Predictive Models Are Missing The Middle
- Teege Mettille
- Apr 17
- 2 min read
It’s decision season, and you’re likely looking at dashboards, probability scores, or color-coded lists telling you which students are “most likely to enroll.”
And yet, something feels off.
You’re not alone. Many enrollment leaders are realizing that traditional predictive models—especially those built around historical correlations—aren’t delivering the clarity or momentum they once promised.
There’s a reason for that.
Most of those models were designed for a different era of enrollment. An era with fewer applications, longer decision timelines, and more linear student journeys. They tend to rely on static inputs: test scores, FAFSA status, ZIP codes, or whether a student visited campus.
What they don’t capture is what’s happening right now.
They can’t tell you who just opened your email but didn’t click.Who completed their application but hasn’t logged in since.Who scheduled a call three weeks ago—and then ghosted.
These aren’t small details. They’re signals. And they matter more in April than anything a regression model said in January.
The middle of the funnel—where students are actively comparing, deciding, and shifting—isn’t linear. It’s volatile. And it's where most of your class is still in play.
Traditional predictive models weren’t built to handle that kind of dynamic behavior. They can estimate likelihood. But they don’t map the real decision-making process. They tell you who might enroll—but not how to act.
And that’s what enrollment teams need right now: not just a score, but a signal. Not just a prediction, but a plan.
The institutions that outperform this spring will be the ones that stop waiting for the funnel to play out—and start responding to what’s actually happening in it.
If you're interested in a conversation on this topic, I'd invite you to check out "Redefinining Data Driven" - an enroll ml webinar from January, 2024, where Mickey Baines, Partner at Kennedy & Company Education joined me to discuss what it means to follow the data in today's admissions environment.
