Even that relies on a model of the utilization of the cafeteria over time. If Goldman wanted to incentivize spending a minimal amount of time in the cafeteria, they should just measure that. For example, offer a discount of (25 - X)% where X is the number of minutes between entering the cafeteria and paying for your food. People would learn to anticipate events such as backed-up food or cash register lines and adjust their behaviour appropriately.