# Choosing a campus transport pilot

Original fictional teaching scenario for this workshop. All campus details, proposals, and observations are invented. Use this text as a closed teaching source, not as evidence about a real university or as published research.

## The decision
Northbridge College has a fictional budget of $30,000 for one term's transport pilot. Its committee must recommend one of two proposals and explain what evidence would be needed to evaluate the pilot. The money can fund either proposal in this exercise; there is no supplied cost breakdown.

**Proposal A: evening shuttle.** A shuttle would connect campus with a nearby transit hub on weekday evenings. The proposal prioritizes students attending evening classes. No tested journey-time estimate, demand forecast, or measured emissions effect has been provided.

**Proposal B: cycling support.** Secure cycle parking and a basic repair station would be added near the main teaching building. The proposal prioritizes people who can and want to cycle. The committee has not been given seasonal participation estimates or evidence about who would change their travel mode.

## The supplied data
The accompanying sample-data.csv contains 16 synthetic commute records across four travel modes. The records have trip_id, mode, distance_km, commute_minutes, and cost_cad fields. Cost is an invented out-of-pocket cost for one trip. A zero in the cost field does not mean the mode has no wider costs. Distance is one-way; commute_minutes is the total trip time. The records are a teaching dataset, not a survey of a campus population.

## Reasoning with limits
A group mean describes these records. It does not show that switching a person to another mode would produce that mean travel time. Distance, route, service availability, weather, mobility, caring responsibilities, and scheduling may matter, but this dataset does not measure all of them. Mode counts do not establish campus-wide demand. A small synthetic dataset cannot estimate actual intervention effects.

A defensible recommendation can prioritize a clear goal, use the supplied evidence with appropriate limits, explain a competing consideration, and request the additional evidence needed. Different recommendations can be reasonable if their reasoning is supported. Neither proposal is designated as the universally correct answer.

## Discussion task
Recommend one proposal for a pilot. Use at least two relevant details from the scenario or dataset, explain one tradeoff, name one limitation, and propose one way to evaluate whether the pilot helps its intended users.