Analytics helps Indiana community college put a dent in the dropout problem


This is a sponsored post commissioned by Pentaho, a Hitachi Group Company. Sponsor Posts are paid posts that appear on all pages of, supporting editorial efforts. For pricing and inquiries, email

Fewer than two out of every three Americans who enter a four-year college program graduate within six years. Thirty percent don’t even complete their first year. The stubbornly high college dropout rate remains one of America’s most intractable education problems. Despite having one of the world’s largest and most diverse college education systems, the U.S. ranks dead last in graduation rates among the 18 developed economies tracked by the Organization for Economic Cooperation and Development.

One community college system thinks it’s on track toward cracking the problem with an approach driven by predictive analytics. It’s learning to spot the warning signs of failure so that early intervention can get students back on track. Based on early results, it may just be on to something.

Ivy Tech Community College is a sprawling network of nearly 120 campuses and satellite locations spanning the state of Indiana. It’s the largest singly accredited community college in the country, with nearly 175,000 students each year. A system that big generates a lot of data – about 100 million new records per day. Ivy Tech is using self-service analytics to bust bottlenecks and get information into the hands of the people who need it.

Ivy Tech defies the common wisdom that educational institutions are technology laggards. In 2015, it became only the ninth higher education institution in the country to appoint a chief data officer and the first community college ever to do so.

The journey started five years earlier, when Lige Hensley signed on as chief technology officer with a mission to make the information technology organization more responsive to the needs of the business. As he saw it, the key was to break down the the barriers that stood between educators and the information they needed to do their job. Self-service would be the antidote.

Democratizing data

Lige hensley, Ivy Tech“We believed we could create a ‘data democracy’ that would empower data-informed decision-making, accelerate institutional processes and speed innovation,” he said.

That meant building a data warehouse, but a more important factor in Ivy Tech’s makeover was to change the way it approached data. In 2010, the college had only twenty-two employees in decision support or data research roles and just 225 people who were authorized to build reports. As a result, those reports often took hours to run, assuming that the latest information was even available, which it often wasn’t. Frustrated administrators were forced to rely on gut instinct to make too many decisions.

“At many colleges, you have to log into four or five systems and dump reports into Excel to try to make sense of them,” he said. “We had 40 or 50 systems like that, so if you wanted to find out how Johnny is doing as a student, well, good luck.”

Brendan Aldrich, who stepped into the role of chief data officer two years ago, shared Hensley’s enthusiasm for democratizing data. Creating an analytics culture is about “more than just access,” he said. “Data should be useful and evolve over time without people constantly having to go back to an IT or analytics team.”

After an exhaustive RFP process, the college selected Pentaho Corp.‘s data integration and business analytics platform as a solution. Pentaho combined data integration and analytics with a self-service approach that meshed well with the institution’s needs. As an open source-based platform with a web-based user interface that could be embedded into existing applications, Pentaho provided both the flexibility and functionality required, without costly desktop licenses.  That was important, as Aldrich and Hensley envisioned their user population scaling into the thousands.

Small teams, big ideas

By leveraging this new new technology foundation, the team of just three engineers and six analysts could deliver a data democracy that empowered faster, more informed decision-making. Called “NewT” (for “New Thing”), the system is based on the concept of curated data sets, or collections of data from multiple internal and external systems that relate to specific functions, such as financial aid, student records and accounts payable. Ivy Tech’s data scientists and citizen analysts have now built about 40 such collections over the past three years, and more come online all the time. About 100 million rows of data flow into the warehouse each day with more expected as additional systems come online.

Aldrich and Hensley’s vision was to flip the traditional data warehousing model on its head. Instead of isolating expertise in the hands of a few technology wizards, NewT was built to make analytics accessible to nearly any employee who wanted it. The 12 analytics experts on staff fanned out across the state to teach employees how to interact with and visually analyze data. Last year, they trained more than 3,000 people — about one third of the college’s total workforce — in just one six-month period.

