Professional Documents
Culture Documents
ESCUELA DE INGENIERIA
Profesor Supervisor:
YADRAN ETEROVIC S.
YADRAN ETEROVIC S.
ROSA ALARCÓN C.
FELIPE GUERRERO V.
ii
ACKNOWLEDGMENT
My thanks goes to all the people at Link, specially the core Project
Management Tool development team –Rodrigo Mansilla, Patricio Pardo, César
Rodríguez and Patricio Valdés– which I was proud to lead.
iii
TABLE OF CONTENTS
Pág.
DEDICATION............................................................................................................. ii
LIST OF TABLES...................................................................................................... vi
ABSTRACT.............................................................................................................. viii
I. PREFACE...............................................................................................................1
REFERENCES ...........................................................................................................36
A P P E N D I X E S ...................................................................................................38
vi
LIST OF FIGURES
Pág.
Figure 2.3: Number of SW-CMM appraisals reported to the SEI by year ...................6
Figure 5.1: Link’s time for moving from level 2 to level 3 ........................................25
Figure 6.2: Link’s time for moving from level 3 to level 5 ........................................27
vii
ABSTRACT
We focus on the strategy guiding that effort, in how to approach both the
SW-CMM Key Practices and the regular activities in a software development
company, and also the importance of a centralized requirements/project management
tool for achieve such goal.
These factors are, in first place, taking the upper level Key Process Areas
(KPA) as projects, and making a project-centered approach to every activity
conducted by the company. In second place and to support that, is the use of a
centralized tool, which makes easier for the project manager to manage the project,
while allowing senior management to feel in control of every undergoing project.
viii
1
I. PREFACE
Okay, maybe SQA or SCM were not that strange to me, but “peer
reviews”? Well, really the most amazing thing I saw was having a full project
management tool in place, that allows clients to place their requirements and then
follow requirements’ complete life cycle, including their approval by their own
bosses, the project creation and management, testing, and finally –only by client–
requirement closing; that was really something.
A few months before, the company had been assessed and become SW-
CMM maturity level 2, and by the time of my arrival, they were working to achieve
SW-CMM maturity level 3 using the IDEAL model.
Today, two and a half years from then, when a new programmer arrives
at Link, he or she possibly thinks I am an extraterrestrial, talking about SEPG, CPI,
SPI and causal analysis.
That tool continuous use, enabled better control over the projects, and
availability of real-time online project's performance data. Having an in-house
developed software that smoothly maps with the SW-CMM model, were
fundamental on Link’s software process improvement success.
The model proposes five hierarchical maturity levels, where each upper
level supposes that its lower levels are completed and institutionalized along the
company.
Defined (level 3), where a more robust process is defined. There are
practices in place about training, group coordination, detailed engineering
techniques, and more sophisticated metrics and project management.
5
Today, the SEI has replaced CMM with CMMI, the model’s natural
evolution after almost 20 years of use and feedback. The SW-CMM model was
supported by the SEI until December 2005; nevertheless, the success factors and the
experience on continuous improvement based on CMM are still valid for CMMI.
Between April 2002 and December 2006, 1,946 CMMI appraisals were
conducted (in 1,712 organizations) [Carn07], so we can say that SPI initiatives are
becoming more and more popular both in the United States and the rest of the world.
Two thirds of the appraisals were conducted outside the United States.
According to the SEI, IDEAL supports all the phases required for
performing SPI, and when you complete the cycle, immediately a new one starts, and
this is repeated over and over. The phases are as follows:
As it was said previously, there are pre-onsite activities and there are on-
site activities. Figures 2.5 and 2.6 show those activities schematically:
Pre-on-site activities last a few months, during which both the lead
assessor and the company start planning the assessment. This isn’t a full time
activity. On-site activities may last anywhere between 2 and 14 days, but they
usually take 5 to 10 days, as it’s shown in Figure 2.7:
From IPSS’s work, [Sanf06] lists fifteen critical success factors, focused
on top-risks. Still, the work needs more detail, but presents a preliminary view of
several factors well known in the industry. The factors, in no particular order, are:
Sponsorship, SPI action plan, assessment, deployment, support, consulting, skills,
work practices, reward system, participation, communicate, change management,
learning, values and history.
In [Jalo02] there are three factors mentioned as critical; for making the
classification easier, in bold we added a key word to summarize each:
In [Guer03], forty-three success factors are listed (see Appendix B for the
full list). Guerrero classifies those factors in six categories, and ranks them according
to their frequency of appearance in the literature available at the time. The two most
frequent factors for each category were identified, defining a twelve-success factor
list, which is presented in Table 3.1.
