• Read carefully the “Supporting Scaling Agile with Portfolio Management: Case PAF.com” for questions 1 & 2
• Write answers to the following questions 1 and 2 in your own words labeling each answer as directed in the
question.
• THEN ANSWER QUESTIONS 3 & 4 BASED ON BEING IN A TEAM COMPANY
• One page for each question.
Questions 1 & 2
1. In the case study paf.com has already converted to the Agile methodology but has problems that it is hoping that
PPM can either solve or alleviate. Starting with the most severe
1.a. describe that most severe problem.
1.b. describe the second most severe problem.
2. Describe in your own words how paf.com is using the “lean philosophy” is solving the problems you described in 1a
and 1b using events in the case study,.
Questions 3 & 4 (look at the attached file named ”Fadoodle-Company Project”) to know about the company team project.
3 Name of your company team and the total amount of company investment funds available and the total amount of VC
(Venture Capitalist)funds wanted and total amount received,
Then, describe in detail how you made or planned to make your industries and markets (parts of the world) attractive to
VCs (Venture Capitalist). Please mention how you contrasted your industries and markets to VC’s. Please end your answer
with your evaluation on how successful your efforts on marketing your industries and markets.
4 Describe in your own words the measures your company took or planned to take to convince VCs that your proposed
projects did not have an unreasonably high level of risk.
Three VC (Venture Capitalist) questions to Fadoodle:
If I invest $3million, when will I get my money back?
How are you competing with other bike rentals companies?
Why the public would choose your bikes over your competitors?
What makes your company standout from your competitors? Are your bikes lighter? More colorful?
Will Fadoodle be the exclusive provider of bike-share within the cities?
I will only be interested in investing if the bike-share market will be regulated so that only Fadoodle can operate in
the given cities. Securing such a deal would cost more money but also create more revenue and licensing values, similar
to taxi cab medallions. If Fadoodle can securee such agreements with cities I will be willing to invest $600,000 per
city for 3 cities for a total of $1.8 million. The $100,000 extra per city is to be used exclusively to negotiate that
these markets remain closed-markets.
Do you have exclusive rights to the bike rental market’s in which you are trying to enter?
What guarentee’s do I have that my money would be invested towards the projcets directly?
What precautions do you have in place to avoid issues such that Divvy had when pentrating the new markets?
Feel free to contact me directly with your responses. I really like your ideas, and would be interested in investing a
solid portion, if not all, of my $2 million into this company.
Fadoodle (Name of the company)
Our story:
Fadoodle is a convenient, bike share company with locations in the cities of Austin, TX; New Orleans, LA; Miami, FL;
Jackson, MS; Birmingham, AL; and Phoenix, AZ. Fadoodle is about enabling people to living GREENER lives by encouraging
the use of fuel free transportation that is good for your health, fun, convenient and affordable.
Customer value proposition:
“Fadoodle is a new kind of bike sharing program where riders can conveniently rent a bicycle and return them at an
affordable price. It provides a green alternative 24/7.”
Why VC (Venture Capitalist) should invest in Fadoodle:
• Only a small percentage of cities across the U.S. have bike share programs. This is an untapped, lucrative area
for Fadoodle.
• Our business model (technology, service set up, relationships with suppliers, city government knowledge) can
easily be replicated and extended into new cities. We have the recipe!
• We need your investment of $3 million to support Fadoodle’s long term growth strategy. Over the next three years,
we plan to open 12 new bike share programs in 12 new cities. We estimate it costs $500,000 to open one bike share
program.
If you invest in our company, we are willing to give you up to 33% equity in Fadoodle. Our company is currently valued at
$9 million.
Company Name: Fadoodle
Industry: Transportation
Primary Service: Bicycle share/rental service to begin in the following cities (6): Austin, TX; New Orleans, LA; Miami,
FL; Jackson, MS; Birmingham, AL; Phoenix, AZ.
Mission: Fadoodle is not just about providing rentable bikes in convenient locations to the great people of Austin, TX;
New Orleans, LA; Miami, FL; Jackson, MS; Birmingham, AL; and Phoenix, AZ. Fadoodle is about enabling people to living
GREENER lives by encouraging the use of fuel free transportation that is good for your health, fun, convenient and
affordable.
Defining the Business:
• Service: See Mission Statement.
