just make some better decisions looking forward. But it’s limited because it’s always looking at yesterday. When you get to prediction and prescription, you’re looking at tomorrow and you’re truly making change and if you don’t understand that change management is part of the process, you’re asking people to change how they do things. You have to build that in. That requires support from the top, that requires education, that requires new metrics

Jack Levis [00:38:27]:
by the way, we deploy new metrics every time we do an analytics tool that are more leading indicator metrics so that you get the change happening in the organization and to me, that’s fun when you hear conversations change, when people start talking about different things than they did yesterday because now they have the insight, you know, that you’re starting to gain momentum and that analytics is not a flavor of the month.

Sanjog Aul [00:38:53]:
Jack, have you ever been challenged by the people around you in the way you are approaching it because you don’t have the results, you don’t have a way to do a rebuttal, but you just request them? Hey, could you hold your patience and have faith in me? Is that what’s the best you could do at that time, or you have a way to explain that they don’t create that challenge for you?

Jack Levis [00:39:16]:
Well, that’s a tough question, and it’s especially tough because we’ve had such successes. But there is a period of time when you’re going through analytics and you’re doing discovery that sponsors or C level folks may become impatient and it’s at that point in time, I’m careful not to promise a date. If you were doing medical research, you couldn’t say the research will be done on a date. So I’m very careful to make sure that senior leaders understand. We made presentations to our C suite to explain how analytics works so they could understand there’s a point in time of discovery, then there’s a point of time of implementation. We build prototypes and we’ll use the prototypes to help make sure that we understand how to get gains, that we understand what the new metrics should be, we understand how to educate people and it’s in that period of time that you truly find out whether you have something that’s implementable, where true gains can be gotten or not.

Jack Levis [00:40:16]:
I can tell you that we’ve been doing analytics so long, it’s really hard to separate it from our business processes. You know, we have a new analytics project that we’re deploying right now that has 700 full time resources and that shows the support that our C suite has for the analytics inside of ups.

Sanjog Aul [00:40:41]:
Eric, if you were to look across the industry, the spectrum, do you think you find either skepticism, which drives better results, or wholehearted sponsorship, which drives better results? What? What? Top level sponsorship and faith has given better results? Because sometimes when somebody’s checked, people don’t do what you expect, people do what you inspect. Right? I mean, that’s what a conventional saying is. So would a healthy skeptic. Top leadership and business unit leaders would bring better results out of these initiatives.

Eric Siegel. [00:41:19]:
That’s a great way of framing the question, sponsorship versus skepticism. You know, I think there should be skepticism disguised as sponsorship. That is, on the surface, it has to err on the side of sponsorship when we’re talking about deploying something new. It also turns out that just culturally, and I feel quite confident about this within the field of predictive analytics, the enthusiasm about using data in this way, about the core technology and how it actually then ends up being so valuable and where the rubber hits the road in frontline deployment, the enthusiasm is so high that you don’t need to light a fire in these new initiatives and with negative skepticism, there does have to be sponsorship. You do have to get things going if it’s your first time. By the way, I’d also like to make a comment following up on what we were talking about recently, about what needs to be communicated and shared and mutually understand along the lines of this stuff not being magic. So how do you resolve the idea that it’s not magic with the idea that it’s predictive because prediction sounds like something magical? The answer is that you don’t actually need to predict accurately.

Eric Siegel. [00:42:35]:
In most cases, it’s about predicting better than guessing and tipping the odds in the mass scale numbers game. So that’s sort of the real concrete way. So for example, if I find people that are three times more likely than average to make a purchase if marketed to, there still may be a low percent that are actually responsive. So it’s not about accurate prediction. But again, we’ve tipped the odds and that has a huge impact on the bottom line.

Sanjog Aul [00:43:03]:
Jack, when you look at your organization, and of course you must be talking to your peers in other companies, do you see a thread, a common thread across the board which has led them to relatively more predictable success versus turbulence?

Jack Levis [00:43:24]:
Well, I think culture is probably the top thing that will lead to success versus the turbulence. You know, we have a culture of advanced analytics at UPS. I mean, going back to the 1940s of using quantitative analysis. I think that’s the top thing is getting the culture there. Usually when I talk to an organization and they ask for advice as to how to get started and these range in all sorts of different disciplines from government to small companies, that’s usually the first question is where are you on that culture scale? Where are you on support from the top and the buy in? Because without the buy in, the analytics are just going to sit on the shelf. So I really find that to be the number one area for truly getting the buy in and going, as you say, from turbulence to results.

Sanjog Aul [00:44:18]:
Let’s take a quick break listeners. When we come back, Eric, let’s talk about the percentage of blue sky type of predictive analytics efforts that we should have in our portfolio as an organization versus something which is more strict, measurable and goal oriented, which will eventually help us get through the initial skepticism and very well on our way to ingraining the predictive analytics into our organization’s DNA. Please stay tuned listeners. We’ll be right back.

Speaker E [00:44:52]:
Bosch Software Innovations is proud to sponsor this program. Visit www.bosch to find out how Bosch can help you improve your operational performance and become a manufacturing industry leader in a connected world. Change the way you predict, manage and produce outcomes. Bosch Connected Manufacturing.

