[1]For years, marketers have used sophisticated software to track consumers’ buying habits and web browsing activity, then crunch this information and—based on the data—make a series of intelligent predictions that allow them to target their sales messages much more effectively.
Now, this same technology is appearing in schools and colleges as well—and observers say it’s a development that could revolutionize education.
Using predictive analytics software, Kennesaw State University in Georgia has been able to hone in on prospective students more efficiently. Sinclair Community College in Ohio has cut its student dropout rate in half.
And the online American Public University System has seen its course completion rate steadily climb.
These breakthroughs come at a key time for U.S. higher education, which is under enormous pressure to innovate and provide better learning opportunities.
As campus officials struggle to meet the Obama administration’s ambitious goal of once again leading the world in the percentage of college graduates by 2020, the challenges are significant: Operating costs are on the rise, while budgets for public institutions are shrinking. Infrastructures are aging and need costly updates.
Changing demographics require that schools change, too, to meet the shifting needs of students. Performance is declining at the same time that expectations are rising. And “working harder” is simply not a sustainable option.
“When we talk with government policy makers and senior education leaders, there’s a recognition that education is the differentiator for national success. Everyone recognizes that education is critical, but [schools] still get their budgets cut all the time,” said Michael King, vice president of global education for IBM. “People in education really grapple with that problem.”
To overcome these challenges, the education field needs new and innovative approaches. And predictive analytics is one such promising solution.
A smarter approach
Predictive analytics encompasses a variety of statistical techniques for mining information gathered across a variety of sources—for example, student information systems, library automation systems, learning management systems, and back-office enterprise systems—and analyzing those current and historical data to make predictions about the future.
The implications of predictive analytics for education are nearly endless.
For instance, schools are using analytics software to track student performance over time, looking at various data points—not just test or quiz scores, but other, more subtle signals as well, such as how frequently students are logging on to an LMS, or how often they’ve contributed to online discussions—to identify those who are at risk of failing or dropping out.
Some schools are overlaying a system that can identify these early warning signs automatically and create a customized intervention plan for students who might need it.
Other colleges and universities are using analytics software to more accurately predict enrollment numbers, which helps them plan in a variety of areas—from budgeting and staffing to anticipating the amount of parking that will be available to students. Still other campuses are using predictive analytics to recruit students who are more likely to enroll and do well.
“Detailed modeling allows you to understand more about your student body,” said Georgia Mariani, education industry marketing manager for SAS. “If you know ahead of time how many students are going to be enrolled—something that fluctuates from year to year—you can decide if you have enough books in the bookstore, enough introductory classes, enough labs. Then you can look at the probability of those students actually graduating. You can look at the effectiveness of the programs in which they’re being placed. You can also look at alumni: Who’s more likely to give, how likely they are to give.”
Some of these uses might seem a little basic at first glance: Isn’t it easy, for example, to look at a student’s assessment scores to predict how he or she will perform in the future?
Sure, said Karen Patch, senior technical architect for SAS—looking at a student’s prior grades to see how well he or she will do is intuitive, but analytics gives educators the ability to look more deeply into the data. “What you’re able to find is hidden trends and patterns that you otherwise wouldn’t be aware of,” Patch said.
Predictive analysis has been used successfully in financial, insurance, retail, and other industries for years.
“This is how you get targeted ads over the internet,” said Alex Kaplan, national practice leader for education at IBM’s Global Business Services division. “These firms are mining data and looking for patterns, such as how likely a person is to make a purchase. It’s a sophisticated use of data, and it can be directly applied to education.”
Sometimes the technology reveals key information that might come as a surprise, helping educators look at situations in a whole new light.
“One really big issue is how engaged students are in school,” Kaplan said. You might think that the more involved in activities a student is, the more likely he or she is to fall behind academically. But Kaplan said schools are finding that the opposite is true.
The more engaged students are in school—using social collaboration tools where they can chat with each other online, accessing websites that contain curriculum materials and lesson plans, spending time in other activities such as drama or sports—the more likely they are to succeed in their studies.
“You might have students who are performing poorly academically, but who are really engaged in school, so you know they’re excited about school but are maybe struggling in a certain topic,” Kaplan said. “Or, you might have someone who is scoring well but is really not engaged at all. That student might actually be at risk of dropping out and would normally fall through the cracks. We’re now in the position to give this information” to school leaders before it’s too late.
