November/December 2011 eSCHOOL NEWS • 17

eSN Special ReportBy Jennifer Nastu

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, teachers at

Georgia’s Gwinnet County Public Schools soon will

be able to see at a glance which students might need

more help. Sinclaire Community College in Ohio has

cut its student dropout rate in half. And the online

American Public University System has watched its

course completion rate steadily climb.

These breakthroughs come at a key time for U.S.

education, which is under enormous pressure to

innovate and provide better learning opportunities.

Smarter EducationPredictive analytics can help schools quickly identify at-risk

students—so educators can intervene before it’s too late

Smarter Education, page 18

18 • eSCHOOL NEWS November/December 2011

eSN Special Report

As school district officials struggle to meet the goalof having all students graduate from high school readyfor college or a career, the challenges are significant:Operating costs are on the rise, while budgets for pub-lic institutions are shrinking. Infrastructures are agingand need costly updates. Changing demographics re-quire that schools change, too, to meet the shifting needsof students. Performance is declining at the same timethat expectations are rising. And “working harder” issimply not a sustainable option.

“When we talk with government policy makers andsenior education leaders, there’s a recognition that ed-ucation is the differentiator for national success.Everyone recognizes that education is critical, but[schools] still get their budgets cut all the time,” saidMichael King, vice president of global education forIBM. “People in education really grapple with that prob-lem.”

To overcome these challenges, the education fieldneeds new and innovative approaches. And predictiveanalytics is one such promising solution.

A smarter approachPredictive analytics encompasses a variety of sta-

tistical techniques for mining information gatheredacross a variety of sources—for example, student in-formation systems, library automation systems, learn-

ing management systems, and back-office enterprisesystems—and analyzing those current and historicaldata to make predictions about the future.

The implications of predictive analytics for educa-tion are nearly endless. For instance, schools are usinganalytics software from companies such as IBM or SASto track student performance over time, looking at var-ious data points—not just test or quiz scores, but oth-er, more subtle signals as well, such as how frequent-ly students are logging on to an LMS, or how oftenthey’ve contributed to online discussions—to identifythose who are at risk of failing or dropping out. Someschools are overlaying a system that can identify theseearly warning signs automatically and create a cus-tomized intervention plan for students who might needit.

“Say a student comes into ninth grade with a D ineighth grade algebra. Using statistics, you can see thatsomeone who made a D in Algebra 1 is not ready forAlgebra 2. But you might also see that if they’re notdoing well in math, they’re probably not doing well inscience,” said J. Alvin Willbanks, superintendent ofGeorgia’s Gwinnett County Public Schools.“Attendance patterns can also be shown: Maybe thatstudent was out of school 20 percent of the time. Thiscould indicate there’s a discipline problem. Using an-alytics can alert teachers to these things, so they can beproactive in putting interventions in place to get the stu-dent back on track.”

Working with IBM, Gwinnett County is creating

what is calls its digital content learning assessment sup-port system, or eCLASS. When it is completed,eCLASS will give teachers the ability to open a pageand see their class roster in a dashboard format. At aglance, teachers will get important information abouttheir students, with the ability to drill down to see in-dividual student results and suggestions for addressingeach student’s academic weaknesses.

In higher education, Sinclaire Community Collegeused analytics to look at registered students who had-n’t paid yet, perhaps because their grants or loans hadnot yet come through. Those students were in dangerof having their registration cancelled. By doing theanalysis and intervening with these students before can-cellation occurred, the college was able to halve its stu-dent registration dropout rate. That allowed the schoolto hold onto its state funding and ultimately increase itsgraduation rate.

The 70,000-student American Public UniversitySystem, a fully online school, uses IBM’s SPSSModeler to measure key student performance, partici-pation, and attendance information to predict when stu-dents are in danger of dropping out, so those studentscan be given additional support. APUS also is part ofa national initiative to study the factors influencing col-lege dropout rates, so colleges and universities can de-ploy analytics technology more effectively in theirstudent retention efforts. (See the side story “Analyticsuse boosts student retention.”)