Speaker A [00:45:24]:
HP is proud to sponsor this program. Tap into our expertise, innovation and services to bring your most important workloads to the cloud. HP is proud to sponsor this program. Find out how the HP as a Service solution for SAP HANA can help you gain instant, impactful business results without capital investment. By logging on to hp.com transform information into intelligence and a competitive advantage with full spectrum of SAP HANA products and services from hp, a global SAP hosting partner. You are listening to CIO Talk Radio with Sonjog All. To learn more about the show, please visit www.ciotalkradio.com. if you have a question or comment, call toll free at 1-866-472-5790.

Speaker A [00:46:28]:
That number again is 1-866-472-5710. Now back to the show. Here’s Sonjog all.

Sanjog Aul [00:46:40]:
Welcome back. So Eric, how about looking at a mix and I’m not sure if that is 100% accurate science, but a good split between the blue sky type of predictive analytics initiatives versus very strict goal oriented. What mix have you seen people deploying for them to eventually get to a stable state where predictive analytics actually starts becoming effective?

Eric Siegel. [00:47:03]:
That’s a great question. I would say that all predictive analytics initiatives need to be defined in terms of being actionable in terms of their results eventually being measurable. So blue sky, we could think of that in terms of what’s an absolutely new thing no other organization has ever done. So I mentioned earlier those application areas that are the more well trodden common areas for businesses. Marketing, credit risk and fraud detection in some cases web optimization, certain things based on prediction and that’s oftentimes really the best place to start. But it depends on the organization. So you can imagine how widely applicable and broad this area is.

Eric Siegel. [00:47:50]:
There’s so many ways to use predictive analytics, that is to say there’s so many things that could be predicted about consumer behavior and so many operations that could be improved with those predictions. So when you stick within sort of the more standard realm, you know, you see more track record at other enterprises in the industry. You see the process more well defined because it’s been done so much so in a sense there’s less risk there but depending on your organization there may be other areas I think UPS is a perfect example of course, where there’s all kinds of operations that are unique to that organization and so there’s unique application areas and unique opportunities for predictive analytics. So typically maybe it’s only 10 or 20% where it’s sort of that innovative new ways of using it, but for some organizations like UPS, it’s probably a lot different than that.

Sanjog Aul [00:48:47]:
Jack, you have any flavor of the percentage split?

Jack Levis [00:48:52]:
I really don’t because we keep a mix of long term projects that are truly game changers, medium term and short term and we try to keep those in the hopper all the time and as one gets done, hopefully we look for the next big game changer. I really can’t tell you the split. I can tell you that to me, research has shown that really a small number of organizations are really doing the predictive analytics or the prescriptive analytics. Even less for prescriptive. So I think those organizations just need to get started, get some wins under your belt. I think Eric, hit it on the nose that as you get your first win and the next one starts coming and the next one it becomes repeatable and then you’ll hit your own level.

Jack Levis [00:49:41]:
There’s different maturity of organizations in the use of analytics. I think organizations should assess themselves Informs the Institute of Operations Research and Management Science has an analytics maturity model. I think it would help organizations to take a little survey, find out where they sit and see if they get some ideas on how to move through that ladder of descriptive predictive prescriptive analytics.

Sanjog Aul [00:50:06]:
Eric, when you look at the type of education that’s been imparted formally maybe within an organization to the people who are doing this job. Do you think that internal training is good enough or would you say even the leaders themselves who want to lead this effort to get best success? What avenues are available for them to be able to get educated and get up to speed and stay up to speed with respect to how predictive analytics is changing?

Eric Siegel. [00:50:35]:
Well, Predictive Analytics World, the conference series that I founded, we’ve got 10 events a year. It’s very much formed to help that incremental education continue and also for newcomers. You know, as Jack mentioned earlier, events like that are not just for geeks and we actually have in the conference program to track one for all audiences, managers, decision makers, newcomers, and another track for the expert practitioner and these events also have before and after a variety of single day training workshops. More broadly though, if you’re coming in brand new to the field and you want to get a more holistic education, there’s a huge number of new certificate and master’s degree programs in predictive analytics cropping up across industry, all kinds of universities. Some of them serve as sponsors. You can see them on the Predictive Analytics World website.

Eric Siegel. [00:51:35]:
They’re sponsors for our events. You can find a complete list of these certificate programs on the KD Nuggets Data mining industry portal.

Sanjog Aul [00:51:49]:
So Jack, have you seen, when you have your team and your business leaders, where would you recommend them going? Of course you’d send them to EC’s conference series, but besides that, would you say that in order to be proactive in terms of developing the skill set of all people from top to bottom, what would you invest in?