• Customers
o City visitors
o City residents
• Competitors
o Public transportation
o Other rental bike companies
o Taxi cabs
o Personal motor vehicles
o Companies that sell bikes / bike shops
o Walking
• Supply Chain
o Design and purchase bike frames directly from manufacturer. Bike frame material has extended life and will double
the average life of the bike.
o Purchase bike components from third-party supplier
o Ongoing bike maintenance will be provided by Fadoodle employees
SWOT Analysis
• Strengths
o Ease of access
o Low overhead
o Green
o Convenient
o Fun
• Weaknesses
o Liability
o Inventory
o Maintenance
• Opportunities
o Alignment with city public transportation authority (ticket system integration)
o Expansion to other cities
• Threats
o Theft
o Vandalism
o Competitors
o Weather
o Relationships with city governments
Define Strategy / Strategic Goals:
• Strategy: Provide high quality bikes at a reasonable price
o Strategic Goal: Obtain and maintain relationships with top bike and operational service vendors
• Strategy: Attain 70% of potential customer base.
o Strategic Goal: Develop a PR, marketing & social media strategy to engage customers
• Strategy: Get involved in the community
o Strategic Goal: Social contribution community involvement (sponsor free bikeathon)
• Strategy: Expand the business
o Goal: Open up a new location every three years
• Strategy: Build and increase business presence on the Web
o Goal: Increase awareness of the business
o Goal: Use Website as marketing tool
o Goal: Implement high end functionality for users to give company competitive edge.
• Strategy: Operational Software Implementation
o Goal: Purchase and Implement key software needed to run business ops.
• Strategy: Turnaround time on investment within 1 year.
o Strategic Goal: Use algorithm to make sure there is appropriate profit margin.
o Strategic Goal: Develop monthly Sales/ Revenue Goals for the first 12 months to meet 1 year break – even point.
• Strategy: Maintenance plan on bikes so that they are kept in good shape.
• Strategy: Explore payment options
o Strategic Goal: Research convenient way to accept payment using modern technology.
o Strategic Goal: Enable customers to pay via friendly user interface and credit card.
• Strategy: Ensure bikes are easily accessible in convenient locations
o Strategic Goal: Make sure analysis of consumer base and transportation routes support placement of locations.
Projects:
Project Name: Expand bike locations in cities
Meets Which Strategic Goal?
• Expand the business
Need:
• To increase use of bike share service. Station needs to be close to each other to provide good and fast service.
Problem Met:
• Increasing demand and not enough supply
Major Deliverables:
• Survey customers to ask for bike station suggestions
• Provide a map of proposed areas to add bike stations to show bike stations
Costs: $6,000,000
Project Name: Website Development
Meets Which Strategic Goal? Describe
• Build and increase presence of business on the web.
o Website creation to increase awareness of business
o the website will provide the ability to find business locations and information related to each location such as
how many bikes are available at each location.
o There will also be mobile version of the app so that people can look up necessary information from their smart
phones.
Need:
• Hardware infrastructure in place to host web services
• Unique URL/IP address
• Web development service to write the code behind the website and all its requirements
Problem Met:
• Awareness of service
• Accessibility to services
Major Deliverables:
• Project Plan
• Technical Requirement Plans
• Vendor Contracts Signed
Costs: $150,000
Project Name: Bike Rental Management (BRM) Software
Meets Which Strategic Goal? Describe.
• Strategy: Maintenance plan on bikes so that they are kept in good shape.
• Strategy: Ensure bikes are easily accessible in convenient locations
Need: The bike rental management software will allow Fadoodle to:
• Track the maintenance of each bike through the bike’s lifecycle
• Track real-time bicycle usage via GPS
• Adjust distribution of bicycles based on demand
• Schedule maintenance
• Schedule bicycle pick-up and drop offs
Major Deliverables:
• Project plan
• Technical requirements
• Cloud-based, 3rd party managed BRM
Costs:
• Program, Design, Test, and Implement: $400,000
• Monthly fee (includes maintenance, support, and 8 licenses): $800
Project Name: Secure Vendor Contracts for Bicycles and Security Racks
Meets Which Strategic Goal? Describe
• Provide high quality bikes at a reasonable cost
o Ensure that quality bicycles are available to Fadoodle customers
Need: High quality bicycles
Problem Met:
• Purchase or rent bicycles? Own? Lease? Outsource?
• Time to research vendors and secure contract
Major Deliverables:
• Contracts
• Relationships
• Bicycles
• Racks
Costs: Estimated $400,000
Project Name: IT Infrastructure
Meets Which Strategic Goal? Describe.
• Build and increase business presence on the Web
• Operational software implementation
o Provide IT Infrastructure and Datacenter to provide a means to conduct business for email collaboration, data
processing, financial transactions, payroll, data storage, etc.
o Enable customers to pay via friendly user interface and credit cards
Need: Acquire network and server hardware and schedule professional services to implement and configure hardware and
business software.
