K-12 Enrollment Prediction Model Plan for the students who will actually walk in the door
The Midtown Consulting Group K-12 Enrollment Prediction Model is an AI-driven forecasting model that predicts enrollment for every school, grade and attendance zone — and now every student. Built with Duval County Public Schools, it was 99% accurate on total district enrollment in 2026.
2026 results · Duval County Public Schools
- District total accuracy
- 99%
- Within 1,661 students of actual 2026 district enrollment.
- Average school-level accuracy
- 97%
- Reliable by school, grade and attendance zone.
- Student-level accuracy
- 80% New
- Correctly predicts where individual students will enroll next year.
- Students modeled
- 130K+
- Across growing and declining neighborhoods alike.
The challenge
Enrollment doesn't stand still. Your forecast shouldn't either.
Families move mid-year. New subdivisions fill up — or don't. Charter and private options keep expanding. A forecast built only on last year's grade-to-grade counts can't see any of it, and it's frozen until next year's report.
Transfers and mid-year moves
Students change schools during the year, and annual counts miss the churn.
Year-to-year volatility
Growth in some neighborhoods, decline in others — often in the same zone.
Housing growth vs. yield
New homes don't always mean new students. Yield varies by development.
Choice and competition
Charter and private school expansion keeps reshaping where students enroll.
How we build your solution
Every force behind enrollment, in one model
We connect the data that explains each challenge — from your student information system to birth records, building permits, census and charter enrollment — and turn it into forecasts, scenarios and planning guidance your team can use.
Transfers & mid-year moves
- Student data
Year-to-year volatility
- Birth rates
- Census
Housing: growth vs. yield
- Permits
- Census
Choice & competition
- Charter data
Cohort baseline
+ machine-learning features
Model
- ForecastsK-12 by school, grade and zone
- ScenariosChoice, housing and charter share
- Planning guidanceFacilities and staffing
How the model works
A cohort baseline, sharpened by machine learning
The model blends a traditional cohort-survival baseline with an ensemble of machine-learning and time-series models. Features and models are tuned together, and the whole pipeline is refined as new data arrives.
Ingest
Births, building permits, census, student information system (SIS) and charter enrollment.
Transform
Geocode, align time windows and join by attendance zone.
Feature ↻ Iterative loop
Yield curves, mobility, choice share and housing pipeline.
Model ↻ Iterative loop
Cohort baseline plus an ML ensemble of regression and time-series models.
Publish
Forecasts, simulations and dashboards, checked through a QA loop.
See it in action
Accuracy you can check, school by school and zone by zone
Zone-level view
Click any attendance zone to compare projected and actual change
Switch between elementary, middle and high school zones. Layer in current and prior-year building permits to see where housing is driving enrollment.
School-level view
Projection vs. actual for every school type and level
Drill from the district total into regular, magnet, charter, ESE and alternative schools — and down to each school by name.
- 2025 (Year 1)
- 98% · off by 2,688
- 2026
- 99% · off by 1,661
Why it's different
Not another annual projection
- 99%
Proven in a real district
Built with and validated against Duval County Public Schools' actual enrollment, year over year.
- 1:1
Down to the student
Every total traces back to individual predictions, so you see who is driving the change.
- What if
Built for decisions
Test choice, housing and charter-share scenarios before committing facilities and staff.
| Capability | Cohort survival | MCG model |
|---|---|---|
| Uses grade-to-grade history | Yes | Yes |
| Adds births, permits, census and charter data | No | Yes |
| Updates as new data arrives | No | Yes |
| Predicts at the individual student level | No | Yes |
| Forecasts by school, grade and zone | Limited | Yes |
| Scenario modeling for choice, housing and charter share | No | Yes |
| Facilities and staffing guidance | No | Yes |
| Validated against actual enrollment each year | No | Yes |
Coming next year
Know why students leave — and who might be next
Today, a student who leaves the district entirely is a black box. Next year the model will train on historic end-of-year withdrawal codes, so districts can see the reasons behind those exits — moved out of state, enrolled in a private school, and more.
See the reason behind each exit
Withdrawal codes turn unexplained enrollment losses into patterns you can act on.
Spot students likely to leave
Predict which students are at risk of withdrawing before the school year ends.
Keep similar students enrolled
Give schools a profile of students who leave, so they can reach similar families early.
Questions we get
Frequently asked questions
How accurate is the K-12 enrollment prediction model?
In 2026 the model was 99% accurate on total district enrollment for Duval County Public Schools — within 1,661 students — up from 98% (2,688 students) in 2025. It averages 97% accuracy at the school level and 80% at the individual student level.
What does student-level enrollment prediction mean?
Instead of projecting only totals, the model forecasts where each student is likely to enroll next year. School, grade and zone forecasts are built up from those predictions, so you can trace any number back to the students behind it.
What data does the model use?
Birth records, building permits, census data, your student information system and charter enrollment — combined with signals like student mobility, school choice share and the housing pipeline.
How is this different from a cohort-survival forecast?
Cohort survival projects next year from grade-to-grade history alone. This model starts from that baseline, then adds machine learning and real-world signals, and updates as new data arrives instead of once a year.
Can we model enrollment scenarios?
Yes. Simulations let planners test changes — a new development, a boundary shift, a new charter school — before making the decision.
Ready to plan your district's next decade?
Talk with the team that built the Duval County Public Schools model. We'll walk through your data and what a forecast could look like for your schools.
Talk to our teamLast updated
A LUMINeRA Solution
