About
PS&J >
A Perspective on Software Process Improvement
A
Perspective on Software Process Improvement
To
be effective, improvements must result in near-term benefits in productivity and
cycle time. This doesn’t mean
that we are not interested in improvements that yield long term benefits, only
that we think that all viable improvements have a short term benefit in addition
to a long term one. The reason is simple: effective continuous improvement
requires sustained management sponsorship.
Sustained sponsorship is impossible without a near-term, quantifiable,
return on investment.
There is so much change in today’s business environment that a
long-term return may simply not materialize due to changed circumstances.
Moreover, given the frequent re-organizations that are common in most
businesses, it is unrealistic to assume the continuity of management sponsorship
necessary to make a long-term improvement.
We
prefer to test any improvement opportunity against a few simple criteria:
-
It
is possible to explain exactly how the improved practice will affect
the bottom line?
-
Is
it possible to quantify the cost of the improvement?
-
Is
the projected impact on the bottom line big enough to be significant?
-
Is
the risk associated with the change low enough to warrant the investment?
-
Is
it possible to objectively measure the improvement in process performance
and relate it to business objectives?
-
Can
the investment required to put the improvement in place be recovered in one year?
Our
experience is that most organizations have no shortage of opportunities that meet these
criteria.
Continuous
improvement requires a well-defined, stable software development process, in
which the staff has been trained, and which they are willing to use.
Improvement is impossible without stability, and stability requires consistency.
Consistency is very unlikely without training.
Training is of no value if it is not put into practice.
This means that training courses are not about transferring information
as quickly as possible, they are about motivating behavior.
An effective training course has to give the participants enough positive
experience with the new practices to motivate using them in their everyday
work when the course is over. A
course needs to convince the participants to adopt the new practice and give
them the hands-on skills to perform it effectively.
Follow-up mentoring needs to be provided to answer all the “real
world” questions that can’t be addressed in a class until the transition to
the new practice is complete.
Good
processes are “lightweight” and they focus on what developers actually do at
their desks. Processes that don’t
have a significant impact on daily activities are most likely a source of
needless overhead. Process documentation should be short and simple with clear
quantitative entry and exit criteria. Processes should incorporate feedback to “close the
loop” and build in continuous improvement.
Projects need formal plans and need to have performance managed to
planned targets, but project plans also need to be “agile”, incorporating an
evolutionary approach that builds in opportunities to learn from mistakes and
continually adjust the plan. Planning
horizons should be kept short, at most a few months, since the environment will
continually change. It should
always be easy to re-plan in the face of changing circumstance.
It
takes measurements to quantify the effectiveness of a new or existing practice.
Moreover, once a practice is in place, its effectiveness will decline without
active management. Rational management requires fact-based decisions based on
measurements. This means that an automated measurement
framework, with some
simple analysis capability for decision support, is one of the first orders of
business when embracing continuous improvement for software.
Data collection needs to be simple and fast. The costs of collection must be generally perceived as
negligible. The data must be used directly by the people collecting it to
improve their own practices, or it won’t be collected very long.
Finally,
improvements cannot ignore the human side of the equation.
There needs to be an atmosphere of trust and a sense of teamwork between
management and the development staff. Everyone
needs to have a conviction that improving productivity, cycle time, and product
quality will grow the business and result in improved career opportunities for
all.
We
believe Six Sigma’s model for continuous improvement offers the best overall
frame work for successful software process improvement and Six Sigma thinking
permeates our approach to process improvement at every level.
That doesn’t mean we think Six Sigma is the only way to go about
software process improvement.
On the contrary, we have also had outstanding results with model
based improvement like the CMM, provided we kept the focus on achieving
measurable business results rather than achieving levels, and provided we introduced a simple
metrics framework up front. Any of the published processes, like TSP
and Extreme Programming can provide a fine
foundation for continuous improvement provided they are integrated with a simple
metrics framework and the developers use measurements to actively manage and
improve process performance.
| 
