Your Enrollment Model Already Knows Which Students to Target. The Question Is What You Send Them.
Your enrollment model already knows which students are persuadable. It scored them, ranked them, and flagged the 20 to 35% of your admitted class that could go either way. Then your institution sent those students the same drip sequence as everyone else.
The movable middle is not a mystery. Nearly every institution with a modern CRM has the analytics to identify it. Technolutions Slate holds approximately 55% of the higher education CRM market, Element451 roughly 25%, and Salesforce about 20%. Predictive scoring is a standard feature across all three. The question is not whether your model can find the students who could tip either way. The question is what you do with the intelligence your model gives you.
What the Movable Middle Actually Is
The movable middle refers to admitted students with a 30 to 60% predicted probability of enrolling. They are distinct from the students at the top of your probability distribution (who will likely enroll regardless of what you send them) and the students at the bottom (who are difficult to convert no matter what outreach they receive). The movable middle is the genuinely persuadable segment, the group where additional outreach actually changes outcomes.
At most institutions, the movable middle represents 20 to 35% of the admitted class. But it accounts for the majority of year-to-year yield variance. When enrollment goes up, it is usually because the movable middle converted at a higher rate. When enrollment drops, it is usually because this segment chose somewhere else.
EAB research shows that earlier, personalized outreach to high-fit students more than doubles the likelihood of a deposit compared to standard communication flows. The data is clear: personalization works on this segment. The problem is that most institutions define personalization as inserting a first name into a template.
Case Studies in Model-Activated Recruitment
The institutions seeing the largest yield gains are not the ones with the best models. They are the ones connecting model output to differentiated action.
Texas Tech University partnered with Othot (now Encoura) to identify admitted students most likely to enroll if they attended the Raider Roadshow yield event. Instead of blasting invitations to the entire admitted pool, they targeted outreach to the model-identified movable middle. The result: a 31% increase in event attendance, enrollment exceeding their goal by 322 students, and retention growing by 2.6% over three years.
Western Connecticut State University used Othot predictive modeling to shift from intuition-based recruitment to data-driven personalization. The result: a 20.7% increase in first-year enrollment from 2023 to 2024, generating more than $2 million in additional net tuition revenue.
In both cases, the predictive model was not the innovation. Every peer institution had access to similar analytics. The innovation was connecting model output to specific, differentiated interventions: targeted event invitations for Texas Tech, personalized communication sequences for WCSU. The model said “these students are persuadable” and the institution did something different for those students than for everyone else.
Why Most Schools Fail to Close the Loop
If the data is this clear, why do most institutions still route their movable middle through the same generic sequences? Three operational gaps explain most of the disconnect.
The first is a timing problem. A NACAC study found that a delay of more than 48 hours in responding to a student inquiry drops enrollment likelihood by 30%. Yet most institutions have no automated triggers connecting predictive model outputs to personalized outreach sequences. The model identifies a student as persuadable, but the communication plan runs on a fixed calendar, not on model-triggered workflows.
The second is a process problem. A 2024 EAB analysis found that institutions focusing solely on top-of-funnel lead growth without aligning downstream communication processes often see no net enrollment gain. Bottlenecks in transcript evaluation, financial aid response time, or application completion cancel out whatever yield improvements the model-driven outreach might produce.
The third is a demographic problem that makes the first two unforgivable. WICHE projects a 15% decline in high school graduates between 2025 and 2030. Institutions cannot grow their way out of yield problems by adding more names to the top of the funnel. Every admitted student matters more than they did five years ago, and the cost of failing to convert a movable-middle student rises each year.
The CRM has the scoring capability built in. The outreach workflows connected to those scores are the bottleneck. The model says “this student needs personal attention” and the system delivers “Dear [First Name], we are excited you were admitted.”
The Personalization Gap and the Physical Channel
Digital personalization helps but is increasingly commoditized. Every institution with Slate or Element451 can send personalized emails with dynamic content blocks, targeted text messages, and customized portal experiences. When every school is doing the same version of digital personalization, none of it feels personal anymore.
Physical, handwritten outreach represents a channel that is inherently personal and dramatically underutilized in admissions. Direct mail has a 95% engagement rate and is interacted with four or more times on average. A handwritten note from a department chair or current student, triggered by a predictive model flagging a student as movable-middle, is execution that matches the sophistication of the analytics powering it.
There is an equity dimension here as well. New America and Brookings research has documented that predictive enrollment models can inadvertently disadvantage students of color and first-generation students when they rely heavily on “demonstrated interest” signals like campus visits, email opens, and event attendance. These signals correlate with family income and proximity to campus. Physical outreach that does not depend on whether a student could afford to visit campus, that arrives because the model identified academic fit rather than engagement behavior, can help mitigate this bias.
The schools winning the movable middle are not the ones with better models. They are the ones connecting model output to genuinely personal outreach that the student has not received from four other institutions that week. The distinction between genuine personalization and template-based personalization is the difference between a yield strategy that works and one that just feels modern.
Your model already solved the identification problem. The question is whether your communication plan matches the intelligence your data provides.
For teams ready to implement a personal outreach strategy, our complete guide to handwritten letters in business covers the practical details from message length to paper selection.
FAQ
What is the movable middle in college admissions?
The movable middle refers to admitted students with a 30 to 60% predicted probability of enrolling. These are students who are genuinely persuadable but have not committed. They are distinct from high-probability students who will likely enroll regardless and low-probability students who are difficult to convert. The movable middle typically represents 20 to 35% of an admitted class but accounts for the majority of year-to-year yield variance, making it the highest-leverage segment for enrollment teams to target.
How do predictive enrollment models work in admissions?
Predictive enrollment models analyze hundreds of data points, including academic profile, geographic location, financial need, engagement behaviors, and demographic indicators, to generate an enrollment probability score for each admitted student. Platforms like Othot (Encoura), Capture Higher Ed, EAB, and CRM-embedded tools (Slate, Element451) run these models continuously across the enrollment cycle, refining predictions as new behavioral data comes in. The models identify which students are most likely to respond to additional outreach, enabling institutions to allocate recruitment resources more efficiently.
What is the most effective way to convert movable-middle students?
Research and case studies consistently show that personalized, timely outreach is the most effective strategy. Texas Tech University used Othot predictions to target event invitations to model-identified persuadable students, resulting in a 31% attendance increase and exceeding enrollment goals by 322 students. EAB data shows that earlier personalized outreach more than doubles deposit likelihood. The key is connecting predictive model output to differentiated communication workflows rather than routing all students through the same generic sequences.