The article introduces the Keystroke-Level Model (KLM), a simple and effective model for predicting the time it takes an expert user to perform a task on a computer system. The model is based on counting low-level operations such as keystrokes, pointing, homing, drawing, and mental preparation, along with system response times. The authors provide heuristic rules for coding methods into these operators and validate the model through an experiment involving 1,280 user-system-task interactions across 10 systems and 14 tasks. The model's prediction error is 21% for individual tasks, demonstrating its accuracy. The article also discusses how the KLM can be used for parametric predictions, sensitivity analysis, and benchmarking, highlighting its potential role in system design. The KLM is compared to simpler models, and the authors emphasize the importance of considering user-computer performance in system design.The article introduces the Keystroke-Level Model (KLM), a simple and effective model for predicting the time it takes an expert user to perform a task on a computer system. The model is based on counting low-level operations such as keystrokes, pointing, homing, drawing, and mental preparation, along with system response times. The authors provide heuristic rules for coding methods into these operators and validate the model through an experiment involving 1,280 user-system-task interactions across 10 systems and 14 tasks. The model's prediction error is 21% for individual tasks, demonstrating its accuracy. The article also discusses how the KLM can be used for parametric predictions, sensitivity analysis, and benchmarking, highlighting its potential role in system design. The KLM is compared to simpler models, and the authors emphasize the importance of considering user-computer performance in system design.