Authors
Paul B. Cragg & Peter Grant
Abstract
Spreadsheet models are increasingly being used in decision making within organisations. With questions about the quality of these models, an investigation was conducted into the spreadsheet practices in ten firms, with an emphasis on the process of building spreadsheet models.
The study showed that spreadsheet models were usually built in an informal, iterative manner, by people from all organisational levels. These people had received very little training in the building of models, which could help explain why 25% of the models contained errors.
Other problems were also found. It was evident that the spreadsheet practices in the firms were inadequate. There is a need for increased training as well as setting and enforcing organisational spreadsheet standards.
Although the study was viewed as exploratory, it indicated a need for further study into the effect of formal design practices on such factors as the incidence of errors and model creation time.
Sample
All of the models investigated were built using relatively informal methods, ie. without formal specification, design, coding, or documentation.
Of the 20 spreadsheets investigated:
- Specification. Only 7 involved specification by others. Although this is low, it can be attributed to the fact that most of the models were used solely by their creator. Two had been built for others without any input from them.
- Design. Design took place with only 8 of the models. Even this design was very rudimentary, mainly involving the design of input screens and reports. Planning the layout of the model only took place with 8 of the models studied.
- Testing. All but one of the spreadsheet models were tested. Model validity would be increased if the models were tested by someone other than their author. This was done in only one case.
- Documentation. Documentation of the models was poor. Only half of the models contained documentation of any kind. Of those that did have documentation, the majority only provided sample input and output which was obtained from a printout of the entire model.
Publication
1990, Operational Research Society of New Zealand, August, pages 131-136
