[NB: This is a trial of using AI to transcribe the audio interview using MS Word and cleaning up the grammar using ChatGPT. Therefore, this is not a verbatim transcript.]
Darren, I'm delighted to have you talking to my audience today. So tell me a bit about yourself, how you got started in higher education, and how you got involved with Helio Campus.
Thank you, Rodney, for inviting me to the podcast. I'm excited for this conversation. I'm a New Yorker and grew up in New York City. I moved to Virginia during my senior year of high school and attended the University of Virginia, where I studied. Throughout my career, I have always been the "data guy." It has been called different names, like data operations, when I first started. I began in the telecom industry and then transitioned into management consulting. Prior to joining higher education, I served as the director of analytics for Rosetta Stone, a company focused on language learning software. In that role, we utilized product logs to gain insights through learner analytics. This experience led me to the University of Maryland Global Campus, where I started in 2011 as an ADP in institutional effectiveness. Later, my role expanded, and I became the vice president of analytics, overseeing all data-related aspects in the Office of Analytics. Due to enrollment volatility and fiscal challenges, the University of Maryland Global Campus faced uncertainty. Working closely with senior administrators, faculty, and staff, my team and department played a crucial role in navigating the institution towards financial stability. Recognizing our impact, the Maryland Board of Regents approved the plan to spin us off into a separate company, now known as Helio Campus. I founded the company in November 2015, and since then, we have grown to serve nearly 170 clients across the country, including community colleges, regional public and private institutions, state systems, and large universities. With about 135 full-time employees and 15 part-time employees, we have overcome challenges on our journey as a tech company.
Wow, that's quite a history. It's very interesting. I didn't know that Helio Campus was one of the largest online institutions in the country, if not the world. They had a significant enrollment at one point. I didn't realize they were facing challenges similar to smaller schools.
Yes, that was during the period of 2013 to 2015. They had a large enrollment base and were pioneers in distance learning, which has now become online education
During my time at the University of Maryland Global Campus, I gained a deep appreciation for the struggles and triumphs that students face in earning their degrees. It was a humbling experience, especially coming from a more traditional educational background. I have great empathy and pride in the work we accomplished at the Global Campus.
Now, for those who may not be familiar with your products, can you explain what data analytics entails? Where does the data come from, and how do you use it for the benefit of the university?
Certainly. At Helio Campus, our flagship data analytics practice is called the student life cycle. It involves tracking students throughout their journey at the institution. We collect data from various sources, including the CRM system, the student information system, the learning management system, and the financial ERP system. These systems capture inquiries, applications, course registrations, retention data, learning interactions, and financial operations. We organize and model this data, providing self-service capabilities for the institution. The real value of the data lies in leveraging data science and analytics techniques. We perform tasks such as enrollment forecasting, yield modeling, financial aid optimization, retention modeling, risk scoring, and more. By using the data and aligning it with the institution's priorities, we can model and predict key metrics for the institution's benefit.
It sounds like you have quite widespread sources of information. How granular does it get? I mean, I'm familiar with products that we used to use in my institution that would provide feedback to the student, the course faculty member, and the advisor. Does it get down to the level of individual course performance?
Yes, we capture data at the most atomic level. We go down to the lowest granularity of data, which is the student, course section, and faculty member. We start with the lowest grain of data within the system. In this era of cheap computing power and storage, our philosophy is to obtain the raw data at the most atomic level. We dump all the data into a data lake, go through a transformation process, and then run it through the metadata layer. The metadata layer adds business logic, rules, and data integrity checks. We cache the data in memory for the business intelligence tool that we use.
Okay, I understand. I'm wondering, though, how it works if a student performs poorly on a midterm exam. Does that raise flags and notify all concerned, or is it more of an after-the-fact reporting?
It's both. We try not to rely on a single factor. We use machine learning to analyze multiple variables and determine the primary ones affecting a specific student. We have a concept called student risk scoring, where the performance of a student in a midterm may be of less concern for one student compared to another. The goal is to gauge relative risk for students and have conversations with them. Initially, when we started this at the University of Maryland Global Campus, we learned some lessons. One of them was that specific outreach, like pointing out a bad grade or a missed class, can be too intrusive and not well-received. Instead, we focus on understanding the top factors contributing to a student being at risk, such as financial variables, academic performance, or lack of engagement. We recommend prioritizing institutional resources, such as advising teams and faculty, to reach out to a select number of students based on the analysis. The approach is to have a conversation with them without knowing the exact reasons behind their situation.