Democratizing data is empowering but also carries with it responsibility. The owners of each collection are charged with ensuring data quality and consistency and for applying a common set of governance principles. Delegating data ownership made sense. “Maintaining consistent logic between 40 data collections is easier than attempting to do so across hundreds of reports written by dozens of developers over a period of years,” Aldrich said.

So did moving data quality control further up the value chain. In another break from the norm, Ivy Tech shifted data preparation and cleansing away from the warehouse and into production. “Everyone tries to clean data in the warehouse instead of where it should be, which is in front-end systems,” Aldrich said. “Our philosophy is that if the data isn’t clean, let’s see it so that we can all fix it.”

Each functional area uses a consistent vocabulary based upon a set of common definitions, along with notes on how to use them. Those notes are published on a wiki with a bridge to Pentaho, so data definitions are integrated with the data itself. When users mouse over a data element, the definition from the wiki pops up instantly.

Basic reporting is easy. Users can drag and drop data elements into and out of existing templates, reports, graphs, and dashboards without the aid of IT or a special analytics team. Each department has a set of commonly used basic reports and templates are used to onboard new staffers, and employees are also encouraged to create new models and to share them with each other through governed public folders. Reports can be shared between users, departments or even across the entire state college system.

The payoff of data democracy was almost immediate. One report discovered 2,800 certificates and degrees that students had earned but didn’t know about. Because the college receives funding from the state for each student who graduates, those unclaimed credentials added up to $5.6 million in funding, which more than paid for the entire analytics project.

But the real groundbreaking project was yet to come.

Why students fail

In 2014, the analytics team got a challenge from Ivy Tech’s then-president Thomas Snyder to apply predictive analytics to the task of reducing failure rates. Hensley had a brainstorm. Educational institutions typically categorized student cohorts demographically. What if they tackled the problem using behavioral characteristics instead? If certain behaviors could be correlated with failure, then students could be intercepted earlier in the process and helped back on track.

Over the ensuing two years, the team combed through hundreds of behavioral factors to identify those that appeared to correlate most closely to failing grades. Because the characteristics of the student population and the academic curriculum vary widely across state, the model had to be flexible enough to accommodate different criteria for different regions. Through constant iteration and testing, behavioral models were fine-tuned for each region.

Those filters yielded a list of more than 16,000 students who are at high risk of failing one or more classes during just the first two weeks of the fall 2016 term. Hundreds of volunteer staff were recruited to call the students and interview them, with the results of each conversation captured in a web form. More than 5,000 students were ultimately interviewed.

The results yielded some surprises. Researchers expected to encounter resistance to their outreach efforts, but most students were flattered to hear from the school. “They couldn’t believe an institution our size was taking such an interest in them as individuals,” Aldrich said.

They were also surprised to learn that failure was associated with a broad range of factors, many beyond the college’s direct control. They included health issues, transportation difficulties and childcare challenges. Interviews were adapted to steer students toward resources that could help.

The ultimate reward came at midterms, when the number of students posting D and F grades fell 3.3 percent from the previous fall term. It was the largest single-year decline in that metric that the college had ever recorded.

The percentage may sound small, but it translated to more than 3,100 students who were able to continue their studies rather than drop out. Even better, the number of students who registered for the following spring term jumped 5 percent, making it more likely that they would complete their studies. Higher graduation rates not only mean increased funding for the college, but benefits Indiana’s economy as a whole, since college graduates earn an average of more than 60 percent more per year than students with only a high school diploma.

And those are only the results from the first year. As the program is refined, dropout rates should continue to fall. The college recently added a new data strategist position to the chief data officer office to work on proactive applications of analytics. These data professionals are being deployed into the functional units to work closely with administrators and professors, and departments define their needs for new data sources in weekly briefings.

The goal, Hensley said, it to “ask instead of reacting.”

The emerging data-driven culture is “outcomes-driven,” Aldrich said. “We don’t do things unless we know why we’re doing them. That leads to a lot of innovation.”

Learn more about Pentaho’s data integration and business analytics solutions at

Image: Ivy Tech via Facebook