14
While none of these factors refers directly to company’s size, the “Size
of improvement effort and visibility” can be considered related to the company’s
size, in a proportional way. The five most frequently cited factors (in order) are:
b) Management commitment.
c) Practitioner’s commitment.
Now, crossing reference among the three sources [Guer03], [Jalo02], and
[Sanf06], you could say that the most critical success factors in SPI initiatives are:
15
This work will explore these factors in the context of Link’s case, but
mainly we’ll try to discover if there are others particular to Link’s case.
Senior Management
SEPG
The rest of this document is mainly related to the efforts for achieving
maturity levels 3, 4 and 5. The path from level 1 to level 2 is described in [Guer04],
which served as a foundational stone for the subsequent developments.
Two objective benefits identified at the time, which have lasted over the
years and served as a key factor for keeping sponsorship are the following. First, the
mean response time to requirements remained close to the committed time (once a
requirement is approved by the client’s boss, Link has 48 hours to provide rough cost
and effort estimates), giving Link an organizational credibility in this matter.
Possibly, some work still needs to be done in approval response time, but for the
time of the present work, this approval served as an effective filter for avoiding non-
software requirements input. Second, the rework rate is approximately 5%, a figure
frequently used for getting funds for process improvement or even for Link’s sales
support.
17
Figure 4.1 shows the requirements workflow, where a client can enter
and close a requirement. The workflow allows the different roles to approve o reject
the requirements and the initial estimations. There are several roles that interact with
the tool: the client, the client’s boss, the project manager, the senior manager, the
developers (analysts, programmers, architects, etc), and the software quality staff.
19
When a client enters a requirement into the system, his or her boss has to
approve or reject it. Every time a requirement is approved, the company’s board of
project managers assigns the requirement to one of them; this project manager makes
a fast initial estimate of cost and development time.
Then, the client’s boss decides whether to carry on with that requirement
or not. If the initial estimate is accepted, the project manager creates a project and
begins an initial stage where the analyst work is required. When the analysis is done,
a formal document with the plan is delivered and a presentation of the solution to the
client is conducted.
If the client wants to continue with the project, the project manager
inputs the project plan into the project management tool, scheduling every task
known at the time and estimating a closing date, according to the duration of the
tasks and the availability of resources.
During the project life cycle, there can be some schedule or cost-related
changes, but all of them are negotiated and agreed between the project manager and
the client, changing either the closing date or the project cost.
Finally, when the project ends, it’s only possible to close it when the
client himself has closed the original requirement. If he or she is not happy with the
20
result, the project remains open and it is the project manager’s job to make things
happen and get the project to a successful end.
For the project manager, together with the regular features already
described, there is a control panel, visible every time he or she enters the system. The
control panel shows for each current project three leds, one for each metric used to
control the project status: Cost Performance Index (CPI), Schedule Performance
Index (SPI) and Defect Density (DD). Each led can turn green, yellow or red,
according to whether the metric is under control, not applicable, or out of control
(See Figure 4.2). Next to every led there is a link, which when pressed shows a full
color control chart, indicating the metric’s current and historic values and its upper
and lower limits.
The daily record of the project team’s work generates these metrics. Each
person involved in the project is assigned to one or more tasks, and specific time
21
slots to work in each task are scheduled as well. Then, once a day (or once every two
days), he or she must enter into the project management tool the actual number of
hours dedicated to each assigned task.
In this way, the database is filled with all the information required for the
CPI and SPI metrics. By task reporting, the employee enters in one day the number
of hours dedicated to one particular task. He or she can also report any scheduled
task and mark it as finished. (See Figure 4.3)
For this iteration, the strategy was first of all to create mini teams that
studied and became “experts” in one particular process area. Each employee was a
member of at least one mini team, keeping in mind the goal of involving everyone in
the SPI initiative. According with the IDEAL model, a diagnosis phase began. In this
phase, each team studied whether each practice of the corresponding Key Process
Area (KPA) was used in the company and which was the level of implementation
and institutionalization, as well as whether each goal was being achieved.
In Link’s approach, the mini teams could meet for at most one hour every
day, so not a single employee was ever required to work for more than an hour a day.
In this way the real trick for the project manager was to schedule and coordinate the
meetings satisfying this constraint, so the management decide to “reserve” one hour
daily for each employee.
The idea behind this strategy was in first place, that this is a small
organization, where individuals are very important so you can’t separate them from
their everyday work. Also, it was a way to test the process generated during the
preparation for the level 2 assessments, using it to generate the project plan and
supervision.