Smarter Education...continued from page 17

Smarter Education, page 20

Dennis CarterAAssssiissttaanntt EEddiittoorr

Analytics programs that use complex data sets toidentify struggling students, suggest the best ways touse campus budgets, and improve faculty and staffefficiency soon could be coming to more U.S. col-leges and universities, thanks to a new initiative fromthe higher-education technology group EDUCAUSE.

EDUCAUSE, a nonprofit organization that heldits annual conference Oct. 18-20 in Philadelphia, an-nounced a three-pronged initiative to bolster the useof analytics technology in higher education.

The initiative will include a “major benchmark-ing study of the state of analytics in higher educa-tion” that could show how many campuses are us-ing analytics to improve decision making and studentsuccess. It will end with a national summit for col-lege and university leaders in fall 2012, accordingto the group’s announcement.

In between, EDUCAUSE will work within thehigher-education community to provide thought lead-ership and training to help schools develop the ca-pacity to use analytics.

“Sophisticated uses of data and benchmarkinghave revolutionized business and industry,” saidDiana Oblinger, president and CEO of EDUCAUSE.“These same techniques can deepen our under-standing of learner success, improve our ability totrack progress, and help us make the best decisionsabout campus resources.”

Campus technologists at the forefront of analyt-ics use have discovered the benefits of computer-based models that show precisely how to save mon-ey and improve student grades, but Oblinger said

analytics is far from mainstream use in higher edu-cation.

“Analytics holds transformative promise for edu-cation, but the field is still in the developmental stages,”she said, adding that EDUCAUSE’s initiative “comesat the right moment for higher education.”

Analytics-based early interventions have im-proved course completion rates demonstrably, ED-UCAUSE said. In the future, analytics will enablehighly personalized learning environments charac-terized by adaptive testing and content displays,point-of-need help, and academic support servicestailored to individual student needs.

A whitepaper published by EDUCAUSE last yeardescribed how a classroom analytics program usedat a large state university helps instructors identifywhich students are likely to fail their class by ana-lyzing classroom participation—including contribu-tions to the class’s online discussion board—as wellas quiz grades and the number of times a studentviews course notes online.

The software analyzes these variables and uses acolor-coded indicator to let professors know whichstudents are most at risk. A green light shows the stu-dent is doing well in class, a yellow light indicates“possible risk” of slipping below a C, and a red lightshows “elevated risk” that a student could fail thecourse after failing to participate in in-person and on-line class discussions and scoring poorly on quizzes.

The program also shows instructors which stu-dents are participating in course discussions the most,and correlates this information to class grades.Professors can send an eMail message to notify stu-dents if they appear at risk, congratulate students for

work well done, or suggest ways to get more fromthe class.

“Most colleges and universities collect and storevast amounts of data—in the LMS, admissions files,student and library records, and other systems,” thereport said. “Analytics applications can mine at leastsome of this data, subject it to statistical analysis, andprepare reports or data visualizations to reveal pat-terns, trends, and exceptions.”

The new EDUCAUSE initiative is supported bya grant from the Bill & Melinda Gates Foundation.

Ed-tech group to push for moreanalytics use in collegesOn the eve of its annual conference, EDUCAUSE announced plans to bring more analytics technology to campuses nationwide

Analytics holds ‘transformative promise.’

eSN

What if…you could predict which students would need help making the grade?

SPSS, an IBM Company has innovative, cost effective solutions so you know

which students need help...and in what areas...long before they actually fail.

Visit our free resource center: www.spss.com/academic

20 • eSCHOOL NEWS November/December 2011

eSN Special Report

While looking at a student’s prior grades to see howwell he or she will do seems rather obvious, predictiveanalytics gives educators the ability to look more deeplyinto the data, said Karen Patch, senior technical archi-tect for SAS. “What you’re able to find is hidden trendsand patterns that you otherwise wouldn’t be awareof,” Patch said.

