Roast & reward: An AI reinforcement learning simulation game

Publication Type

Case

Publication Date

6-2026

Abstract

Stepping into the role of a coffee shop owner, the player makes a simple yet critical pricing decision to maximise the shop’s rewards. Each day begins with this question: What should the price of a cup of coffee be today?

Bear in mind that demand for coffee shifts with changing conditions, such as the weather (sunny, cloudy, or rainy), nearby events that drive foot traffic, whether a competitor’s shop is open, and the competitor’s price. Setting a price too low may reduce profits and deplete inventory quickly, whereas setting it too high may turn customers away and lead to excess stock.

As the week unfolds, the player learns through trial and error by gathering feedback on the number of cups of coffee sold, revenue, and profit. Unsold inventory at the end of the week also incurs a penalty. Therefore, the challenge lies in balancing sales to minimise wastage while avoiding an early sellout in the week.

The game begins with a one-week orientation, followed by the actual 21-day simulation. At the end, the player’s performance is compared with that of reinforcement learning (RL) and machine learning (ML) agents running in the background. Success in this game comes from observing patterns, experimenting with prices, reflecting on the outcomes of daily decisions, and learning which price strategy/policy works under different conditions.

This single-player game is designed as an experiential introduction to RL. Students will learn to explain key RL concepts, how RL learns through repeated interaction with its environment, and differentiate RL from other types of ML. By the end of the game, they will be able to explain why context matters when making decisions and differentiate between exploration and exploitation in refining a strategy. They will also learn how rewards and penalties provide feedback that shapes future behaviour.

Keyword(s)

exploitation, market conditions, competition, demand analysis, tradeoff analysis, sales forecasting

Discipline

Artificial Intelligence and Robotics | Business Administration, Management, and Operations

Area of Excellence

Digital transformation

Research Areas

Marketing

Data Source

Field Research

Industry

Beverage industry

Geographic Coverage

Middle East

Temporal Coverage

2026

Education Level

Executive Education; Postgraduate; Undergraduate

Publisher

Singapore Management University

Case ID

SMU-26-0007

Comments

For purchase of the case and supplementary materials via The CMP Shop, please access the following link:

The links to purchase the case and supplementary materials on The Case Centre and Harvard Business Publishing is available via The CMP Shop.

SMU Faculty/Staff can download the case and supplementary materials on iNet with your SMU login ID and Password via The CMP Shop

Additional URL

https://ccx-shop.smu.edu.sg/products/roast-reward-an-ai-reinforcement-learning-simulation-game?variant=44510005067818

Share

COinS