Models and Theories in Human-Computer Interaction/GOMS and its unpredictability

While GOMS can provide several advantages, such as minimal effort, cost, and time to market, it is plagued by a strong weakness of user unpredictability. This is especially apparent when looking at complex multi-tier application development.

I would like to present an example that applies to my professional experience. Currently I am developing a new sales application for a mid-size trailer manufacturing company. This application needs to have both online and offline capabilities, and will be used in the field by sales representatives at potential clients place of business. Since there are a range of environments that the sales representative will be using the application, it is extremely difficult to assume every possible scenario that could apply. Therefore, when evaluating the application using the GOMS method, we cannot factor in the environmental factors. In example, can we predict a user being interrupted in the middle of an operation? How about user fatigue in the system? Also, since this is an online / offline application, what if another factor such as network connectivity comes into play? Could this disrupt the flow of the application and cause the user to need to use an alternate path / method?

The unpredictability of GOMS definitely puts this method at a severe disadvantage over alternate methods when applied to certain scenarios, as demonstrated above. While GOMS may be appropriate for a select few scenario where user unpredictability does not have as major of an impact, there are a significantly larger number of scenarios where GOMS would not have a place.