Higher education is in the business of bringing students to their school and ensuring they get a degree. But graduation rates are suffering across the country, with one in three students never finishing college and over half of students finishing college in six years rather than the traditional four. Predictive Analytics could be applied to identify who is likely to fail and help them on their journey before they begin drifting. But just how are the insights generated being applied to help students succeed? And does Predictive Analytics stand to be the next big wave on campus or is taking on this weight just a part of the Freshman 15?

Contributors

• • Ellen Wagner, Co-Founder, Chief Research and Strategy Officer, PAR (Predictive Analytics Reporting) Framework,
  • Ashutosh Nandeshwar, Director, Advancement Information Systems & Analytics, California Institute of Technology

Transcript  (Can Predictive Analytics Boost Graduation Rates?)

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 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]:
Good morning 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 ctralive and hashtag Higher Ed. Today’s topic is Can Predictive Analytics Boost Graduation Rates? And our guests for today’s show are Ashutosh Nandeshwar, who’s the director of Advancement Information Systems and Analytics with the California Institute of Technology and he’s also the author of book Tableau Data Visualization Cookbook. Good morning, Ashutosh. How are you?

Ashutosh Nandeshwar, [00:01:02]:
Just great, Sanjog. Thank you very much for having me on the show.

Sanjog Aul [00:01:05]:
Oh, the pleasure is all ours. We also have David Crane, who’s the Assistant Provost and CIO with Southern Illinois University. How are you doing, David?

David Crane [00:01:13]:
Good morning. I’m doing great.

Sanjog Aul [00:01:15]:
Great, great to have you. And finally we have Ellen Wagner, who is the Co Founder and Chief Research and Strategy Officer with the PAR Framework or the Predictive Analytics Reporting Framework. Good morning, Ellen. How are you doing?

Ellen Wagner [00:01:27]:
I’m doing very well. Good morning, everyone. It’s great to be here.

Sanjog Aul [00:01:31]:
Great to have you as well. So on orientation days, as we know, in colleges, speakers used to ask incoming freshmen to look to their left and right and recognize that there may be not everyone left by the end of the four years and with the tradition of this four year plan becoming less of the norm, schools are looking to increase these odds, either by admitting students more likely to succeed or help those who are struggling. So on a previous show, we had discussed how big data can help in all aspects of higher education. But today we wanted to focus on this one goal and the use case with predictive analytics. So, Ashutosh, the first question I’d like to pose for you in terms of the challenges that colleges are facing in terms of improving graduation rates, what are some of the obvious ones and some others which we may not recognize as the challenges?

Ashutosh Nandeshwar, [00:02:25]:
I think the first thing that we know that for past 50 to 60 years, all the research that has been done, the graduation rates and the retention rates, the first year freshman retention rates have stayed the same so on national average it’s I think about 70% so only 70% of students who get admitted in the first year actually stay in the second year also and the graduation rates are also very low. So only 50% of the students who actually start their college within six years, under 50% of them will actually complete their degree so that’s a big, challenge snd a lot of research has been done in this area for last 50 to 60 years, various theoretical frameworks, various practical applications, but very few institutions have been actually successful in making any kind of a dent in this problem. As you said, some colleges are trying to maybe recruit and admit better prepared, better prepared students.

Ashutosh Nandeshwar, [00:03:24]:
But still there are some problems that are not only academic preparedness, but they’re also the persistence that’s required to complete the college. So not only facing the challenges and leaving them at it, but actually facing the challenges and then be able to overcome with that. And that is something that colleges try to do with their first year counseling and other and I think David and Ellen would be able to speak better about those because I don’t have any experience in that administration but what I’ve seen from my experience that college preparedness is only one factor. If that were the only factor, we would have seen a lot of students graduate but I think there are a lot of other factors that are, that come into play, which again, I’m thinking that David and Ellen might be able to answer better.

Sanjog Aul [00:04:09]:
So David, would you think that if the first year student doesn’t go on to the next one, could that be because there is a disconnect between what expectations were set for them by the parents and other people before they went to college and then it was a rude awakening and they said this is not what I wanted, or it is just, you know, the flakiness, if you will, of that age at which the kids actually go to college. So where are we missing the mark?

David Crane [00:04:39]:
Well, I think there’s a little bit of both. We at Southern Illinois University are an access university. Actually we have kind of a unique role in that we’re a top hundred research university but also we have almost 50% first generation college students every year in our freshman class. And this year we’re up to about 40% minorities in our freshman class and so we absolutely, with so many first generation college students, have a lot of kids come to college that don’t have adequate expectations and thus our retention rate has been far below the national average

David Crane [00:05:20]:
and within the last year or two, we’ve put a lot of focus on it and frankly are going to see 6 to 7% increase from freshman to sophomore retention in just one year with the numbers that are just coming out this fall. So we’ve put a lot of focus on how do you increase student success and thus increase retention? And so it is a lot of focus. I think as we’re talking about today, I think predictive analytics and big data can play a big role. But I think one of the biggest problems has just been the economy in general. The fact that so many of our students have to work so many hours on jobs outside, you know, the university as well as, you know, we have ever shrinking number of faculty and ever shrinking qualifications of faculty because we’re forced to use more non tenure track, fewer professor level and so, you know, the economy has been a big challenge for our university and we’re trying to use technology to offset some of those challenges.