• Purchase Hardware
• Purchase Software
• Schedule Professional Services
• Test
Major Deliverables:
• Project plan
• Technical requirements
• Data Center Creation
• Server Implementation
• Software Implementation
Costs: $500,000
• Server Hosts: $100,000
• SAN: $85,000
• Routers / Switches: $45,000
• Software Licensing: VMware, Veeam Backup, MS Server 2012, MS-SQL, MS Exchange: $100,000
• Additional Data Center Costs: $100,000
• Professional Services: $60,000
Project Name: Advertising
Meets Which Strategic Goal? Describe
• Attain 70% of potential customer base.
o This will allow our company to reach out to even more customer. In order to extend our company throughout Texas
we will need to do more advertising and give out promotions.
Need:
• Specialist (in order to)
o Preparing for Advertising and Promotions
o Evaluating Advertising and Promotions
Problem Met:
• Generate awareness in communities of Fadoodle service
Major Deliverables:
• Analysis
• Plan
• Launch
• Evaluation
Costs: $100,000
Project Name: Customer Service Automation Software and Service
Meets Which Strategic Goal? Describe.
Strategy:
• Operational Software Implementation
Need:
• First line of support for customers who contact Fadoodle
• Software and third-party service to categorize and prioritize customer help tickets and route call to correct
parties
Major Deliverables:
• Project plan
Costs:
• Setup: $10,000
• Annual fee: $50,000
Future Projects:
• Development of APIs for integration with city public transport ticketing system.
Project Ranking and Costs:
Projects Pts Cost
Secure Vendor Contracts for Bicycles 20 400,000
IT infrastructure /Hardware 19 500,000
Bike rental management (BRM) software 16 400,000
Financial management suite 15 200,000
Website Development 14 150,000
Advertising 10 100,000
Opening additional bike station location 5 6,000,000
Customer Service Automation Software and Service 1 60,000
Supporting Scaling Agile with Po rtfolio Management: Case Paf.com
Kristian Rautiainen
Aalto University School of
Science and Technology
kristian.rautiainen@tkk.fi
Joachim von Schantz
Paf
Joachim.vonSchantz@paf.com
Jarno Vähäniitty
Aalto University School of
Science and Technology
jarno.vahaniitty@tkk.fi
Abstract
This paper is a descriptive case study of how one
department at Paf, Paf.com, introduced portfolio
management to help support scaling agile software
development. Paf.com had experienced problems with
long time-to-market due to thrashing, which was
caused by frequently changing priorities due to an ad-hoc prioritization process and handovers. Also, there
was lack of visibility into projects entering and
progressing in the development pipeline. No structured
way of starting projects was enforced company-wide,
and too many parallel projects got started. As a result
of introducing a structured portfolio management
process, the number of ongoing projects has
dramatically reduced, from over 200 to 30, reducing
thrashing. Listing all projects in priority order in the
Paf.com backlog provides visibility into what is
currently ongoing, helping coordinate the work of
multiple Scrum teams. The portfolio follow-up function
provides progress data on the projects, helping
managers make more informed decisions, considering
the whole portfolio.
1. Introduction
During the past 10 years, agile software
development methods have gained acceptance in the
software industry. Originally, the sweet spot for agile
software development was one co-located team of 3-8
persons working on one product [3]. Lately, the trend
has been in scaling agile to the enterprise level and
applying lean philosophy to software development, see
e.g. [14-16, 22, 23]. Some amazing results have been
reported about applying fully distributed Scrum [28,
29], but not many can boast such impressive results.
The transition from the agile sweet spot to
enterprise-scale agility is not easy. Kalliney [9] reports
“portfolio problems” when scaling Scrum, specifically
regarding product management and enacting the
company vision, managing cross-team risks and
dependencies, and handling the silos of knowledge and
skills.
The process for achieving balanced resource
allocation in terms of value maximization, strategic
alignment, risk level, and the number of ongoing
projects is called new product development portfolio
management [5], or portfolio management for short. It
has been mostly addressed in literature on new product
development (NPD), but lately also in literature on
scaling agile and lean. However, experience reports
and case studies in the context of lean or agile are
scarce. This paper aims to add to that body of
knowledge by providing a descriptive case study of
how one company, Paf, introduced portfolio
management to help support scaling agile software
development. We also report on initial results and
challenges met.
The paper is structured in the following way. First
we present related work on portfolio management from
literature on NPD and especially literature on agile and
lean software development. Second we describe the
research methodology and provide case company
background information. In Section 4 we provide the
results of the case study. Finally, we sum up the paper
with its contributions and suggestions for future work.
