For instructors — what this teaches and how to run it
The three findings, in order
Variability on its own barely moves a finish date. Good weeks cancel bad weeks.
Dependency on its own costs nothing when everyone is reliable.
Put them together and the losses stop cancelling, because idle crew time cannot be recovered while surplus work just queues up. That gap is the whole lesson.
What the challenges are really testing
Level 1: reliability beats horsepower. Buying capacity when the average is already adequate is wasted money.
Level 2: throughput is capped by the weakest average in the chain. Boosting four crews and leaving one behind changes nothing at all — it scores worse than buying nothing.
Scoring
Scores come from 200 simulated futures, not the single run students watch. A good decision can still get an unlucky run, and students should see that distinction explicitly.
Everyone gets identical futures, so scores are directly comparable. Collect them for a class leaderboard.
Debrief questions
Which crew lost the most capacity, and was it the slowest or just the furthest down the chain?
Your schedule software gives one duration per activity. After this, how would you present a finish date to a client?
Where is the "queue of waiting work" on a real project, and who is paying for it?
Source
Mechanic adapted from the Parade of Trades game: Tommelein, Riley & Howell (1999), ASCE JCEM 125(5); Choo & Tommelein (1999), UC Berkeley P2SL.