Sanjog Aul [00:06:21]:
Ellen, what do you see happening?

Ellen Wagner [00:06:24]:
Well, as I’ve been listening to both of my co presenters speaking this morning, I’m struck by in fact even your question about the first time full time freshman. We’ve been looking at, oh, probably about 35 institutions now that have been working within this collaborative framework effort with which I’ve been involved. Probably 85% of the students within our 2 million student data set don’t look a whole lot like a first time full time freshman and I think this also underscores another part of this big conversation. College students don’t look a whole lot like the 18 to 22 year olds that we all remember. They are certainly a big part of who’s coming to college but college students look a whole lot like all of us right now. We are of all ages,

Ellen Wagner [00:07:19]:
we come in full time, we come in part time, we’re working and I think to David’s point, the economy might have something to do with what’s going on but I think what we’ve been seeing in our research is that there are certainly going to be cognitive or preparation issues, but one of the biggest issues for people trying to graduate on time or within a particular framework is that life interferes with the completion of college. So any one of us that’s looking at trying to keep people in school longer or get them to the point of a high value credential or a degree needs to be looking at a much bigger slice of variables that are likely to push someone off their path so that we can look not so much at one size fits all interventions that wrap around all students, but using these predictive data to really get the nuance that we can find for each individual student and then start looking at the things that for that student are going to help them proceed to complete their particular pathways to success.

Sanjog Aul [00:08:27]:
Now David, would you think that if based on Ellen’s response? People are coming from all walks of life, so our traditional assumptions of somebody’s expectations not met or they are people who are just starting out and they may be not making as much in terms of being able to support their education and that’s why they’re walking away because this kind of changes the landscape and the traditional mindset of a young first time college student coming, their expectations not met and they are aggressively thinking, I’m going to change the world. So after first year they say school is not going to help me change the world. I’m going to go out and start a startup or things of that nature. So are we living in that world truly or is this a world that we don’t understand and that’s why we’ve not been able to solve this problem.

David Crane [00:09:14]:
I think it’s a mix. I mean, I think every year we have a higher percent of non traditional students. We have higher percent of distance education via either, you know, remote campuses or military bases or online learning and I think that’s going to continue to grow. But at the same time, we still as a university are predominantly, you know, first time freshmen and transfers. We get a lot of transfers from junior colleges in Illinois. We have a very robust junior college system and so really, you know, our main population is still that 18 to 22 year old, but the other set of students is continuously growing every year.

David Crane [00:09:58]:
And so really, you know, there are almost two different paradigms and we have to plan and make sure that students in each are very successful.

Ellen Wagner [00:10:06]:
Yeah, I think that’s really the trick here because the idea of the single vision of a college student, I mean, college students look really different, have lots of different demands and so for any one of us looking at use of predictives to start looking at what we can do for each individual student, remember that when we’re talking about some of these predictive analytics, we’re not just talking about something that sweeps across all of us and generalizes. We’re able to take very specific records on each individual that contributes information and really dive down to that individual student. So in addition to David’s point that we’ve got to be mindful of the types of students coming in. And I think this anticipates one of the other questions we may be looking at here in a little bit, we’ve also got to make sure that we are using our data to answer the question that’s in front of us and to really look at what problems we are trying to solve with the data rather than having the Data there where we sort of look for patterns and fit ourselves into what the patterns are telling us just right off the top that we can see. I mean this is, I wish it were true that predictive analytics were going to be the magic eight ball that just simply provides us with all the fully formed answers and, and actually that’s, I think that’s going to be one of the things I hope we get into because predictive analytics as the magic pill is a somewhat incomplete perspective.

Sanjog Aul [00:11:33]:
So Ashutosh, when you listen to both David and Ellen, when you look at predictive analytics as a tool, and Ellen said it right, because we can very easily start taking a technology and say, man, this is going to change the world. It’s going to be the magic build. Typically, or usually that’s not the case. So if you look at the intent of utilizing a technology such as predictive analytics or this approach of solving it, is the intent to actually make sure that we work on the higher purpose, which is to get these graduations rate up or to be able to retain people by changing their buying behavior so that they buy more courses. Is this more altruistic or is this more business centric in your view?

Ashutosh Nandeshwar, [00:12:17]:
I think that’s a great question. I agree with David and Ellen that it’s not only the traditional first time freshman students. So here is a story that I want to tell you so when I started first time in institutional research, that’s how I started working in my first full time job. And this was the problem that I really wanted to study, that it first time student retention because many universities have this problem. I was working for a public, public state university and I started looking at the data and one experienced researcher who had been there for more than 30 years came and told me, he said, you know, yeah, you have all these tools, which is great, but at the end you have to look at the individual person. Every record in your data set is actually a person.

