User-Reality fit is the functional correspondence between the user’s knowledge of the relevant set of acts and information cues that could achieve the task goals (e.g. a sales forecast) and the actions supported and information available in the organisational and business process context of the user’s work (the substantive-task/real-world task environment). It is the user’s potential for substantive task performance, unaided by technology.

User-Representation fit pertains to the user’s potential performance in the modelling task (e.g. producing a sales forecast from a computer-based model). It is the degree of functional correspondence between the user’s knowledge of the set of acts (e.g. model manipulation and analysis) and information cues (variables) that could achieve the desired goal (e.g. a computer based sales forecast and associated model). The task-environment here is the modelling task/technology-as-representation (e.g. the variables in the model and the analysis methods supported by the model’s formulation).

User-Tool fit pertains to the user’s potential performance in the operational task. It is a measure of the functional knowledge the user has of the acts and information cues supported by the technology-as-tool in carrying out operational tasks (i.e. a correspondence between what is known by the user and what is supported by the tool). Such knowledge, for example, could include the commands to execute data and model manipulation tasks in the sales forecasting scenario).

Representation-Reality fit is the degree of functional correspondence between the technology as representation’s knowledge of relevant acts and information cues relative to the required acts and information cues of the substantive task (e.g. sales forecasting), in reality. It is the system’s potential for performance, assuming it is effectively used. For example, in a decision support system, representation-reality fit is a measure of how well the model embedded in the decision support system functionally approximates the real world decision environment.

Tool-Representation fit is the tool’s potential performance in manipulating the representation (i.e. the modelling-representation task-environment). It is the degree of functional correspondence between the representation manipulation and processing acts supported by the tool, and the acts required in manipulating the representation to achieve the modelling task goal (e.g. an appropriately specified sales forecasting model and suitably organised time series data).

Tool-Reality fit recognises that the demands of the modelling task are somewhat determined by the demands of the substantive task. Consequently, it is the tool’s potential for manipulating appropriate representations of reality. Tool-Reality fit is the degree of functional correspondence between the technology as tool’s knowledge of (support for) procedures of data and model manipulation that are implied by the substantive task, and consistent with the actions and information supported in real-world context (i.e. the substantive-reality task-environment).

When considering the fit to performance relationship it is necessary to identify the specific task-environment of interest — the goals of the task and its place in the hierarchy of operational tasks, modelling tasks, substantive tasks. Indeed, location in the task hierarchy determines the relevant set of performance measures. In practice, performance is a multi-attribute construct; it can rarely be defined in terms of a single all encompassing measure (in part because good measures are often hard to come by, and in part because users have multiple goals to satisfy).

In its most general form, performance can be measured in terms of effectiveness and efficiency. Effectiveness relates to the output quality of the task (the task as product). Efficiency relates to the costs of inputs (information cues, required acts) for a given level of output (product). At higher levels in the task hierarchy, effectiveness measures dominate user attention. For example, in the substantive task of producing a sales forecast, forecasting accuracy is the key performance measure. Of course, the forecast must be completed in reasonable time (i.e. the required acts cannot be too onerous) and the relevant data must be readily available at a reasonable price (i.e. the information cues employed must be accessible economically), but these are matters of efficiency. At the higher hierarchical levels these efficiency performance measures act more as minimum requirements that, once satisfied, become relatively inconsequential. For example, the time it takes (a function of the required acts) to produce the sales forecast will not be of consequence unless it delays the production of the forecast past the point when management can act on the information. This is not to say that the manager will not prefer tools that take less time (e.g. require fewer keystrokes to manipulate data or models). Rather, it suggests that in the context of the super-ordinate task of producing a good forecast, enhanced modelling and data manipulation capabilities are likely to be more important than interface improvements (within some ‘reasonable’ limits).

In contrast, at lower levels in the task hierarchy efficiency issues take on greater importance (e.g. operations such as downloading files, printing documents, or booting up a machine always seem to take too long). Thus, at lower levels in the task hierarchy efficiency measures become more important, largely because effectiveness (product or output) performance tends to be binary (e.g. either the document printed out correctly or it did not, the data were sorted appropriately or not).

For each of the three key tasks in the task hierarchy, performance can be more formally defined as a function of the different fits, in principle, as follows:

More completely, any fit based predictions and explanations of performance must recognise the dynamic nature of fit. Fit is subject to change over time as users learn and systems are refined. Pragmatically, it is also important to recognise that perceptions of fit may often differ greatly from what is actually the case (Davern, 1996; 1999).