Predictive analysis has been used successfully infinancial, insurance, retail, and other industries foryears. “This is how you get targeted ads over the in-ternet,” said Alex Kaplan, national practice leader foreducation 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’sa sophisticated use of data, and it can be directly ap-plied to education.”

Sometimes the technology reveals key informationthat might come as a surprise, helping educators lookat situations in a whole new light.

“One really big issue is how engaged students are inschool,” Kaplan said. You might think that the more in-volved in activities a student is, the more likely he orshe is to fall behind academically. But Kaplan saidschools are finding that the opposite is true. The more

engaged students are in school—using social collabo-ration tools where they can chat with each other online,accessing websites that contain curriculum materialsand lesson plans, spending time in other activities suchas drama or sports—the more likely they are to succeedin their studies.

“You might have students who are performing poor-ly academically, but who are really engaged in school,so you know they’re excited about school but are maybestruggling in a certain topic,” Kaplan said. “Or, youmight have someone who is scoring well but is reallynot engaged at all. That student might actually be at riskof dropping out and would normally fall through thecracks. We’re now in the position to give this informa-tion” to school leaders before it’s too late, he said.

Now, Kaplan said, instead of relying solely on theirintuition and observation, educators and administratorscan have quantitative data to help them make better,smarter decisions. “And it’s not just about the individ-ual student; it’s also about how the school thinks aboutinstruction,” he said. “Once [officials] identify patternsand trends, they might learn that everybody is strug-gling with the periodic table in chemistry, so it givesthem a deeper insight into the instructional process.”

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 onerecent survey 90 percent of respondents said they hadattained a positive ROI from their most successful de-ployment of predictive analytics—and more than halfachieved a positive ROI from their least successful de-ployment. Now, IBM has developed what it is callinga “vision for smarter education,” creating an analyticsframework for schools that builds upon its expertise inanalytics, business processes, and technology integra-tion.

IBM’s new framework combines predictive analyt-ics with “intervention management” technology, whichcan trigger a specific intervention that is unique to eachstruggling student’s needs and then deliver this reme-dial or supplemental content directly to the student. Theentire process occurs through a single dashboard inter-face, IBM says; the solution is in beta-testing now andwill be commercially available for schools and collegesearly next year.

Analytics features are even appearing in popularLMS programs. Instructure, whose Canvas LMS pro-gram is available as either an open-source version thatschools manage themselves or a cloud-based modelhosted by the company, plans to release a version inearly 2012 that includes predictive capabilities.

On their course roster, instructors will see green, yel-

Smarter Education...continued from page 17

Dennis CarterAAssssiissttaanntt EEddiittoorr

Phil Ice knows numbers never lie.Ice, the director of course design, research,

and development for American PublicUniversity System (APUS), has watched re-tention rates at the 70,000-student onlineschool steadily climb with the continued analy-sis of in-depth information that shows when astudent might be on the verge of dropping out.

If a student’s test scores are dropping, par-ticipation numbers are low, and disengagementis evident through various statistics, the num-bers suggest that student might not last muchlonger at APUS.

How can professors and university officialsknow precisely which students are in dangerof giving up on their education? One surefirestrategy is to examine how many days havepassed since the student last logged onto his orher course website.

If it’s been a while since the online student checkedthe site for syllabus updates or discussion sessions,APUS’s analytics system will flag the student as a po-tential dropout.

“We hone in on things such as a student’s percep-tion of being able to build effective community,” Icesaid.

APUS professors and instructors use informationdetailing a student’s online engagement with a com-prehensive survey to create a model of student reten-tion and satisfaction, according to the university.

“We’ve been using analytics before it was a buzzword in higher education,” Ice said.

APUS uses IBM’s SPSS Modeler, which uses keystudent performance, participation, and attendance in-formation in part to measure a student’s social pres-ence, along with a student’s perception of online learn-ing’s effectiveness.

These two factors have proven to be reliable vari-ables telling professors and instructors how likely a stu-dent is to drop out of classes.