Ashutosh Nandeshwar, [00:13:02]:
And that struck a card to me I still remember that because it’s very easy to look at, get lost in the data and say, oh, I’m going to, and try to generalize that but at the end it’s still that one person that we have to care about and I think that’s why we are seeing the newer trends of not only creating patterns. I think we have to attack this problem from any angle possible because we do have to help these students graduate. It’s only fair that we help them. So one of the ways is that obviously you look at the first time freshman students and you try to make them pass every year so at least when they get to the third year the chances of succeeding are higher than if they were to leave in their first year. So one of the newest things that are happening is the learning management systems and using predictive analytics in learning management systems and looking at the course data and their online course behavior data, trying to not only predict but coach them.

Ashutosh Nandeshwar, [00:14:01]:
So based on, let’s say they missed an online course or a day logging in, then some sort of flags go out to advisors or faculty or the student himself or herself and based on other students history showing here are some things that the student could do better if he’s not getting the grades that he should be getting. Or based on his or her profile trying to predict that this is the grade that you are going to get but if you do these things, these other things, you might be able to move to a higher grade and Michigan University of Michigan has developed some of these technologies. I think one of them is called Ecoach and Purdue also had some of these signal systems and I think this is the new trend that we are seeing

Ashutosh Nandeshwar, [00:14:47]:
and this is a hard problem this is not an easy problem. Look at a lot of text data and trying to come up with some patterns and be able to provide some counseling and coaching because I think it’s again up to that individual person and it’s more about perseverance and persistence than the college preparedness so we have to help them in those in that manner also.

Sanjog Aul [00:15:11]:
Let’s take a quick break listeners. We’ll be right back. And Ellen, let’s explore this a little further based on what Ashutosh mentioned, that we are trying to utilize aggregate data across multiple students and then trying to find a pattern and then try to make decisions and or suggestions for an individual student for that person to be successful. Now that’s not new. When we start talking about predictive analytics applied to an individual based on what they do, that’s more laser focused because the traditional method may not have been successful or that’s not accurate because that’s not working. So what’s the tweak? What’s something new that we are attempting to do where this one person is a population of one and you study that person’s behavior, the environmental factors, the background and everything else and then perhaps allow creation of an environment for that individual so that they increase their chances of success and graduation.

Sanjog Aul [00:16:05]:
Please stay tuned listeners. We’ll be right back.

Speaker A [00:16:14]:
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 F [00:16:44]:
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 [00:17:16]:
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 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. That number again is 1-866-472-5710. Now back to the show. Here’s Sanjog Aul.

Sanjog Aul [00:18:01]:
Welcome back. So Ellen, I’d like to quickly pose the question again. Is we are a person or a population of 1? We have our own environmental factors that influence us and which will take us whether through the graduation or have us drop out if we try to go the traditional route of collecting a lot of different data, trying to find patterns and then apply them with the hope that this is going to be successful. I’m sure that’s what we’ve been doing all along and it has not worked. So what’s new now?

Ellen Wagner [00:18:28]:
What’s new now is that we do have opportunities to sample from lots and lots of different data sources. Ashutosh was talking about using learning management system data where you could actually look at point of instruction information and the other thing we’re able to start doing is to look at displays information that we’re collecting, to start looking then at the bigger numbers for doing predictions about where the patterns are going to move and from there not so much using the data. It’s less about using the data for reporting and it’s more about using the data for insight. So that if you know that you are looking at trying to personalize learning experiences as much as you can, but you’re also trying to maximize the efficiency on what the institution can do. I mean, this is really the nut of where I’m seeing the value for the predictives is that we’ve got to find a place where we can personalize the experiences so that they are relevant and meaningful for the maximum number of students while also making certain that the institutions providing these experiences can afford it because if you happen to be an institution that is an open enrollment institution where you take all students, you’re going to have a lot more heterogeneity in your student population, which makes it critical to respond to everyone

Ellen Wagner [00:19:53]:
but the capacities that you have for touching everyone will be compromised unless you can get inside and see the different patterns sooner. So I think what’s different now is that this isn’t so much about top down telling people what it should do, although I think we’re going to see that for a while. This is more about providing institutions with data so that they can do a better job and can be more responsive. So it’s almost like flipping the formula. We’re not so much just reporting where we’ve been in that reactive way as we are looking at using data for insight to be proactive about planning for the things that we have to do to be more relevant for more students going through our student pipeline.

Sanjog Aul [00:20:32]:
So David, I asked this question, is it altruism or is it business when it comes to helping a student graduate? Because if you’re going to get business like in terms of solving this problem, we’ll get anxious and we will also try to immediately move towards a more aggregate form of understanding of how students behave versus looking at one student at a time, which is a more expensive proposition. So what’s the mindset with which we are even approaching this problem?

David Crane [00:20:59]:
From my perspective, it’s very much linked. The student success leads to retention, which leads to graduation, which helps recruitment and so I really think that they’re linked. We’ve looked at studies that have shown over and over that it’s much cheaper to keep a student than it is to recruit them originally and so the number one factor of course for retention is student success and so if we can put students on the path to graduation and succeed in our programs, you know, we keep that student and you know, for every student that we keep, there’s basically a hundred thousand dollar impact to the university for keeping that student for a four year period and so, you know, it’s very important that we keep that student from a business perspective and by putting them on the path to graduation and then having them graduate, it empowers, you know, us to do so and also helps our recruitment.