Does it also provide assistance to faculty by offering templates for emails and connecting them with students or advisors? Or is that responsibility left to the faculty?
That's an important aspect. Initially, when we started this, we thought the data part would be the most challenging. However, what you mentioned about utilizing the data is now the key. It involves developing and testing intervention strategies. While we provide recommendations and best practices, the institution needs to take the data and act upon it. They should have a dedicated team focused on testing various intervention strategies and rolling them out to the institution. That's the second part of making an impact in student success. It requires a systematic approach to intervention strategies and measuring their effectiveness. This rigorous process is what I was referring to when describing the high response we received to the empathetic email outreach.
I hear more and more about connecting students, especially freshmen who often question why they are learning a specific topic. Is there a connection to industry needs and employment?
Let me talk a little bit more broadly. Our focus is not only on data analytics but also on assessment and credentialing. We aim to help institutions align themselves better for the future and take a holistic approach to planning, assessment, and decision-making. We have three main product offerings: data analytics, assessment and credentialing, and financial intelligence. Assessment and credentialing focus on measuring student learning and providing credentials that can be mapped to job opportunities. Curriculum mapping is an important part of this process, linking learning objectives, skills, and careers. However, mapping learning objectives to specific jobs is still in the early stages of development.
Certainly, if students can see the connection between their studies and industry needs, it would provide them with more motivation.
So, you know, one of the hot topics that keeps coming up in the news is the sustainability of institutions. When we consider our own experience at the University of Maryland Global Campus, as an independent institution relying on tuition revenue, we've experienced significant fluctuations in enrollments and tuition revenue. This has had a considerable impact on an institution like UGC. While flagship public selective institutions with large endowments face fewer pressures, there is a significant financial burden on a large portion of higher education.
Our financial intelligence focuses on ensuring that institutions operate efficiently, effectively, and sustainably. We offer three main products. The first is our newest product, financial modeling. It addresses the need for scenario planning, especially during the pandemic. The financial modeling software enables institutions to assess different scenarios and their impact on financial sustainability. It helps analyze net operating surplus, changes in net assets on the balance sheet, liquidity, and other factors, allowing institutions to plan beyond the next fiscal year.
We also provide a benchmarking consortium that compares administrative and academic costs, particularly related to labor expenditure. This allows institutions to perform internal benchmarking between schools or external benchmarking against peer institutions. It helps identify areas where schools may be overspending or underspending.
Lastly, our product called academic performance management focuses on the sustainability and financial aspects of academic departments and programs. It involves understanding the contribution margin, instructor humanization, curriculum subscription rates, and other financial factors.
When it comes to return on investment, one of our primary goals is to help institutions adopt a more holistic approach to planning, assessment, and decision making. Through data analytics and modeling, we can analyze admissions funnel or financial aid data, leading to increased yields, new students, and net tuition revenue. By focusing on student success, institutions can retain and graduate more students, preserving net tuition revenue, which is vital for the overall operating budget and revenues.
Well, that's a mouthful. Yeah, I can see there's a lot there. What can you give me an example of? Obviously that these schools before they contract with you there, we want to know what the return on investment is,, what are some examples of of where schools save money by using your software?
I cannot provide specific examples without naming schools, we have seen scenarios where institutions were on the wrong track. These mistakes can vary, such as poor financial planning, inadequate student retention strategies, or inefficient allocation of resources. Our aim is to help institutions avoid such pitfalls and optimize their operations.
So, um, I can provide an example that comes to mind regarding data analytics and how it helped an institution make a policy change. The institution had specific policies for awarding merit-based and need-based aid. However, our analysis showed that their merit-based aid policy was punitive towards economically disadvantaged students. They were providing aid to students who did not yield a good return on investment, as these students either did not enroll or transferred out of the institution. Meanwhile, their policies were not supportive enough for their core students who wanted to stay and graduate.
By utilizing data, we demonstrated that they needed to make a couple of changes. Firstly, they should adjust the mix of aid by decreasing merit-based aid and increasing need-based aid. This shift in strategy would be more beneficial for them. Secondly, we assisted them in refining their financial aid policies related to merit-based aid, tailoring it more towards their core students rather than aspirational profiles. This approach was driven by data and aimed to guide policy changes that would yield positive outcomes, such as attracting new students, reducing unmet financial need, and improving yield and retention rates.
Regarding your question about pricing strategy, it is indeed a significant concern for enrollment managers in higher education. Most institutions have not undergone a reset, and the average discount rates for private institutions are typically well above 50%. This means that, on average, students receive aid or scholarships that cover 50% of the tuition price. Public institutions have seen a steady increase in discount rates, which are now in the 30s or around that range.
