How A Preschool Operator Lifted Labor Productivity 50%
A PE-backed preschool ops team trained an ML model on attendance to staff 100+ sites hour by hour — lifting labor productivity 50% and expanding EBITDA margins 33%.
50%+
Lifted across 100+ preschool sites
The Challenge
A private equity-backed education platform operating 100+ preschool sites had grown through M&A, leaving disparate data systems that could not communicate with one another. Leadership had no real-time visibility into staffing levels across sites. Determining which sites were over-staffed required hours of manual data pulls and Excel analysis — preventing tight labor management in an industry where labor is the only significant variable cost.
What They Built
SaxeCap integrated data from dozens of discrete systems across the business into a unified platform. They built a machine learning model trained on historical student attendance data to predict, hour-by-hour, how many students of each age level would be in each classroom. That output fed into an optimization model that recommended staffing adjustments to maximize site-level EBITDA while meeting all state-mandated student-to-teacher ratios. Recommendations were pushed to site principals in real time; principals could accept, override, or provide feedback that trained the model. The entire system ran continuously with no generative AI — pure classical ML and operations research.
SaxeCap began by solving the data unification problem that had made real-time staffing visibility impossible: integrating data from dozens of discrete, incompatible systems accumulated through M&A into a single unified platform. This foundation was non-negotiable — without it, any optimization model would be working with incomplete and unreliable inputs.
With clean, unified data, SaxeCap trained a machine learning model on historical student attendance patterns to predict, at an hourly level, how many students of each age group would be present in each classroom across every site. Because state regulations mandate specific student-to-teacher ratios by age level, the prediction granularity had to match the regulatory structure exactly. The attendance predictions fed directly into an operations research optimization model that calculated the minimum staffing configuration needed to meet all regulatory requirements while maximizing site-level EBITDA. Recommendations were pushed to site principals in real time. Principals could accept recommendations, override them with local context, or flag errors — each interaction becoming feedback that continuously improved model accuracy. The entire system ran on classical ML and operations research, with no generative AI involved, delivering 50%+ labor productivity gains and 33%+ EBITDA margin expansion.
AI Role
EBITDA margins expanded by more than one-third (33%+) — transformative for an industry where labor is effectively the only variable cost.
Infrastructure
Proprietary labor optimization platform (unified data and recommendation layer) • Machine learning attendance prediction model (classical ML, historical training data) • Operations research optimization model (staffing constraint solver) • Site principal interface (real-time recommendations and override/feedback loop)
Integration Points
Dozens of discrete site-level systems integrated into centralized data platform • Historical attendance data pipeline feeding ML training and inference layers • Optimization model connected to attendance predictions and state regulatory ratio constraints • Real-time recommendation output pushed to site principals with feedback collection loop
Impact
Labor productivity at the preschool platform increased by over 50% following deployment of the AI-driven staffing optimization system.
EBITDA margins expanded by more than one-third (33%+) — transformative for an industry where labor is effectively the only variable cost.
SaxeCap's standard approach delivers demonstrable EBITDA uplift within approximately two months of engagement start, including diligence, analysis, and first system deployment.
Implementation Complexity
Best Fit For
Private equity funds ($3B+ AUM) and their portfolio companies ($20M–hundreds of millions in annual profit); services businesses (field services, healthcare, education, business services) with human-capital-intensive workflows; companies with proprietary data assets and opportunities to automate manual, rule-bound processes.
