Remarks
from SPC Trial for an Emergent Organization
Ayça
TARHAN, EUROPEAN SEPG, 12-15 June 2006,
Abstract:
The
application of statistical process control (SPC) techniques for software is
rare due to such requirements as high maturity, rational sampling, and
effective metric selection. Companies that invest time and money on a process
improvement model like CMMI or ISO 15504, can also
take the advantage of following a well-founded framework to establish the
infrastructure required for SPC implementation. For other companies, however,
the path to SPC implementation is not that clear due to lack of targeted
guidelines. Existing studies frequently focus on potential benefits of SPC
results rather than providing guidelines based on practical evidence. The need
for such knowledge encouraged us to perform a case study in order to answer two
basic questions: 1) Can we identify procedures to guide SPC implementation? 2)
Can an emergent organization apply SPC techniques following these guidelines?
The case study was performed on review process of a system and software
development organization which has 50 staff and currently pursues process
improvement studies to achieve CMMI L3. While performing the case, we used a
specific procedure to evaluate the suitability of review process and metrics
for SPC. We worked on existing review process data of 200 data points,
collected by established company-specific procedures for two years. We
translated the review data to a form appropriate for comparison among different
projects and products. We investigated the utilization of review metrics as
review effectiveness, defect removal effectiveness, and review open period with
respect to defects. As control chart is one of the most sophisticated data
analysis tools within SPC, we demonstrated practical evidence on the
utilization of SPC via control charts. Our experience has showed us that with
established guidelines for rational sampling and metric utilization, an
emergent organization can apply SPC techniques and attain the ability to
understand its processes based on quantitative data.