APUS recently joined the Colorado CommunityCollege System, Rio Salado College, the University ofHawaii System, the University of Illinois-Springfield,

and the University of Phoenix in a nationwide initia-tive aiming to improve analytics and increase its use inhigher education.

The Predictive Analytics Reporting (PAR)Framework project, launched last May by the edu-cational technology group Western InterstateCommission for Higher Education’s Cooperative forEducational Technologies (WCET), will examine sixcritical data sets as a single sample that could makeanalytics more effective on college campuses.

The six colleges and universities participating inthe PAR project will create a pool of information from400,000 students who will remain anonymous inthe research. Analysts will use this massive collec-tion of student records to assess factors that affect re-tention and learning outcomes among online students.

Ice said about nine in 10 colleges use some formof statistical analysis to determine retention and learn-ing strategies on campus, but only 9 percent use his-torical data to supplement current numbers—and onlyabout 1 percent of U.S. colleges and universities use“deep data mining” to explore why dropout ratesmight be rising.

“That number really needs to expand, especiallyaround the crisis of student enrollment,” Ice said. “We

need to let the data speak for themselves.”Campuses large and small have not been

immune to spiking dropout rates. University of Kansas officials worked

with a data-mining company in 2010 to pin-point strategies to keep students enrolled af-ter a university report showed that 28.7 per-cent of freshmen from the fall 2007 semesterhave left the campus.

Kansas’s 28.7 percent dropout rate amongfall 2007 freshmen was significantly higherthan peer institutions, said ChristopherHaufler, a University of Kansas professorwho chaired the school’s student retentiontask force.

The university’s average retention rate af-ter one year is 80 percent, he said, comparedwith 85 percent to 90 percent at peer insti-tutions.

Kansas’s retention system uses analyticsto raise real-time online flags for students who fallbelow a certain grade or involvement “threshold”designated by the university.

Nationwide, fewer than three-fourths of two-yearcareer college students return to school after theirfirst year, according to research released by the non-profit Imagine America Foundation. Just 57 percentof public community college students return after oneyear, and 68 percent return after a year at a privateinstitution, according to the research.

Other schools, such as Western Iowa TechCommunity College (WITCC), use student tuitionpayment data to gage how likely a student is to leaveschool.

Educators often have a difficult time tracking theengagement level of online students, who—unliketraditional students—don’t interact with their pro-fessors and fellow students every day in class.Analytics tools, like the kind used at APUS, couldmake that task much easier, Ice said.

“By identifying patterns of performance, we be-lieve we can create an approach to applying predic-tive analytics … that will help practitioners and stu-dents alike spot barriers to success before theybecome problems,” he said.

Analytics use boosts student retention

Smarter Education, page 22

eSN

22 • eSCHOOL NEWS November/December 2011

low, or red dots next to students’names, indicating howat risk they are of failing or dropping the course.Clicking on a student’s name will take instructors to apage where they can see more detailed reports basedon that student’s grades, class participation, assignmentcompletion, and outcomes (whether he or she has mas-tered the content).

At the bottom of the course roster, instructors willsee a list of students who are most at risk in the class.The software also will contain dashboards for admin-istrators to view the same information for entire de-partments or schools.

Getting started with predictive analyticsFor school district leaders interested in using pre-

dictive analytics, the first step is to understand the kindsof data you want to analyze. Then, you must create adata warehouse that pulls this information from all

available sources, including student information sys-tems, learning management systems, enterprise soft-ware, and other areas. And the data need to be both con-sistent and trustworthy. “Having this information allin one place gives schools an enormous amount ofleverage,” IBM’s Kaplan said.

The next step is to think about the key performanceindicators you want to be able to compare. Some com-mercial analytics software might have certain optionsbuilt into the system, while others might give users theflexibility to design a nearly unlimited number ofqueries themselves. Whatever option you choose, youshould recognize that your needs are likely to evolveas you dig deeper into using the system.