Sanjog Aul [00:21:56]:
So Ashutosh, if you look at the data points which you would have liked to see collected, if you had a population of one, that student with the environment that they came from and was working towards making that individual be successful in terms of graduating, what would that be as compared to when you’re looking at an aggregate population? And where would we be missing the mark when we are looking at a population of 2000 versus population of 1?

Ashutosh Nandeshwar, [00:22:25]:
Before I answer that, I would like to comment to your previous question about altruism or business. Also, I was just reading about some articles last night about how President Obama’s new plan of ranking every institution and graduation rates are one of the bigger factors or characteristics of that ranking system. So I don’t see if that system is implemented. I don’t see how this question doesn’t become a business type of a question. But at the same time, I think it’s only fair that we help each and every student. As David said, we have to empower them to graduate and put them on a path of success. So I think if you get a chance to talk about the new rating system, I would like to do so with other guests than you. Sanjog, going back to your question about looking at the aggregate data and individual data, I think on aggregate level we all know what the biggest predictors are.

Ashutosh Nandeshwar, [00:23:16]:
It’s the college preparedness, high school gpa, their ACT scores, their SAT scores, their parents education levels, the financial aid factor, how much money they need, the EFC and how much money they already have and what financial aid is the institution offering. Those are some of the bigger factors but when we did the research we found that it was very hard to predict the first time retention. There were no attributes that stood out. Again, high school GPA explained 50% of the problem, but the 50% of the problem still remained. So we were not able to answer. So we moved on looking at the first time Freshman data and try to predict the second year retention. So we got slight improvement but we got biggest improvement when we looked at a third year retention, looking at the first year data but trying to predict whether they would stay after three years.

Ashutosh Nandeshwar, [00:24:06]:
That has had the biggest improvement in our research and the biggest factor in that was the education levels of parents. So that somehow is a signal for again that talking about perseverance and persistence is somehow the parents are able to explain the value of education and guide them. So here’s another example. When I was working for this big public state university, I used to see a lot of first generation students and there were a lot of facilities for first generation students. But my supervisor who was working there as a director, he also had his daughter in the same college and since the first day he was able to walk with her and show her oh, here’s where the first year counseling centers are, here is where you go and register for your courses, here is the registrar’s office, here is the student center.

Ashutosh Nandeshwar, [00:25:04]:
That made a huge difference because his daughter knew who to count on, who to ask questions to and he was right there to be able to provide all those answers and provide that guidance. So I think that is very critical and that is one of the biggest factors in student success but that’s on aggregate level. When you talk about individual level, again, I would like to go back to this ecoach example, I think that’s one of the game changing technologies. If we can focus on that individual and look at all the attributes about that individual, just like what Amazon and Netflix does it for us based on our behavioral data, not only our transactional data, but our behavioral data, how we are doing in our courses, what notes are we taking? Are we asking any questions during the class? If you’re missing something.

Sanjog Aul [00:25:52]:
So Ellen, if you were to look at like the way Ashutosh responded and he mentioned about parents background, etc, and then you look at individuals data points, does collecting that data point at an individual level becomes financially impossible for an organization to start doing things at that level? Because there are specific instances where while your aggregate data might show that this person is very likely to graduate, but then the individual factors any day take precedence over are more important and they have bigger influence over the aggregate the data that you might have collected with respect to that student.

Ellen Wagner [00:26:39]:
Well, the world in which I spend my time is the one that lives between the institution and governance. So I’m smiling as I’m hearing the conversation. It would be great to have that one to one opportunity for all students because that is in fact going to be one of the places that will have some direct and significant impact. The thing that I start seeing is that when institutions are going to be funded by a state legislature that the opportunity for that level of personalization is really not something that a legislature is going to look at. That’s a great ideal and there will certainly be opportunities and I think emerging new models of education that are going to push us more toward that. I think David mentioned online learning or distributed learning as a way to provide people with more pathways. But you know, when a state legislature looks at education reform or progress, they’re looking at really three things. They’re looking at progress, they’re looking at retention, and they’re looking at completion.

Ellen Wagner [00:27:45]:
So I think the challenge for many of us working with these data is that we’re not talking about a single framework. We’re talking about how do we leverage analytics to help people make tough decisions about resource allocation that helps us get as close to that personalized learning experience that we all know is something we’d love to have and we’d like to have for our kids. Back to what an institution says that they can actually afford to provide to touch graduation, retention, and completion. So this is where I think sometimes we have to make sure that the metrics by which we are being held accountable are in fact, the ones that we are measuring. So that we, I think we all mentioned at the beginning of the show that analytics in and of themselves are not new. What we’re looking at now is hope for application of analytics to answer questions that we’ve not been able to answer before and so the balance between the perfect experience for each individual student and what each of our institutions, whether they’re public or private or for profit or what have you, these are business conversations on top of pedagogical conversations and I think this is why we all get pretty confused.

