The promotion worked.

This is an authored teaching scenario, inspired by the opening of Enter the Possible. The people’s responses are written for the scenario, not generated by AI. There is no winner, learning score, or claim that these simplified outcomes predict a real restaurant.

• The restaurant begins with 24 open orders. Each fifteen-minute window brings 20 new requests.
• Your first decision sets intake at 20, 10, or 0 orders per window. It persists for the three windows; pausing does not cancel accepted orders.
• The current team completes up to 10 meals per window. In this authored run, consent to simpler dishes raises capacity to 12 from the second window onward. Relief costs $45 and raises capacity to 16 only in the third window, after arrival.
• At 6:30, one waiting customer must leave and receives an $18 refund under every option. A wider refund offer causes up to five more waiting customers to cancel in this scripted response. Cancelled orders are removed before meals are completed.
• The credit option commits $3 to each remaining customer waiting at the start of the third window. Refunds and credits are different obligations. The figures are commitments, not profit.
• Open orders = previous open orders + accepted orders − completed meals − cancellations. Declined requests were never accepted.
• Rewinding restores the decision history and reuses the same demand, capacity rules, and external update. Future loyalty, food quality, and communication time are not numerically modeled.

Real use would require checking these assumptions with the people in the situation. That is part of the point.