Education leaders also should consider what actions,if any, they want certain indicators to trigger, such assending a message about tutoring options to a studentwho is considered at risk of failing. Again, some soft-ware programs might contain certain built-in interven-tion choices, while others might not.

In an April 2010 whitepaper called “7 Things YouShould Know About Analytics,” the higher-education

technology group EDUCAUSE noted that analyticstechnology is a powerful tool that can help schools“identify where and when certain investments will havethe greatest benefits.” But the organization cautionedthat even the best data algorithms can result in mis-classifications of students, “in part because such pro-grams are based on inferences about what different sortsof data might mean relative to student success.”

In other words, predictive analytics software is onlyas useful as the data it draws from—and the supposi-tions that can be made from how those data interrelate.

That’s why it’s important to have as complete a pic-ture as possible of a student’s academic history. Andthat’s the motivation behind states’efforts to create lon-gitudinal data systems that can follow a student’sprogress from pre-kindergarten through college grad-uation and on to the workforce—a development thatIBM’s King finds encouraging.

“If you can track a student across multiple schools,you have a common view of the student that can helpdrive understanding,” he said.

Other countries are starting to look at how data sys-tems can help their educational systems fuel econom-ic growth, by identifying students coming through thepipeline as individuals with particular skill sets that cansupply the country’s needs in various areas. By takinga “P80” view of students—that is, a single longitudi-nal view of each student from pre-kindergarten throughage 80—policy makers and education leaders are ableto look at the “supply chain” of people moving intothe workforce and discern whether there are enoughteachers or nurses, for example.

In order to compete globally, this is the direction theUnited States needs to take as well, King argues: “It’simportant that our educational leaders and policy mak-ers recognize that that’s where we’re heading. It’s acompletely different competitive dynamic. We have tostart thinking not just about our schools, but … how wetransform our state systems by bringing together a P80view of each student.”

Jennifer Nastu is a freelance writer who writes fre-quently about education and technology.

Smarter Education...continued from page 20

Predictive analytics software can be a powerful tool, but it’s only as useful as the data it draws from.

eSN Special Report

Like most states, Minnesota contains manyschool districts that are quite small; one district,for example, has just 500 students. Fortunately,the state is home to TIES, a joint-powers coop-erative that is owned by 46 Minnesota school dis-tricts. TIES (which is actually a school district bycharter in itself, though it doesn’t have any stu-dents) offers analytics packages to its membersvia Cognos, a company that was purchased byIBM in 2007.

TIES gives districts the opportunity to usehigh-powered tools they otherwise wouldn’t havebeen able to afford, said Ben Silberglitt, the or-ganization’s director of software applications.“We help districts set benchmark target scores onlocal assessments, so they can use those assess-ments as indicators for state assessments,” saidSilberglitt. “This helps them identify early onwhether students are on track, not just by studentbut by student population.”

Schools then are able to pinpoint where theyare effective and where they are not. “For exam-ple, they might learn how effective they are withEnglish language learners who are below target.They might learn they are really effective withSpanish-speaking students, but not with studentswho have other languages spoken at home,” hesaid. “Or maybe it’s attendance: This particularpopulation has a low attendance level and is hav-ing trouble in school. That helps [the schools] planfor student success.”

TIES began using Cognos technology sevenyears ago, but the predictive analytics element hassignificantly come into play in the last five years.TIES created a data warehouse to give its mem-ber districts not just easy access to their data, butthe ability to connect disparate sources of data—looking, for instance, at how attendance impactsinstruction. “Enterprise software systems are notdesigned to get data out in order to answer thosekinds of questions,” Silberglitt said. “So puttingthat warehouse layer on top has been great interms of answering those questions on the fly.”

The districts do not have to approach TIESwith their questions; TIES provides not only thesoftware, but also the training to use it effective-ly. “When districts want to move forward with acertain model, we show them the process behindanswering those questions, such as what types ofreports they need to run,” he explained.

The school districts that have taken advantageof their access to predictive analytics have foundthat using data in this way has made them muchmore effective.