Ellen Wagner [00:29:05]:
We’re comparing apples to zebras.

Sanjog Aul [00:29:08]:
David so now if what Ellen says is true, and of course that becomes a problem, then should we even attempt to solve this problem in its entirety, or we would just chip away till we find a magic pill?

David Crane [00:29:23]:
Well, the reality is that it is a very complex problem, except that with technology advances, technological advances, we’re solving some of those problems. We absolutely are transitioning to digital course materials. So not only for the textbooks, but for the online learning components in addition to learning management system. I mean, in other words, we have this big predictive analytics program in our math department, which is taking our core math courses, like the freshman level basic math courses, and using predictive analytics and we’re using a Pearson product not only for the electronic textbook, but for all of the online lab components. And so we’re taking all of that data and we can tell how much a student is studying. We can tell if they’re doing their homework, if they’re getting the homework right. We can tell if the, you know, there’s certain concepts that they continue to miss in the homework.

David Crane [00:30:23]:
We can tell how much time they’re spending, you know, overall in studying on the class we can tell if they’re attending class and we’re marrying that with, like, the preparedness data. So we’re doing placement tests before the math class even starts and so we can tell in four weeks, our predictive model is 85% accurate who’s going to select succeed in the course. And by seven weeks, it’s almost 95% accurate and then the key is to take that data and to have an accurate response to the data. In other words, there’s a certain group of students that are not going to succeed

David Crane [00:31:01]:
and so we’ve created parachute courses so they can transition to a lower level math class and still get credit for it and depending on their major, they may or may not have to go back and retake, you know, the higher level math class, but it gives them a chance to succeed and readies them to take the class again if they need to or desire to. But then other students can be saved and you can tell by the predictive analytics that if they go to tutoring sessions or they join group study, or they just spend more time online and so what we’re trying to do is we notify the instructors, we notify the advisors, and then we’re really big on the notion of a quantified self and so basically we want to provide feedback to students so they know when they’re in trouble and they know what they could be doing differently and so we are popping up alerts when they log on to any of our systems that show them if they have warnings, show them that they’re in trouble in individual classes and why and what they can do about it and we feel like, you know, those have all helped our grade distribution just since they’ve been implemented.

David Crane [00:32:13]:
You know, we’re seeing higher A’s, B’s and C’s and fewer D’s, F’s and withdraws and so we’re now actually expanding that program to other classes. Last spring, we started with our University College student success course, and now we’re looking at, you know, English courses and so we believe that we can use technology as kind of that medium to provide feedback to the student as well as to provide timely alerts to instructors and advisors. So I’m not saying it’s a magic pill, but I believe that the technology now exists that we can do customized education in a very affordable manner.

Ellen Wagner [00:32:52]:
What’s cool about David’s stuff here is that you’re talking about all the intervention. And this is something we alluded to a little bit earlier. I think all of us did, that the predictions are important, but it’s really just the first step that if you know who’s at risk, then you have to take the next step, which is doing something about it and I will say that for all of the progress we’ve been seeing in our predictive works, the model telling people who’s at risk is the first step. It doesn’t matter if you know who’s at risk if you can’t do something about mitigating the risk. So the focus on interventions is absolutely where I think we’re going to see the payoff when it comes to the predictive analytics, because without action, the predictions really are of relatively limited value.

Sanjog Aul [00:33:38]:
Let’s take a quick break. Listeners will be right back. And Ashutosh, I’d like to take what Ellen just mentioned about the quality of decisions being made based on the data and findings that becomes critical and what is the impediment in being able to take the right decisions, most objective decisions, which will eventually, number one, help graduation rates, but also not offer way too many parachutes, that is lower the standards at which a student of course expects the world to work with them when they come out in real life and also be able to work harder and smarter. That’s what the colleges are for, not to get them graduation, but to help them set the foundation in terms of what the real world is and prepare them for the real world. Let’s explore this when we come back. Please stay tuned.

Speaker A [00:34:30]:
HP is proud to sponsor this program. Tap into our expertise, innovation and services to bring your most important workloads to the cloud.

Speaker F [00:34:42]:
Bosch Software Innovations is proud to sponsor this program. Visit www.bosch-si.com Connected Manufacturer 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 [00:35:13]:
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 Jog 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.

Speaker A [00:36:14]:
Here’s Sanjog Aul.

Sanjog Aul [00:36:16]:
Welcome back. So Ashutosh, we all know that of course we can churn all the data we want, but the values in the actual interpretation and corresponding actionable items that we identify which we put into place. Now Ellen did mention about I think it was David who mentioned about parachutes and that’s basically lowering the standard for somebody to get graduation and in some cases you may do it. But if that becomes a norm, then we are lowering the bar as compared to what the real world is for the students just because we want to get them to graduate. Now that is not of course welcomed by someone who’s looking at it objectively. So where do you think the disconnects are? Or some improvement is required in terms of the interpretation and the decision making by the schools and colleges when they are looking at predictive analytics and trying to make the most of it?

