Building quality into software products

Measuring quality

Gillies (1992) details five approaches to measuring quality from the purchaser's point of view. These are: 

  • Simple scoring 
  • Weighted scoring (or phased weighted factor method) 
  • The Kepner-Tregoe method 
  • The Cologne combination method 
  • Polarity profiling

Because of deficiencies with these methods and because of their older date Gillies suggests a new technique called LOQUM (LOcally defined QUality Modelling) which involves a three step approach:

"LOCRIT: a knowledge elicitation exercise to derive the relevant quality criteria and associated measures.

LOCREL: a further knowledge elicitation to define relationships and conflicts between criteria. 

 LOCPRO: A profiling tool to display a graphical profile to represent the overall quality of the system".

A more scientific approach was proposed by both McCall and by Boëhm. Typical examples are:

Mean Time Between Failure (MTBF)

MTBF = T-tot

N

T-tot = the total time period N = the number of failures in T-tot

Complexity is measured by McCabe (1976) "as a cyclomatic number, based upon graph theory, that seeks to estimate the number of linearly independent paths through a program".

This level of detail is beyond the scope of this paper. Gillies (1992, p40-43) cites numerous references to quality measurement.