Skip to content

Calculating Usage Profiles

A usage profile is a description of all possible usage situations (states) and all expected usage scenarios (transitions) for a specific user (usage) group, e.g. web store customers, car drivers or classes of medical personell. In this way, it is easy to describe different classes of users who use the system in different ways, e.g., visit websites in different ways. Profiles are represented by different assignments of transition probabilities in a statistical Markov chain usage model. These values can be computed in different ways:
  • Uninformed Strategy

This is the simplest but least accurate strategy for creating usage profiles. All outgoing arcs of a state within the usage profile are assigned equal probability values. While this approach does not require any information about system usage, it can be used when such information is not available.
  • Informed Strategy

Informed probability generation requires the collection of field data, which is then mapped to a usage profile. This method is the most accurate way to obtain information about system usage and is useful when updating/extending existing systems. Field data can be derived from direct logging and monitoring of user interaction usage, or from screen recording, and can then be analyzed or mapped directly to the model structure.
  • Intended Strategy

Probabilities are assigned according to the intended use of the system and are highly dependent on the quality of the information sources. Development teams often have a variety of information sources for the required customer specification. In addition, application specialists, customer support, and application trainers usually know the different customers and their specific ways of working very well.
  • Constraint-based Strategy

Since customer experts cannot be expected to have expert knowledge of statistics and Markov chain theory, the information gathering process must be as informal as possible. This means that customer experts are allowed to describe their knowledge in terms of constraints. Instead of assigning fixed numbers to transition probabilities, they must postulate relationships between them. These constraints can then be automatically transformed into fixed probabilities using algebraic techniques.