“They can make decisions more quickly andare more proactive about changing their ap-proaches. They might discover, for example, thatthey need to apply more of their resources to thethird grade, because they have a cohort there whohasn’t grown as much as the other cohorts in thelast two years. When they learn that, they knowthey need to shift things around. And we knowthat leads to better outcomes,” Silberglitt said. —J.N.

Analyticsimproveseffectiveness of Minnesotaschools

eSN

eSN

November/December 2011 eSCHOOL NEWS • 23

Randy Sumrall, chief information officer forthe Education Service Center (ESC) Region10 in eastern North Texas, knows the impor-tance of case management for individual stu-dents—that is, looking at where at-risk studentsare in their education, discovering where in-tervention needs to take place, deciding whatthat intervention will be, and giving them theadditional instruction and follow-up they need.

So he was happy when, in 2005, theTechnology and Data Services division of ESCRegion 10 partnered with IBM to providethought leadership to its client districts on theuse of data warehousing and business intelli-gence to become more data-driven, and hewas hired to lead the initiative.

ESC Region 10 is one of 20 education ser-vice agencies for the state of Texas. It providesadministrative, instructional, and informationservices to 80 districts and 32 charter schools.ESC Region 10 and IBM built Empower, aneducational data warehouse that could drawfrom multiple silos of data, such as student per-formance, student demographics, finance, andhuman resources, and provide comparisonsacross multiple school districts. The systemruns on technology from IBM Cognos.

ESC Region 10 piloted the system with nine

districts and began production with 15 districts.The biggest element the districts looked atwas student performance. For example, saidSumrall, one district gave its principals a “snap-shot” of their students multiple times throughthe year to pinpoint the students’ specific needs.

But at that point, Region 10 discovered thepiloted districts were all it could handle.“Without charging unaffordable fees, we couldnot scale the system,” said Sumrall. “The datawere much more diverse than we expected,with no data standard, even within a single stu-dent or business database vendor.”

At that time, the Texas Education Agencyand the Michael and Susan Dell Foundationannounced their intention to create a state datawarehouse based on a data standard to be de-veloped. The state data warehouse would pro-vide student performance dashboards through-out the state.

“We knew immediately that this would be-come the solution we were pursuing,” Sumrallsaid, and they transferred their intellectualproperty and assistance to the new TexasStudent Data System. Now, Sumrall and theESC Region 10 are helping to support two dis-tricts in a limited production of the TexasStudent Data System. One solution stemming

from the state system is the Reveal DropoutEarly Warning System. It uses data from thestate data warehouse to provide a compre-hensive view of students at risk of droppingout, automatically and dynamically.

“At Region 10, we believe the data ware-house is only a part of the needed school in-formation solution. There are four major areaswhere school information needs to be man-aged, and they are interconnected,” saidSumrall. “They are data warehousing and re-ports on student and teacher performance, pre-dictive analysis of potential instructional andintervention solutions, management of contentthat is correlated to state standards with qual-ified built-in assessments, and case manage-ment to unify the efforts of multiple specialiststhat wrap around the needs of the individualstudents.”

As the Texas Student Data System is beingdeveloped, ESC Region 10 is helping its dis-tricts understand what data they have, whatdata they want, and what they want to usethis information for. “If you want to be a data-driven district, you need to know what you wantto collect. Often there’s a total disconnect be-tween getting the data and knowing how touse it,” explained Sumrall. —J.N.

Texas system provides early warnings to help stem dropouts

eSN

Learn how IBM’s advanced case management strategy can enable institutions with greater insight and information at http://www.ibm.com/software/data/advanced-case-management.

Managing student growth and success requires a coordinated set of instructional processes and a single view of student information. With advanced case management capabilities from IBM, educators can identify “at risk” students at an early stage, track intervention programs and capture the impact on individual student performance. These tools can enhance a teacher’s observation of student progress with the insights to identify and deliver more effective learning. Collaboration and communication tools can also help all members of the learning community connect and focus on student success.