Ashutosh Nandeshwar, [00:37:12]:
So I think it’s a very complicated problem. And first, I think I agree 100% with Ellen, that action must succeed analytics for your action, for your success is defined by that action, any analysis so to paraphrase someone, 100% of your analysis is wasted if your reader doesn’t take any action on it so it is very critical and that problem is not only the predictive analytics problem, it is a problem of explanation also. So we might be able to come up with really complicated, high performing, highly accurate problems, but if there is no audience for it, or if the person who is coming up with analysis is not able to explain it in a manner where it is not only descriptive, but it is actually prescriptive, then again analysis is failing and that person is actually responsible for that. So it’s not only so, it’s a multi part problem.

Ashutosh Nandeshwar, [00:38:03]:
First you need to have the analysis that is accurate and high performing, but it should be explainable also. But the person who is presenting should also be in the position to be able to explain it to other administrators who may not understand all the analysis and they possibly don’t even care. They just want to know what actions that they have to prescribe to other administrators based on data. So I think that’s one part of the problem. The second problem that you mentioned about parachuting, so I don’t know in what context David was saying, I don’t think he was saying that we want to lower our standards. I think if he possibly meant, I don’t want to speak for him, but I think the idea here again is for that individual, for that student, what is suitable for that student, what would make him tick, what would make him succeed and that’s the prescription we have to provide. It’s not about lowering the standards

Ashutosh Nandeshwar, [00:38:53]:
I don’t think anybody would approve of that but it is again, finding the success path, the path of most success for every student and providing that to them.

Ellen Wagner [00:39:04]:
That’s right. Pathways to success, that may not be the first one that they try, but if you you know, one of the things we saw with so many of our students when we were looking at, you know, where were the pinch points, where were the places in the system that people would fall out? And sometimes we could use our great big fancy schmancy analytics to do very refined analyses, but sometimes we could use the data that we had collected to actually look at something like, oh my gosh, that ad drop is going to push students to a point where they’re not going to be able to recover within the semester if they drop out of that course. What can we do to say, move an add drop date forward or to change the conditions under which one can add or drop a course so that if somebody discovers, oops, I made a mistake, because sometimes we do, then you give them an alternative to be successful and to not just make it binary. Yes, you’re going to succeed or no you don’t. So parachuting for me was, it sounds more like putting people on the right path rather than basically saying, oops, too bad, out of here.

Sanjog Aul [00:40:07]:
So in the working world, David, this is a question for you. We look at hiring the right person goes a long way in terms of ensuring that person’s success. Basically hiring the right person and putting them in the right role. If we had to now take this whole phenomena and try to bring it to the way an individual is invited into a school, not just that they have the money and they can pay for it, we get them in, but what kind of courses are they able to take? In what way are they oriented in and in what position? And there’s no specific position, but what set of courses? Again, perhaps it comes back to how do you induct a student? In what insights do you think predictive analytics could be providing us more than what was happening earlier? So that you are setting that person up for success and you have to do less work as damage control.

David Crane [00:41:00]:
So we actually have implemented a new advising system that uses predictive analytics. So based on their high school record and their ACT scores and all the pre college work that they’ve done, we can predict their success in various majors. So someone may be an engineering major or they want to be an engineering major, and yet they’re not good at math, we can show them statistically what their percentage chance of succeeding in the engineering major would be and yet we can also recommend, you know, you’re not good at math, but you’re really good at English and so these are maybe some majors that you can try instead and ultimately, you know, it’s really all up to the student, because we don’t want to tell someone that wants to be an engineer that they absolutely can’t be, because you can defy the odds and, you know, you can work hard and you can overcome some of your limitations. But I think that it’s very important in setting the expectations of what is going to be required to succeed in various majors and what their chance of success really is and then as we get more data on the student, what classes they take here at the university and what their success in those classes are, that predictive model gets even more predictive.

David Crane [00:42:15]:
It gets stronger and so, you know, we can tell them, look, you’re not doing well here, but you could transfer to this other similar major and do very well. And I think that that’s a big part of orienting them to the university and making sure that they succeed. But then another thing that we’re doing is, you know, we do have a student success course that is mandatory for all of our incoming freshmen to take that really, you know, talks about what it takes to succeed not only in college in general, but at college here at Southern Illinois University and I think that’s one of the reason our student success and retention numbers are going up, is because we do require that student success course and so, you know, I think a lot of it’s just about setting expectations of what it’s going to take for that student to succeed, which also gets back to, you know, customized learning and quantity, quantified self. We just want to keep the student informed and let them know what their path is and where they stand on that path at any given time.

Sanjog Aul [00:43:17]:
Ashutosh, I have a question here. If you were to most academic institutions don’t say this out loud, but student is actually their customer and traditionally we have all been taught not to say no to a customer. So if a student comes in and that person’s history says that you have less chance of being successful, or it’s like a writing on the wall and then we go and advise them to say, you can take it, you might have to work harder but we leave it at that but we are the one then left holding the bag later on because it was a writing on the wall. But we simply advised versus saying, sorry, we cannot get you admitted to this course because of the obvious things we see.