Now, Kaplan said, instead of relying solely on their intuition and observation, educators and administrators can have quantitative data to help them make better, smarter decisions.
“And it’s not just about the individual student; it’s also about how the school thinks about instruction,” he said. “Once [officials] identify patterns and trends, they might learn that everybody is struggling with the periodic table in chemistry, so it gives them a deeper insight into the instructional process.”
The higher-education technology group EDUCAUSE, meanwhile, has launched a new initiative to encourage the use of analytics technology among colleges and universities. The project will culminate with a national summit for campus leaders to explore analytics use in more detail next fall.
Analytics features are even appearing in popular LMS programs. Instructure, whose Canvas LMS program is available as either an open-source version that schools manage themselves or a cloud-based model hosted by the company, plans to release a version in early 2012 that includes predictive capabilities.
On their course roster, instructors will see green, yellow, or red dots next to students’ names, indicating how at risk they are of failing or dropping the course.
Clicking on a student’s name will take instructors to a page where they can see more detailed reports based on that student’s grades, class participation, assignment completion, and outcomes (whether he or she has mastered the content).
At the bottom of the course roster, instructors will see a list of students who are most at risk in the class. The software also will contain dashboards for administrators to view the same information for entire departments.
IBM has worked with industry leaders for years, helping businesses use predictive analytics as a data-driven system for managing risk and improving their return on investment.
In fact, the company says, in one recent survey 90 percent of respondents said they had attained a positive ROI from their most successful deployment of predictive analytics—and more than half achieved a positive ROI from their least successful deployment. Now, IBM has developed what it is calling a “vision for smarter education,” creating an analytics framework for schools that builds upon its expertise in analytics, business processes, and technology integration.
IBM’s new framework combines predictive analytics with “intervention management” technology, which can trigger a specific intervention that is unique to each struggling student’s needs and then deliver this remedial or supplemental content directly to the student.
The entire process occurs through a single dashboard interface, IBM says; the solution is in beta-testing now and will be commercially available for schools and colleges early next year.
Examples of analytics use in higher education
Sinclair Community College used analytics to look at registered students who hadn’t paid yet, perhaps because their grants or loans had not yet come through.
Those students were in danger of having their registration cancelled. By doing the analysis and intervening with these students before cancellation occurred, the college was able to halve its student registration dropout rate. That allowed the school to hold onto its state funding and ultimately increase its graduation rate.
Worried that it might not have enough computing capacity for its growing online program, Sinclair College also used predictive analytics to gauge when it would run out of server space and was able to take steps to ensure it would have enough capacity to serve all of its online students.
The 70,000-student American Public University System, a fully online school, uses IBM’s SPSS Modeler to measure key student performance, participation, and attendance information to predict when students are in danger of dropping out, so those students can be given additional support.
APUS also is part of a national initiative to study the factors influencing college dropout rates, so colleges and universities can deploy analytics technology more effectively in their student retention efforts.
Purdue University has been using a home-grown system called Signals for a few years now.
Analyzing factors such as how often a student reviews online resource material, participates in course-related chats, and communicates with professors and teaching assistants, the system gives students “stoplight indicators”—red, yellow, or green—to encourage or warn them well before midterm exams and direct them to appropriate supports if necessary.
Data crunchers at Kennesaw State felt their system of IT-centric reporting was holding the university back. The system required that staff members put reporting requests into a queue; then a team of developers would have to write queries to fulfill these requests.
The process could take as long as two weeks, and then—if the person requesting the information discovered that he or she didn’t ask the right question—revisions could take another several days.
“Any sort of number crunching or analytics had to occur on desktop software like Microsoft Excel, and not a lot of that was even being done,” said Erik Bowe, chief data officer for the university. “There was no integrated data warehouse and no uniform set of reporting tools. The reporting system we had for half a decade was not allowing us to look forward and grow.”
Bowe put SAS’s Enterprise Intelligence Suite for Education into place, which now lets administrators directly access the information they need to plan classes, develop budgets, and track student progress. The IT staff is free to look at the big picture and help the university hold down administrative costs.
