Textbook of Usability

Predictive modelling in usability refers to methods that predict aspects of user performance (task time, error rates, cognitive load) using quantitative models rather than empirical testing with human participants. It answers the question "which design is likely to be faster, more accurate, or less demanding?" through calculation rather than experiment.

The main predictive tools:

• Fitts's Law — predicts pointing time for target acquisition
• Hick's Law — predicts decision time for choices
• Keystroke-Level Model (KLM) — predicts expert task time by summing operators
• GOMS variants — model goal hierarchies and selection rules
• CPM-GOMS — models parallel perceptual, cognitive, and motor operations
• Cognitive load estimation — counts working memory items per step
• ACT-R — full computational cognitive architecture for detailed predictions
• Cogulator and CogTool — practical tools that automate predictive modelling

Predictive modelling is most valuable:

• Early in design, before prototypes exist, for comparing alternatives
• For routine expert tasks where the model assumptions hold
• For identifying bottlenecks in workflows
• For justifying design decisions with quantitative evidence

The limitations are equally clear: models predict specific aspects of expert, error-free performance. They do not predict satisfaction, learnability, novel exploration, or the full range of human variability. They should be combined with expert review and user testing for comprehensive evaluation.

The appeal is clear: predictions can be made without writing code, building prototypes, or recruiting participants — making them among the cheapest forms of usability evaluation.