Sanjog Aul [00:44:02]:
And please look at something else that partially is of course, a management decision. Other is the way we are trying to solve this problem. As I said, if you don’t hire the right people, maybe this is not the right place for them to Work similarly, this major is simply not right for them. So if you want to really try your luck, go into another school, but don’t accept any student who knowingly is accepting a major or an area which the details and the data shows right there that they will not be as successful. You see where I’m going with this?

David Crane [00:44:37]:
Yeah, absolutely. It’s a gray area because the college readiness isn’t always predictive. Right. So a student may not have really applied themselves in high school and may decide to apply themselves in college. So when they’re first coming in the door, you know, that’s really hard for us to do. I mean, we absolutely set standards within each college. So our College of Engineering may not admit a student that is admitted in College of Liberal Arts and so in that way, we are saying no to some students that we just don’t think would succeed in certain programs.

David Crane [00:45:11]:
But it is tough because each of those colleges is being pressured to keep their enrollment up and in addition to, you know, normal budgeting, there’s performance based budgeting for each school to encourage increased enrollment and retention. And so, you know, we have to be really careful not to lower those standards and put students in a position where they can’t succeed and, you know, at the end of the day, it really takes a strong advisement model but again, predictive analytics, one of the things that’s really helped us with is we can show the student, you know, that based upon your test scores and your high school GPA and what you’ve done in your first year here at SIU, similar students have only succeeded in engineering 15% of the time or 20% or whatever, you know but if you went to this other program, your success rate for similar students is 75%. So you would have a much easier time succeeding. But at least with predictive analytics, we can provide them the information that they need to make an informed decision.

David Crane [00:46:18]:
But yeah, at the end of the day, as long as they’ve been accepted by the college, we let them make that decision.

Sanjog Aul [00:46:24]:
Let’s take a quick break, listeners. We’ll come back and Ashutosh, I’d like to pose a question around the collective intelligence which could have come to rescue if we had the data beyond the regular transactional data, what’s the student’s GPA, what is it that they have done in the past, et cetera, versus having a connection with the high school, which in turn is carried through and that data, that analytics allowed you to make the right hires or right induction of the students who would be a natural fit for a type of course, and then you perhaps will have less of an issue of trying to help them get through the graduation later on. Please stay tuned listeners. We’ll be right back and Explore.

Speaker F [00:47:08]:
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 [00:47:40]:
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 a 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.ciotalk. if you have a question or comment, call toll free at 1-866-472-5790.

Speaker A [00:48:44]:
That number again is 1-866-4725790. Now back to the show.Here’s Sanjog Aul

Sanjog Aul [00:48:53]:
Ashutosh, this is Sanjog asking you this question. As a student who may have lived through high school and now coming into college, if you could figure out a way to talk to my high school teachers or even elementary school depending how far back we go, and learn as analytics of how I learned what ticked me off versus what made my eyes glitter and then that data is passed along to the college where the advisor has been armed with that data and then helps in getting the right major selected. My chances of graduating and my inclination of graduating because I’m coming from one environment to another with a seamless transition would make the percentage of probability much higher. Do you think this is like boiling the ocean or this is even being considered as part of doing your predictive analytics?

Ashutosh Nandeshwar, [00:49:52]:
I think any analyst would say more data is better. We all want more and more and more and more data, but I think at some point of time we have to look at and see, find out whether this data is being useful and that’s why I guess we use machine learning and predictive analytics to find out which attributes stand out. But to your question about using high school level data, so essentially we’re talking about, let’s say transcripts for every course that they have taken, what grades they received and how well they did in each type of course, I think it would be useful to find out what their passion areas are. Just going back to my example, as David was saying, some students start blooming later. It was certainly a case for me. I didn’t do that well in high school or even in college. It was after only grad school when I started finding my passion areas and started focusing in.

Ashutosh Nandeshwar, [00:50:41]:
So there is a danger in over predicting or overfitting the model. But I think it could be using again for generalizing that saying oh, the students take these types of courses and they do well in math type of courses and we know that they would do well in these type of math courses than college. And again, I think that’s when this E coach type of system from University of Michigan could be very, useful when it’s looking at their high school performance and comparing it with other students performances and saying if you right now you are on a trajectory to achieve or get a C, but if you do these other things you might be able to get a B or an A. I think there it could be of some use.

Sanjog Aul [00:51:22]:
Alan, if you were to look at examples of predictive analytics being applied in different educational institutions. What are some of the hindsight 2020 type learnings? Because one is to say we can do this, we can do that and life will be good. Another is to see the actual examples of what has worked, what has not worked and what is showing promise.

Ellen Wagner [00:51:44]:
I think that people tend to be more successful with their analytics initiatives when they actually know what it is that they’re looking for and that they’re trying to solve a problem rather than going in and saying, well I think we’re going to do analytics, let’s look for a pilot place to just try it out. Because the motivation for being able to solve a real problem does I think, focus on identifying the data sources that will answer real questions. So I mean I think the place that I’ve seen, well, it’s just one of those. I think it’s a maturation, a maturation issue because most of us know how to work with analytics and because we’re curious about it. The tendency to turn every analytics initiative that we’re doing into some type of a research project is almost overwhelming. And I think that trying to look for the business intelligence that we can pull out just tends to be, let’s put it this way, I think that there are going to be lots and lots of ways that we’re going to be able to use data to do all kinds of projections or forecasting both for individuals or different areas but right now, looking that far ahead makes it really hard to not get caught in the complexities of this.