Jack Levis [00:52:07]:
So again, I invest in Predictive Analytics World. I say it one more time, it’s not just for the heavy analysts, but also as I mentioned, the Institute for Operations Research and Management Science. Join an association like informs. Talk to people that are like minded, learn what analytics can do and what it can’t do. If you’re looking for a team, you know, INFORMS has a certification. Go test your analytics folks, see if they’re really just good at writing queries or if they really know how to do the predictive work. So I can’t say enough about sending your folks to conferences, sending them to network, getting them educated. I think that changes the culture inside your organization

Jack Levis [00:52:51]:
and as I said, I thought that was the number one thing. I know that Eric’s got a conference coming up, you know, it’s not far away and I think it’s going to be very worthwhile

Sanjog Aul [00:53:00]:
Coming to leadership. Jack, when you look at everything else, I think success or failure to quite an extent falls on the shoulder of the leaders, whether they are the people who are driving predictive analytics or the ones who are supporting and sponsoring what kind of leadership have has to be demonstrated by an organization’s leaders in whichever capacity they are in order for this predictive analytics related endeavors to be effective.

Jack Levis [00:53:31]:
So I think you have to understand that analytics is a journey and that’s going to require support from the top. It’s going to require guidance. You’re going to need to make analytics part of your culture. Everything analytics, the processes, the procedures, the methods all become the same thing. You bring on the right people and as we said, send them to the right conferences. You need to relook at your metrics. If your metrics are all the same old lagging indicators, then you’re not going to make a change. We’ve changed metrics to be more leading indicators and those leading indicators will then change conversations and make people be more analytical in their thinking and most important, understand that change management is needed and understand there’s going to be chaos along the way.

Jack Levis [00:54:24]:
I’ve got two quotes on my wall and especially as we’re working on some of these big analytics projects. One talks about chaos. It says that’s where great dreams are born and before the beginning of great brilliance there must be chaos. Before a brilliant person begins something great, they must look foolish to the crowd. Which goes into the second quote, which is about three stages of truth or I call it three stages of projects. The first is ridicule, the second is violent opposition and the third is acceptance as if self evident and leaders need to allow their people to go through those stages to get to acceptance as if self evident.

Sanjog Aul [00:55:05]:
Eric, any final words of advice where based on all the different leaders that you see and their respective organizations, the way they’re showing success, would you appeal to the other leaders who are getting started in the journey or maybe in the midst of it to do something new, different or better in order for this predictive analytics to be successful and effective?

Eric Siegel. [00:55:26]:
You know, I thought that what Jack, the way Jack put all that was remarkably well put. You know, within finite scope application area project such as targeting marketing. At this point it’s becoming more and more clockwork over across enterprises the industry is starting to really understand how these things work but when you start looking at a unique enterprise like UPS and or for any large organization, sort of all the different cross intra organizational deployments, that’s where there’s a tremendous amount of creativity needed there needs to be a real meeting of the minds between experts in predictive analytics and groundbreaking deployments of it, as well as the on site C level and VP level that really understands the organization as a whole. And these two minds need to come together and that takes a fair amount of exploration but when you get that union, that synergy between these two areas, that’s where the enterprise really evolves.

Jack Levis [00:56:40]:
Everybody talks about going from data to information to knowledge, but I think that’s short sighted. I think after knowledge is wisdom and that’s when you’re using models to tell you how to operate and that makes a new person as wise as someone who’s been there a long time but even after that, I think there’s clairvoyance. I think when models and data is smart enough, the transaction by transaction we’re predicting that a problem is going to happen and solve it before somebody knows something’s wrong will look clairvoyant and that’s truly when we’ll start seeing huge, gains beyond what we can envision today.

Sanjog Aul [00:57:17]:
On behalf of the show and our listeners, I’d really like to thank you both Eric and Jack, for sharing your thoughts on how to make an organization and get the most value out of predictive analytics and making predictive analytics effective.

Eric Siegel. [00:57:30]:
Thanks Sanjog.

Jack Levis [00:57:32]:
Thank you Sanjog.

Sanjog Aul [00:57:34]:
Thank you so much again and listeners. Please like us on Facebook search for CIO Talk Radio and be sure to follow us on Twitter. Thank you so much again for listening to CIO Talk Radio. This is Sanjog Aul your talk show host till next week. Take care and God bless.

Speaker A [00:57:49]:
Thank you for tuning in to CIO Talk Radio. To learn more about about the show, please visit www.ciotalkradio.com. please join Sun Joke all next Wednesday at 7am Pacific Time, 9am Central Time and 10am Eastern Time for another hour of CIO Talk Radio on the Voice America Business Channel. CIO Talk Radio is brought to you by HP and Bosch Software Innovations.

Download Podcast

Apple Podcast, Google Podcast, Spotify, Pandora, iHeartRadio, SoundCloud, TuneIn, and Stitcher. Find other syndication channels here or search CIO Talk Network podcast on any other app.

Explore More

• • Can Predictive Analytics Boost Graduation Rates?
  • Pros and Cons: Predictive Coding – Should I Invest?
  • Complex Case Law eDiscovery: Is Predictive Coding a Panacea?
  • How to be an Effective Analytics Leader?
  • Creating a Winning Data and Analytics Strategy
  • Data Analytics as a Team Sport
  • Building High-Performing Data and Analytics Team
  • Have Enterprise Data and Analytics Programs Delivered on the Promise?