Pricing and optimizing financial aid, which is part of our work, are crucial considerations for enrollment managers. While some institutions have implemented price resets, the majority have chosen not to do so for various reasons. Many institutions continue to increase their sticker prices annually, as there are still students who pay the full tuition amount, and that revenue is necessary. However, the rising discount rates pose challenges similar to those in other industries. A discount rate of 50% or more would typically indicate a mispricing issue. The concept of price resets has been widely discussed, and although some institutions have adopted this strategy, the majority have opted against it.
Got it. I'm much like you, a technologist, although I haven't been deeply involved in the field for a while. But I'm interested in AI and machine learning. You mentioned machine learning earlier. Could you tell me more about these new technologies and how you're using them?
Certainly! We have been using machine learning for a long time. When I was at the University of Maryland Global Campus, we initially worked with a vendor before developing our in-house capabilities. Now, we have it as a core component of our campus. AI, on the other hand, is a form of machine learning, particularly referring to large language models. Currently, what is most prominent in everyone's mind is generative AI, which involves the use of these large language models for various purposes. As a company, we are exploring how to integrate generative AI into our products and operations. We have many ideas and some ongoing pilots, but it is still a relatively new area, and we must proceed cautiously.
It's worth noting that we don't own these large language models, like ChatGPT, for example. The company that produces them, OpenAI, retains ownership of the data used to train these models. Therefore, we must be careful and not upload proprietary or confidential data, especially student data.
We are looking into obtaining "walled garden" versions of large language models that we can control and ensure perfect security. However, we are still in the early stages, both technologically and from a privacy policy perspective. Cloud services such as AWS, Azure, and Google are also developing these services. Currently, we are in the testing, piloting, and research phase, but we have long been utilizing more traditional machine learning methods.
In traditional machine learning, we ingest large amounts of data and use different algorithms to predict various outcomes. For example, in forecasting models, we use time series forecasting to estimate future enrollments with confidence intervals. For student risk scoring, we might employ random forest algorithms to calculate the relative probabilities of different outcomes and identify the most likely scenario. These are more mature models, while generative AI is relatively new to the scene.
Indeed, handing over existing functions to run on "autopilot" without careful scrutiny is not advisable. Just like with Tesla Autopilot, it's necessary to maintain hands-on involvement and oversight.
Regarding data science in higher education, it differs significantly from other fields. Predicting human behavior presents unique challenges because we have blind spots in the data. Unlike machine data that can accurately predict machine outcomes, we lack information about students' day-to-day experiences and subjective factors. For example, we may not know if a student is unhappy with their roommate. Predicting human behavior requires more caution and careful consideration than predicting machine outcomes. Therefore, our modeling philosophy emphasizes a "human in the loop" approach. We capture data, have inputs reviewed by advisors or data scientists, and incorporate additional information to influence the models. This concept of involving humans in the loop is crucial for higher education.
No, I certainly understand. I was just thinking about some of your financial inputs. Do you have any inputs regarding the economy or interest rates? Do they factor into your calculations at all?
In our case, we don't include those as direct inputs. Our focus is primarily on the students' ability to pay. When students enroll, they declare their intent and fill out the FAFSA form, providing us with information about their financial situation. Based on that, we calculate their "unmet need," which refers to the remaining cost of tuition after considering their financial aid, such as scholarships, grants, and other forms of assistance that reduce the overall cost. We assess their ability to pay based on factors like family contribution and work-study opportunities. The unmet need represents the portion they would need to cover through loans or other means. So, our approach revolves around understanding and analyzing students' unmet need and ability to pay, and how that impacts factors like retention and graduation rates through statistical analysis.
Understood. Well, it seems like we're running out of time, but before we wrap up, I'd like to give you an opportunity for some closing remarks. Are there any new features or functionalities on the horizon, or any final words you'd like to share with our potential audience?
Absolutely. Instead of focusing on specific product features, I'd like to emphasize a broader message that we're trying to convey to the market and our clients. All these elements are interconnected. It's crucial to consider academic programs, student learning outcomes, career outcomes, and the operational efficiency and sustainability of universities together. Integrated planning is key, where all these aspects are analyzed and aligned. It's important to take a holistic view and combine data to make informed planning and decision-making processes. That's the approach we're striving for, providing our clients with the tools, data, and best practices to achieve this holistic perspective.
Well, listen, I learned a lot. I think that was great. I appreciate your time with my audience and I wish you the best of luck.
Thank you very much. Appreciate your time today as well.