Ellen Wagner [00:52:54]:
Answering some questions. Here’s an example. When we started doing our work with the PAR Framework, we had a very specific question that multiple institutions were saying. What would happen if we actually aggregated our student data and created a single federated data set where we could actually do the types of predictions that Amazon or Netflix are doing on their data? Well, it was pretty exciting to think about doing it. We did learn pretty quickly that because we’re dealing with people and not with financial transactions, that even when we have data, that making decisions about someone’s career path is not just going to be a data driven decision. You look at data to inform where you’re going. So I think I’m taking this a little bit further down the path than you may have expected. But I think for me, the balancing act between what we can find and what we can do about what we found just is if people are playing around with the research sides of this, answering questions really aren’t the intended outcome.

Ellen Wagner [00:53:53]:
It just makes it really, you know, research is a really great thing. But I think we all have to decide, are we exploring because we want to know more about this intellect, because we’re intellectually curious, or are we finding ourselves in situations where our feet are being held to the fire, where we must accomplish certain goals that we can’t do unless we get smarter about the impact of what we’ve been doing? So you look at your data so you can make your corrections. I think that’s an important shift. Looking forward rather than looking backward, looking proactively rather than reactively.

Sanjog Aul [00:54:21]:
So, David, when you look in your world and the way you see moving forward, the people who are on the fence of trying this, are there certain cautions that you’d like to offer some pitfalls and gotchas?

David Crane [00:54:36]:
Well, I would say that adoption of technology by faculty is one of our biggest issues. Just because they may have been teaching the same course for 30 or 40 years, exactly the same way. And really, to get the data that we need to be successful in intervening timely in a timely manner, we really need to transition to more data driven instruction methods and so we have a lot of instructors that are very interested in that. We have whole departments who have wholeheartedly jumped into that pool. But we have others that are very resistant to that and so we have started a faculty technology mentoring program where we’re picking a faculty member from each department, fully training them in all of our technology and then they become the technology champion, you know, out in their department with their peers and we feel like that’s helping.

David Crane [00:55:32]:
But honestly, getting accurate data in the system that we can then use to intervene and provide data back to the student has been really our biggest hurdle and we’re trying to overcome that.

Sanjog Aul [00:55:47]:
Ashutosh, one last comment from you. If you were to look at where this is going and we started the conversation that it’s a complex problem, do you think two years down we are still to look at the same level of this being an issue which is being just we are chipping away into it, or do you think there is a realistic timeline by which we can look back and laugh at it and say, yes, we nailed it.

Ashutosh Nandeshwar, [00:56:14]:
I think it would be very oracle for me to say that in two years we’d be able to solve this problem. We have been trying to solve this problem for past 70 to 80 years. I think it’s very hard to put a timeline on it unless we are able to attack the problem from all angles. Aggregate levels coming up with different solutions, counseling centers also providing individual level data, providing arming the faculty and graduate assistants to also be able to help in classes. Then I think we can make definitely certain improvements. One last thing I want to comment on is two books that I read were kind of very transformative for me one was the Grit, the book called Grit, and the other was The Growth Mindset. In both of the books in the Grid book, they studied all the schools from the Harlem School district where they had the KIPP program, where they put the students from the elementary school

Ashutosh Nandeshwar, [00:57:08]:
and their only objective was to put these kids through college and make them successful. And they were coaching them individually every day and these were coming from lower economic backgrounds, so it was a challenge for them. But many students did succeed in high school, but very few did succeed in college and they found out that grit was the perseverance was one of the biggest factors and same is with growth mindset that we have to make students believe that it’s possible for them to really grow into whatever they want to and they must have this growth mindset. So I think we have to answer question not only from technology, but from advising also.

Sanjog Aul [00:57:45]:
On behalf of the show and our listeners, I’d really like to thank you, Ashutosh, David and Ellen, for sharing your thoughts on how predictive analytics can help boost graduation rates.

Ashutosh Nandeshwar, [00:57:57]:
Thank you very much.

Sanjog Aul [00:58:01]:
Thank you. And please like us on Facebook. Search for CIO Talk radio and be sure to follow us on Twitter. Thank you 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 [00:58:14]:
Thank you for tuning in to CIO Talk Radio. To learn more about the show, please visit www.ciottalkradio.com. please join Sun Joke all next Wednesday at 7:00am Pacific Time, 9:00am Central Time and 10:00am 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

• • How to Make Predictive Analytics Effective
  • Pros and Cons: Predictive Coding – Should I Invest?
  • Complex Case Law eDiscovery: Is Predictive Coding a Panacea?
  • Reforming Education With Big Data
  • Is Data Analytics helping increase the impact of K12 Education?
  • Keeping Higher Education Viable Through Technology
  • Education Faculty’s Survival in a Digital Age
  • Can IT Improve Education Governance & Business Sustainability?