Inspiration

Public service systems operate under real constraints — limited capacity, fluctuating demand, and operational uncertainty.
When reviewing Toronto’s shelter occupancy data, we were struck not by isolated spikes, but by how persistently close the system operates to its limits.

This inspired us to ask a deeper question:

Is the system merely busy — or structurally strained?

Rather than focusing on prediction, we chose to frame the problem as one of operational resilience and system sensitivity.


What We Built

We designed a layered analytics pipeline to transform raw administrative records into decision-support insights.

Our workflow included:

  • Cleaning and standardizing daily occupancy data
  • Constructing rolling utilization and volatility metrics
  • Measuring peak pressure using rolling 95th percentiles
  • Building an interpretable Capacity-Pressure Index
  • Simulating a +5% demand scenario to assess system resilience

The core idea was to move beyond descriptive charts and build a structured operational model.


Key Insights

Our analysis revealed:

  • Several locations operate under chronic high utilization, indicating structural strain.
  • Some sites exhibit peak fragility — stable on average but vulnerable during surges.
  • A modest 5% demand increase significantly raises the share of near-full and over-capacity days.
  • Temporary capacity loss amplifies pressure in already strained locations.

These findings suggest the system operates close to critical thresholds.


Challenges We Faced

  1. Distinguishing structural strain from temporary fluctuations
    High utilization does not always imply fragility — volatility and peak analysis were necessary.

  2. Designing an interpretable model
    We avoided black-box approaches and instead constructed a transparent composite index with clear weighting logic.

  3. Translating technical analysis into operational insight
    Our goal was not simply to analyze data, but to produce planning-relevant conclusions.


What We Learned

This project reinforced that:

  • Public service systems should be analyzed as dynamic operational systems, not static capacity reports.
  • Average utilization can understate risk — peak metrics matter.
  • Small demand shocks can have nonlinear impacts in near-capacity systems.

Most importantly, we learned how structured data modeling can support more resilient public planning decisions.

Built With

Share this project:

Updates