Identifying individual students who are at-risk of dropping out is often a subjective exercise in higher education. Additionally, for each student the institution intervenes with, we must answer the question of why are we intervening? And what resources will we provide this student? It is critical to increase retention and graduation rates, while remaining conscious of intervention resources, budgets and university revenues.
Bringing rigor to the process via predictive analytics and student segmentation will not only give you more accurate results, but also identify these individuals with a longer lead time to plan and execute interventions. When you have identified students that are at-risk of dropping out, how do you now interact with them and how do you successfully intervene?
Not all at-risk students show the same behaviors, and not all are at-risk for the same reasons. We have identified six steps that will help you to work through the process of grouping students in to similar segments for cohesive intervention planning.
1. At-risk Student Identification — The first step is identifying individual at-risk students in the first place. The SightLine strategy is to use a previously developed predictive model that flags students for being at-risk with a 1–2 semester lead time, providing enough time to meaningfully intervene and retain these students. This is just one example of a method to identify at-risk students, your institution could use a method as simple as intervening with students below a 2.0 term GPA. Although, a more analytically rigorous method is recommended for improved at-risk student identification.
As an example of actual findings from a predictive modeling process, we identified several primary risk factors for students at a Midwest University. Here are a few of the key findings, out of many factors that were analyzed.
Students with fewer than 15 total term credits have a higher average risk of dropping out within the next semester.
Being employed on campus is hugely beneficial, even for students who historically have struggled academically.
Participation in certain student activities or clubs is better than others.
Financial factors play a large role in drop-out risk for certain student segments.
Additionally, we identified which overall and term GPA thresholds put students at higher risk of dropping out, dependent on how long the student had been attending the university.
After applying the predictive model to each student and assessing the risk factors above (along with other factors), we identified individual students who are at-risk for leaving the university.
2. Intervention Types — As stated above, not all at-risk students are considered at-risk for the same reason and they should not all receive the same type of intervention. Some students may benefit from conversations with their academic advisors, attending extra sessions at learning centers, or meeting individually with a success coach. Others need to address financial literacy challenges or find campus employment; additionally some students may qualify for and would benefit from additional scholarships.
3. Segmenting At-Risk Students into Similar Groups — After creating a list of the individual students who are at-risk for dropping out, it is critical to determine the best path towards intervention. For each student that the institution intervenes with, we must answer the question of why are we intervening? And what resources will we provide this student? Rather than providing a completely individualized intervention for each student, we use a hierarchical segmentation/clustering algorithm to find groups of students that behave similarly, within the total at-risk student population.
In other words, we use a mathematical formula to measure the distance between all students based on qualities such as academic progress and standing, university engagement, employment, scholarships and grants, financial need and socioeconomic standing. We use this distance calculation to find students who are close together and classify them as a single student group. The students within each segment/cluster/group, have similar qualities and therefore may be candidates for the same intervention strategy. We can then develop 10 or so intervention and messaging strategies, rather than more than 500 individual interventions.
Below is an example of three different clusters or segments of students. Each group is differentiated from the others based on academic progress, need-based financial aid, and academic performance and each group should receive a different intervention approach.
4. Applying the Intervention and Messaging — To effectively apply intervention methodologies, a strategy for both the intervention itself and, more importantly, the messaging to the student is critical. For the client case study, we are discussing, we identified 10 homogeneous groups within the overall at-risk student population. This became the foundation for institutional discussion around applying appropriate interventions and how to approach the student. Such approaches will vary by student group as shown in the two following examples.
Segment A: One of the groups in our example consisted primarily of first year students, receiving Pell grants who had low academic performance during their first semester and did not participate in any on campus activities or clubs. We would recommend speaking with this group of students about whether they are looking for employment and inform them of on-campus job opportunities. This student segment may be employed off campus, due to their financial need and this may be evident by their lack of on-campus participation. Being employed on campus provides flexibility of work schedules that are conscious of academic priorities, opportunities for work experience in their academic field, and engagement with campus advisors and faculty.
Segment B: Another group from our case study, had the highest need-based institutional scholarships, were receiving Pell grants, and were the highest performing academically across our at-risk student population. These students are highly driven and motivated academically, with high financial need. If funds are available for additional scholarships to boost students on to graduation, this group of students should be candidates. We would still assess each student individually for scholarship eligibility but working with students in this segment is a good start and will help apply scholarships to students with the highest impact on retention outcomes.
It is important to emphasize that specific messaging and even the tone of messaging should vary by intervention strategy. For employment nudges and additional scholarships, the focus would be on the positive opportunity, stating the ways these opportunities would support the student. For other student segments that are struggling academically, the goal may be to meet individually with academic advisors, in which case the messaging would be much more individualized with a supportive tone.
5. Key Project Stakeholders — The question of who should be involved in the identification of, and assistance with, at-risk students may vary depending on your unique institution. There may be many moving parts to this type of project considering the offices that may be involved and individual advisory outreach. We recommend that this be a part of a larger strategic conversation between key stakeholders of resources across campus. This would include the Dean of Students, the Student Success Office, Learning Centers, Academic Advisors, the Academic Affairs Office and Deans of schools or colleges. It is important to discuss a centralized versus a de-centralized approach for student interventions. Individual offices or academic departments may have different resources at their disposal, and it may be beneficial to have each implement specific intervention methods.
6. Evaluation of Efforts –It will be difficult to know whether intervention efforts were fruitful and if valuable intervention resources are being allocated appropriately without evaluating outcomes. We recommend two different approaches to evaluating intervention efforts.
Quantitative Metrics — Possible metrics to track may include; response rates from intervention emails, whether advisory meetings or coaching sessions were attended, did employment rates go up for students struggling financially, and whether students adjusted course loads when prompted. Over time, this process will help the intervention team adjust resource allocation and determine which interventions are useful, which are difficult to implement, and which interventions students are resistant to.
Student Feedback — Students are surveyed incessantly, which is why we do not recommend university wide surveys for targeting at-risk students. If, on the other hand, students have received support through campus employment or scholarships, they may be much more responsive to questions regarding their satisfaction and whether or not the intervention made a difference to them.
Adopting a data-based at-risk student identification, student segmentation, and intervention process provides an opportunity to reduce data and knowledge silos across the institution. By quantifying and evaluating efforts, the institution may share metrics across various intervention resource centers and key stakeholders. This provides the basis to continue conversation around improving the intervention process and understanding where students may be hitting roadblocks.
SightLine provides an individualized approach to identifying groups of at-risk students and tailored intervention applications through our concierge analytics. Contact SightLine to learn more about these methods and share if you found this article interesting and if you agree with our methods. SightLine believes in sharing best data practices and the math under the hood.
Originally published at https://sightlinedata.com.