That means Reinoehl's team has a big concern for the quality of student life. Retaining students is a core part of his mission, and retention remains a core problem for most universities. It's not just the obligation to ensure a student doesn't slip through the cracks - retention also has financial ramifications.
I recently talked with Reinoehl and his colleague, Cilla Shindell, Director of Media Relations about the problems universities face recruiting and guiding students into successful graduates. Reinoehl shared the data challenges his team has faced, and how Alteryx has helped him to improve retention rates.
Struggling with data and disparate systems
When I was a college grad myself. I worked in college admissions. The technology was archaic. So what does it take to support 8,500 undergraduates and 2,600 graduate students today? Reinoehl, who has worked at the University of Dayton since 2004, said the first big technical change happened in 2010, when they moved to an ERP system. Prior to that, they had a number of home grown, one-off systems:
Honestly, I don't think we were very sophisticated at all in our use of data. The systems we had couldn't begin to do that.
Their ERP install gave Reinoehl a view of what was possible with better data:
The ERP system sits on top of Oracle. One of the advantages to that is we could use SQL's query language to extract data... Operationally that was a key moment, just to be able to use the data that we were collecting.
College recruitment faces market pressures
Today's predicaments require a different level of data analysis. Reinoehl gave me three reasons college recruitment is more challenging today:
- Fewer eligible students - the number of traditional aged high school graduates is declining, especially in the midwest and especially in Ohio. "There's just not as many prospective students available," says Reinoehl.
- Affordability - Median incomes have remained relatively flat since the depression in 2008. Meantime, university operation costs are rising (utilities, salaries, etc). "It's becoming more and more challenging to maintain affordable offerings."
- Competition with public/subsidized universities - "As private education becomes more expensive, we're competing more and more with the public subsidized institutions from a pricing perspective."
Competing with data - a retention scenario
These market pressures have instigated a focus on data, with retention being a prime example. Reinoehl and I discussed retention from an analytics-and-results perspective. It comes down to two things:
- Build a risk assessment model that incorporates all the variables that might impact groups of students. Combine a quantative model with a qualitative outreach and/or monitoring of individual students. On the quantative side, Reinoehl talked about how 20-25 retention variables can be taken into account that divides the student body into high retention risk groupings, e.g. low income, undeclared majors, distance from home. Then you take pro-active action with those students. The qualitative side could involve working with deans and professors to ensure alarming/concerning interactions with students are noted and acted upon.
- Develop and automate "positive intervention" steps that can mitigate problems. One scenario Reinoehl gave: the dreaded killer combination of courses: "If a student takes three very difficult courses at the same time versus maybe just scheduling one, that has a dramatic impact on retention." Warning students about difficult course combinations can make a difference.
The need for data blending - and discovering Alteryx
These objectives led Reinoehl's team to start working with Tableau for data visualization. But underlying problems nagged. Two years ago, at a Tableau user event, Reinoehl discovered Alteryx:
When I saw the Alteryx tool, I was sold within literally 15 minutes, because the person at the conference demonstrated the ability to easily blend multiple data sources.
And why was "data blending" so important?
Our data files were becoming so big that it was really time-consuming to try to do something in Tableau, and find out it wouldn't work. We were spending all the time blending the data. Alteryx is just phenomenal in its ability to create files very quickly, to enable me to filter on the fly, and to do it all visually.
Things moved quickly from trial to go-live:
Basically, they sent me a test instance. I was able to start playing with it right then. I went live literally thirty days later.
Using Alteryx, Reinoehl's team connects directly the operational store that pulls in all the data from transactional systems on campus, including their ERP system. That data is quickly "blended" with any other relevant data source:
I'm able just to pull it straight from the data store, and then I can blend in flat files, CSV files or other data files, and then ultimately I push it into Tableau to do the visualization.
From back end chores to useful analysis
It's about shifting time from back end data chores to useful analysis. Example: a recent predictive challenge around predicting housing demand:
I had all the enrollment data. The housing data sits in a different system. The operations team in housing just sent me a CSV file. Literally I could, in thirty seconds, merge that into the master enrollment file which has thousands of records. If I had to do that in an Excel file or import it into a SQL query language tool and create a table - all those other mechanisms take forever. This allows me to do it literally within a minute, and then start to draw insights. That's what I like about Alteryx - it takes a big chunk of work and makes it a lot simpler, so that we can spend our time thinking through the problems and how we ought to approach solving them.
Another perk? Reduced IT dependency. He can blend in data and create new reports without throwing requirements over the wall. Reinoehl speaks highly of his IT team, but he doesn't want to go knocking every time he needs a new report either:
Reporting is becoming more complex. There's always the time-consuming process of trying to translate for our IT team what we really need from a report. They don't have the end user functionality expertise. This gives me the ability to create dynamic reporting. I know exactly what I want. It just cuts down on all that back and forth with our IT team.
Using data to make the retention business case
And as for that 3 percent increase in retention? Reinoehl credits the business case he made for additional resources based on the data:
I think we were able to achieve a three percent increase largely because I was able to create a conceptualization of what our problem was through data. Then what we were able to do is ask for greater resources around our retention efforts.
Long story short, that resource bump included investment by the school deans into staff who are directly involved in student success outreach efforts. Reinoehl identified nuances in retention in each school, and proposed a strategy accordingly, which included addressing that "killer course combination" problem:
I think that whole system was necessary to see our three percent increase. I made a case for that based on doing the analysis of all the data sources we had in Alteryx. That's where this came together.
What's next - predicting student outcomes
So what's next? Reinoehl wants to take the predictive side of this further, tying their results into student success:
Retention is a nice thing to move. We're proud of that. The more exciting one would be if we can start to drive six year, or four and six year graduation rates higher, and then start to be able to predict outcomes as well, based on the desired outcome a student has - maybe a placement in a service organization or a graduate school.
That could mean looking at percentage acceptance rates at different graduate schools, and discovering patterns or anomalies. And: building in triggers and contextual actions, such as alerting a student about a needed prerequisite course, and "link into the course registration right then, so they can take the action. That's where this comes together in a powerful way." But they aren't across that finish line yet:
We're not quite there yet. I think that's the next evolution is predicting those outcomes.
Sounds like a worthy goal to pursue.
Updated October 19, 5:30 am ET, with the final quote at the end that is important to building in actions based on the data.