SEPG. This generates new work to be done, such as changes in the project
management tool, new checklists, etc. The mini teams, or the company developers in
the case of the project management tool, developed all that.
The next step was obviously the formal CBA-IPI, where no improvement
opportunities were found and where Link achieved level 3. Once the assessment was
completed, the IDEAL’s learning phase began, where we analyzed and discussed the
SPI cycle that was ending and the software development process strengths and
weaknesses, and implemented some minor changes. Then, we were ready to begin a
new iteration.
There was no peer review experience inside the company, so in this case,
it’s easy to describe the whole process for implementing that practice.
Peer Reviews, refers basically to the process in which other person than
the one who created some work product, but with similar skills, reviews some piece
of code or another product. It’s expected that conducting a fresh review of some
product, in the eyes of an expert, could help find some hidden defects faster than
using testing techniques.
In the meantime, SPI’s project manager was organizing the ideas and
keeping the consistence between all the changes proposed by all the mini teams
working in all the practices of the SW-CMM maturity level 3.
Later, the SEPG approved those PIP’s and released a new process (a new
chapter on Link’s methodology), named (of course) Peer Reviews, where all the
activities that a project needs to carry out for doing peer reviews are described.
The cost of the assessments was about USD 20,000 for both, the informal
and formal approaches. Thus, the total cost for getting from level 2 to level 3 in
fifteen months was about USD 185,000, a figure we think more than affordable for a
small size software development company in a country like Chile, considering that
the main cost (89%) was absorbed by internal employees in their regular labor days.
Even if this cycle looks more expensive than the first one (US$185,000
over US$121,000), it was possible to keep all the time the support of senior
management, whom received monthly updates and reports with the status of the
process improvement project.
At this point, every single employee was very enthusiast about the SW-
CMM model and was very motivated to continue the process improvement effort. It
can be acknowledged that the continuous involvement of the people and the frequent
delivery of results (the formal assessments) are vital factors to succeed in every
process improvement initiative.
As a result, when the next cycle started, the first question was, “Where
do we go from here?” Obviously, the IDEAL’s diagnosis phase begun, a new SPI
project was created, new mini teams were formed, and so on…
After 14 months of hard work, Link was formally assessed and achieved
the level 5 with the following strengths:
Figure 6.1: SW-CMM assessments distribution for all the companies assessed
to September 2005.
The average time for moving from level 3 to level 5 is 38 months, and
the “smallest observed value that is not an outlier” is 18 months, so Link is one of the
fastest companies in the world for moving from level 3 to level 5 (14 months).
Next figure shows all the registered data about the time to move up from
maturity level to the next one. Note that the 38-month for moving from level 3 to 5 is
obtained by adding the medians for moving from level 3 to 4 (25 months) and
moving from level 4 to 5 (13 months).
Talking about costs, the SPI project took 4,330 man-hour or USD
52,000. Including the development of the project management tool (USD 22,000),
and both the informal and formal assessments (USD 20,000), the effort’s total cost
was about USD 94,000.
As you can see in Table 6.1, getting from level 3 to 5 was cheaper than
moving from level 2 to level 3, and still cheaper than moving from level 1 to level 2.
29
an employee must enter into the system the number of hours worked in the task
previously assigned to he or she by the project manager.
The project management tool solves this issue in a very elegant and
simple way. This is a major strength of the tool and of course, one of the most
relevant factors to keep projects under control and eventually to achieve some of the
goals required by SW-CMM.
Once a tool like the project management tool is available, you need to
develop a strategy to make the employees to use it and to report about their work. In
Link’s case, there is a monetary incentive for reporting the number of hours worked
using the project management tool.
In the same direction, by getting the entire project planning and tracking
inside the software, it wasn’t necessary to maintain a lot of documents. Instead, Link
actually used only one main document (and several secondary small documents) for
each project, avoiding a lot of bureaucracy and, among the “each resource reporting
task completion”, give time to the project manager to actually lead the project instead
of spending several hours doing paperwork.
From the time of the project management tool deployment to the time of
the level 5 assessments, around 1,400 projects had been conducted using this tool.
This generated a considerable amount of data that were useful to statistically
determine which metrics should be used.
Because of all that, it was only necessary to develop within the project
management tool a control panel (see chapter 4.2.1), where the project managers and
senior management could see at any time the historic and current values of those
three metrics, and graphically analyze them.
- Defect Prevention
For working in these three KPAs, two process mini-teams were formed
(as usual in Link’s strategy, using the IDEAL approach). The first one, studied
32
The common factor was that the created processes, defined that the three
KPAs will be satisfied by keeping three strategic projects; each one with its own
particular activities, but projects at last.
