Abstract
The difficulties of the non-programming user in developing quality software-requirements specifications are well known and described in the literature. Typically, the user will work with a software expert to develop the specifications. During this process the user must learn new terms and concepts, and must attempt to identify the required functions of the resulting software product.
The goal of this investigation was to evaluate the capability of non-programming users (in this case experienced inventory managers) to develop specifications for an inventory error-control system. Participants specified tests from a set of available tests to detect possible errors in inventory change-records, i.e., in inventory updates. Costing-aids were provided that were analogs of aids that are expected to be presently available. The costing-aids provided an increasing data on system cost: the total system cost, the total system cost plus the costs of each part of the system, and an automatic sort of previous designs according to cost.
Results shewed that problem-complexity had a strong effect on performance; greater problem-complexity resulted in more costly designs. Further, the more complete costing-aids tended to degrade performance on the simple problems, yet they improved performance on the more complex problem. It was concluded that the ability of non-programming users to develop least-cost requirements specifications was poor when using any of the aids that were made available. The experimental results, however, suggested new approaches for procedural and computational support for the non-programming user, that could lead to more suitable costing-aids.
Get full access to this article
View all access options for this article.