Academics
  /  
Undergraduate Program
  /  
Client Project Challenge
  /  
Project
Optimizing Dog Volunteering Shift Organization

Optimizing Dog Volunteering Shift Organization — Evanston Animal Shelter

Picture of a dog at EAS

September 25, 2025

How do you run a three‑hour volunteer shift with 14–20 dogs and 5–12 volunteers in a new facility full of doors and blind spots—while keeping dogs from crossing paths? The team paired a clear playbook with a mixed‑integer optimization model that outputs a safe, efficient schedule volunteers can follow.

Problem Overview

Evanston Animal Shelter (EAS) moved into a new building with four separate rooms and more blind spots.

Map of EAS 

Shift leaders (Kennel Captains) want dogs out of kennels as much as possible, but no two dogs can occupy the same confined space and everyone must avoid surprise encounters in back‑of‑house corridors. With volunteer counts and dog counts changing each shift, KCs face a daily trade‑off between safety, task completion, and meaningful dog‑volunteer time—exactly the kind of “too much to juggle, too little time” problem many organizations recognize.

The solution:

  • Two deliverables: a Python optimization model that produces a per‑volunteer schedule (and Gantt charts/Excel output) and a poster with plain‑language recommendations KCs can use immediately.
  • How it works: the model resembles a Job-Shop Scheduling formulation—walking and cleaning are “tasks,” areas of the shelter act like “machines,” and the objective minimizes total shift completion time so volunteers can spend more time with the dogs. It encodes safety (only one dog in a confined space), sequencing, and “one task at a time per volunteer” constraints, then produces an optimal shift schedule.
  • Technology: Python + Gurobi (academic/non‑profit license), with Excel outputs and Gantt visualizations for KCs.

The simple analysis (quick wins)

  • Keep Segment 1 moving. When the time in the first corridor (Segment 1) increases to its observed extreme, the total walk time can grow by ~34 minutes—so fast passage there prevents backups and reduces risk in a known hotspot.
  • Walk “adoptable” dogs first. Model outputs favor walking adoptables before back‑of‑house dogs, which reduces overlaps and makes cleaning more efficient.
  • Do (almost) all cleaning during first walks. With typical volunteer counts, cleaning naturally fits in round one. If the shift has only five volunteers, save 2–3 kennels for the second round so cleaning finishes before dogs return.
  • Be surgical about the second walk. The team computed the maximum feasible duration for round‑two walks to finish the shift in three hours; in extreme cases (e.g., 20 dogs / 5 volunteers) the second walk should be trimmed to about 7 minutes—or skipped in favor of safety and sanity.
  • In low‑volunteer, high‑dog scenarios, second walks must be brief or skipped; in more balanced shifts, 2–3 minutes trimmed from round‑two walks can keep the day on time without sacrificing safety.

Quote:

“Gayoung, Pedro, Rachel, and Tomas did fantastic work. I am impressed with everything they achieved in only 10 weeks and the quality of their write‑ups and presentation.

The students were very proactive in asking questions and finding out information about the situation at the shelter, what is a good solution, and what type of software to use so that the shelter can use their code later if needed.”
— Daniela Hurtado-Lange, volunteer, project lead, and professor at Kellogg

Team

EAS Team: Gayoung Kim, Pedro Lacombe Farina, Rachel Silverman, Tomas Serna.
Faculty advisor: Prof. Klabjan.
Report date: June 10, 2025.