[Development Factors and Product Quality]
The test factors are relative to the programming effort.
System engineering is relative to the total of test and
programming effort and project management is relative to
the total of programming,test and system engineering effort.
The number of defects in the last row in the table is based
on using average C programming language FP to KSLOC ratio.
The parenthetic test factor in the first row is for a CMM
level 1 organization while all others for the first row are
for CMM level 2-5 organizations. Meta Group and C. Jones factors
are for an average maturity organizations,which are at the "
bottom half" of CMM level 1.
===========================================================
[Programming Productivity and Test Factor/Product Quality]
Programming Productivity implied in IBM FSD development factors:
C and other high level languages, low complexity code = 255-650
SLOC/PM. The high end of the range results from increasing
maturity of the development environments. Av.. = 450SLOC/PM
Programming Productivity for MetaGroup factors:
Worldwide average productivity measured over 770 projects = 650
SLOC/PM.
Programing Productivity for factors reported by C. Jones:
Average productivity for commercial and system software using
average FP conversion factor for Clanguage = 960 SLOC/PM
The higher programmer productivity reported by MetaGroup and
C. Jones may be because of the lack of software engineering
discipline employed by the observed projects towards code reviews,
design, unit test, etc., which increases programmer productivity,
but pushes defects into system and integration test phase,
hence requiring more test resources. Higher product defect
rate also supports this conjecture.
===========================================================
[Product Defect Ranges]
Given the product size and early discovery defects (either actual
or estimated), The independent test defects can be tracked against
the Rayleigh curve to reach the target product defect rate.
Criticality based product defect ranges that have proved
to work for some benchmark applications, such as the space
station, shuttle and large commercial systems, that can be
used as reasonable product defect goals are shown below.
===========================================================
[Other Product Defect Averages]
The defect numbers from C. Jones were calculated using average
lines per FP for C programming language and average defect rate
for a SEI SW-CMM level. The minimum defect rates are given in
parenthesis for each level.
The defect rates observed by Meta Group over 770 worldwide projects
that Delivered programming productivity of twice the average for
all the projects. These companies used 4GLlanguages more than the
average projects did.
===========================================================
[Development Effort and Quality]
Historical Trends
IBM FSD:
a.Four times additional test effort and twice the project
management to reduce the defect rate by half
b.Eight times additional test effort and five times project
management effort to reduce product defect rate to 1/10th
Putnam:
a.25% increase in schedule to reduce product defect rate by half
B.50% increase in schedule reduces product defect rate to one fourth
文章定位：