The other two KPAs were also implemented by creating projects derived
from internal requirements (the Software Projects Manager was the client). Both
projects had a part-time project manager, who manages his project according to the
process definition developed by each mini-team.
33
VII. CONCLUSIONS
It’s important to understand that every company can interpret the model
in a way that works for the company. Rather than moving the whole company to a
rigid and bizarre model, you need to write their own software process in the way it
makes sense for itself.
The main strategy followed by Link was to take the IDEAL model and
develop an accordingly methodology, and to use a tool for supporting such
methodology
From this experience some lessons can be pulled out. Some of these
lessons are well known in many software development process implementation
efforts, but some of them are clearly unique. In our opinion, these unique lessons
made the difference for having succeeded and still being successful in the projects
that Link undertakes.
First, there is the great vision and support from the top of the
organization (management commitment). Maybe this is the obvious one, but the
secret to maintain that support is basically to manage the process improvement like
any other project.
In particular, the project must include a realistic but aggressive plan and
should follow it as it was “sold” to management. Also, you should be able to show
34
results at any given time during the project. If you can summarize the spirit under
SW-CMM it is that it allows you to establish and meet compromises.
Based on the above, you can realize that the critical success factors
covered by known literature were full accomplished in Link’s case, but looking at the
figures, Link’s time to move from one level to he next one were always in the fastest
group.
As described in section 5.3, the yearly cost of SPI per software engineer
at Link is pretty close to the industry average, so the main difference was the time
spent on each SPI initiative. Each SPI cycle was by far faster that the industry
average, resulting also in a lower final cost. The question now is, “what makes
Link’s case that fast”?
To support all that, the project management tool was vital. That tool
avoids a lot of bureaucracy and allows project’s managers to be in control of their
projects at all time, without having to spend several hours writing reports (task
completions, % of advance hour by hour, etc.).
The use of this tool was a critical success factor, as this tool heavily
reduced the bureaucracy and made work really flow.
REFERENCES
[Sanf06] SAN FELIU T., GARCIA S. and GRAETTINGER C., Critical Success
Factors (CSF) in SPI Bibliography, Proceedings of the First
International Research Workshop for Process Improvement in Small
Settings, 2006
37
APPENDIXES
39
a) The Project Manager is responsible for leading him project’s peer reviews.
b) Peer Reviews are conducted for defect finding in work products and opportunely
correcting them
A.2.2 Goals
A.2.3 Definitions
A.3 Roles
A.3.1 Responsible
A.3.3 Training
A.4 Inputs
a) Matrix: Pre-workability
A.4.2 Infrastructure
b) Meetings Room
c) Data Show
A.5 Activities
Responsible: Reviewed
Description: The reviewed removes all defects identified during peer review,
which the project manager determines.
The pm must verify that this activity is accomplished.
Before conducting peer review, the work product been reviewed must
been finished inside the scope of a project and configured into the Configuration
Management System
A.7 Outputs
A.8 Revisions
Software Projects’ Director: The SPD gets a periodical report from the
Project Managers, generated from the Project Management System’s binnacle.
- Project Team
- Stakeholders
44
It’s important that the project manager is aware of the following factors
during the peer review:
A.9 Metrics
Unit Hours
Definition The effort is the sum of the hour tagged “RP” in the Project
Management System
This category includes the factors affecting the whole organization, like
organization’s size, business goals, mission, etc. Changing these factors involves
changes in the general direction of the organization, which is usually driven by a
strategic plan, dictated for several years. The main managerial level affected by these
factors is senior management, because this level dictates organization’s strategy.
4 Organizational structure
3 Application diversity
2 Project size
2 Organization size
1 Number of customers
1 Personnel rotation
teams, as they are the target for SPI. These factors cover all the managerial levels,
because there are cultural issues for each of the levels.
18 Management commitment
6 Climate of change
2 Change agents
1 Being a subcontractor
47
18 Practitioners involvement
5 SPI incentives
2 Champions
This category includes factors under direct control of the process team
about the management of the SPI project itself: Number of people involved, time to
devote to the effort, resources, etc. These factors affect the process team—it’s their
responsibility to manage them. The factors included in this category are assigned at
the project management level where decisions affecting the factors will be adopted.
This category includes low-level SPI details like the design and the
deployment of the improvements. These factors are the technical core of a SPI
initiative. These factors affect the process team, because they are the most likely to
manage them, and are assigned at the project management level where decisions
affecting the factors are adopted.
5 Assessment
4 Improvement sequence