Kennesaw State uses the new system to improve planning and budgeting in a variety of ways. For example, said Bowe, administrators created a historical timeline of information around the high schools from which the university typically recruits the most students.
“We looked at five-year trends …of what types of students were coming in, whether they were remedial students, students with special needs, students looking for traditional programs,” Bowe said. Officials discovered that over a five-year period, the university had experienced a 30-percent decline in student recruitment from a particular high school.
When officials dug more deeply into the data, they learned that the enrollment of that high school also was decreasing. This knowledge helped officials plan their future recruiting efforts.
They also looked at the demographics of those students coming in over the five-year period, by ethnicity, race, and gender. This demographic information was useful because the university recently had received a grant from the Goizueta Foundation to improve its recruitment and retention of Hispanic students.
The analytics allowed them to see where their strengths lay in terms of recruiting Hispanic students, leading to better decision-making in leveraging the grant funds.
In yet another example of analytics use, administrators examined the university’s retention, progression, and graduation (RPG) data. The associate vice president of RPG asked for a report that looked at first-time freshmen who graduated in six, seven, eight, or nine years.
“We pulled the report together in about three hours, and it revealed that two programs in particular were bottlenecks, causing our students to delay graduating in a four- to six-year cycle,” said Bowe. Knowing where these bottlenecks are, the university can now allocate resources to help students and staff in those programs be more effective. “We’ve shifted our focus from report writing to assisting people in getting the right data out of those reports,” Bowe said.
First steps in using predictive analytics
For school leaders interested in using predictive analytics, the first step is to understand the kinds of data you want to analyze. Then, you must create a data warehouse that pulls this information from all available sources, including student information systems, learning management systems, enterprise software, and other areas.
And the data need to be both consistent and trustworthy. “Having this information all in one place gives schools an enormous amount of leverage,” IBM’s Kaplan said.
The next step is to think about the key performance indicators (KPIs) you want to be able to compare.
Some commercial analytics software might have certain options built into the system, while others might give users the flexibility to design a nearly unlimited number of queries themselves. Whatever option you choose, you should recognize that your needs are likely to evolve as you dig deeper into using the system, as Kennesaw State’s experience indicates.
“That’s one mistake we did make,” said Bowe. “The metrics and measures bubbled up from the lower echelons of our organizations, they didn’t drive downward. There wasn’t a concerted effort to identify the KPIs we needed to track.”
He added: “We realized we didn’t necessarily agree on what all the KPIs were that we needed. Now, we’re just in that process.”
Campus leaders also should consider what actions they want certain indicators to trigger, such as sending a message about tutoring options to a student who is considered at risk of failing. Again, some software programs might contain certain built-in intervention choices, while others might not.
In an April 2010 whitepaper called “7 Things You Should Know About Analytics,” EDUCAUSE noted that analytics technology is a powerful tool that can help institutions “identify where and when certain investments will have the greatest benefits.”
But the organization cautioned that even the best data algorithms can result in misclassifications, “in part because such programs are based on inferences about what different sorts of data might mean relative to student success.”
In other words, predictive analytics software is only as useful as the data it draws from—and the suppositions that can be made from how those data interrelate.
That’s why it’s important to have as complete a picture as possible of a student’s academic history. And that’s the motivation behind states’ efforts to create longitudinal data systems that can follow a student’s progress from pre-kindergarten through college graduation and on to the workforce—a development that IBM’s King finds encouraging.
“If you can track a student across multiple schools, you have a common view of the student that can help drive understanding,” he said.
Other countries are starting to look at how data systems can help their educational systems fuel economic growth, by identifying students coming through the pipeline as individuals with particular skill sets that can supply the country’s needs in various areas.
By taking a “P80” view of students—that is, a single longitudinal view of each student from pre-kindergarten through age 80—policy makers and education leaders are able to look at the “supply chain” of people moving into the workforce and discern whether there are enough teachers or nurses, for example.
In order to compete globally, this is the direction the United States needs to take as well, King argues: “It’s important that our educational leaders and policy makers recognize that that’s where we’re heading. It’s a completely different competitive dynamic. We have to start thinking not just about our schools, but … how we transform our state systems by bringing together a P80 view of each student.”