2. Related work
2.1. Portfolio management in new product
development
The concept of portfolio management is not new. It
has appeared over the decades under various names,
such as R&D project selection, prioritisation or
resource allocation. The focus of research has been in
the area of new product development (NPD) and on
developing quantitative techniques and methods for
project evaluation, selection and prioritisation. [4]
Portfolio management has been adopted in the
industry, but not without problems. A recent study of
30 companies [2] shows that while the companies have
adopted portfolio management practices, they still
struggle with completing projects within schedule and
lack a broad overview of ongoing projects. The main
reasons behind this were; (1) very different types of
projects are included in the managed portfolio, and (2)
not all projects and smaller activities are managed as
part of the portfolio. Therefore as much as 50% of the
product developers’ time can be allocated to work that
is not seen by the portfolio managers. This creates a
gap between required and available resources that can
go unnoticed. Since portfolio management is
essentially about resource allocation, the inherent
complexity of the issues involved keeps it far from
being a mechanical exercise.
Successful portfolio management is about
achieving a balance between the four potentially
conflicting goals of 1) maximising the financial value
of the portfolio, 2) linking the portfolio to strategy , 3)
balancing it on relevant dimensions, and 4) ensuring
that the total number of ongoing activities is feasible .
Different portfolio management techniques, such as
financial and economic models, scoring models and
mapping approaches emphasise these goals differently.
[5, 17]
Setting up proper portfolio management has been
recognized as challenging even for the best of
organizations [19]. In practice, portfolio management
is often realised through integrating techniques for
project evaluation, selection and prioritisation with a
phased review process for ongoing development
projects [5, 17, 34]. Phased review processes organise
product development projects into a sequential set of
phases having themes (beginning with idea generation
and ending with product launch and maintenance),
with a corresponding business and prioritisation
decision point (i.e., the review) at the end of each
phase. Development work is conducted during the
phases, along with gathering the information needed to
pass the next review. [4, 6]
Literature describes two basic alternatives for
implementing the portfolio management process in
practice. The first one emphasises decision-making
through in-depth reviews for each ongoing project and
manages the portfolio in a bottom-up manner (e.g.
gates dominate [5] and model I funnel [34]). The
second is top-down, with decisions based on looking at
the portfolio as a whole (e.g. portfolio reviews
dominate [5] and model II funnel [34]). The first
mentioned is better suited for larger firms in mature
businesses with dedicated resources and fairly static
portfolios, because the emphasis is more in making
sound go/kill decisions for individual projects than re-prioritising the entire portfolio every few months.
Likewise, the second is more appropriate for smaller
companies in fast-paced and fluid markets because it
allows for more dynamic resource allocation through
periodically reviewing the entire portfolio. [5, 34]
2.2. Portfolio management in agile and lean
software development
Advocates of agile methods tend to see much of the
literature on managing new product development as
fundamentally incompatible with agile software
development because it tends to view development as a
separate “phase” in the cycle of realizing an idea into
an actual working solution [12]. For example, phased
review processes as described in Section 2.1 are by
some seen as incompatible with the basic principles of
agile development [14, 15]. Whatever the case,
integrating agile software development with phased
review processes may not be straightforward due to
attitudes, as well as a number of actual contradictions
in e.g. level of abstraction in planning and feedback
cadence [10].
Not much has to date been written on how to set up
portfolio management so that the result would
compatible with agile and/or lean principles. Still, the
ideas and practices of portfolio management have
begun to emerge also in agile software development,
mainly through the introduction of lean principles to
software development, see e.g. [11, 12, 14, 15, 22, 24,
25]. However, the use of the term portfolio
management differs between the authors.
Leffingwell provides in his book [16] and its blog
companion (scalingsoftwareagility.wordpress.com) a
very high-level view to portfolio management. At the
portfolio management level the company’s executives
define investment themes that drive the resource
allocation and thus investment priorities of the
company. Epics or epic-scale initiatives are used to
express the portfolio vision in practice. These guide
upcoming product releases.
Poppendieck & Poppendieck [22] provides another
viewpoint to what could be interpreted as an approach
to portfolio management. Possible development efforts
that take up people’s time are first classifies by type,
for example as strategic business initiatives, feature
upgrades, infrastructure upgrades, and maintenance.
Then the desired cycle time for each type of effort is
decided. The investment levels for the categories are
set by determining how many initiatives of each type
should be carried out within, e.g. a year. Or, in the case
of e.g. maintenance, a reservation is made of how
much of the total capacity the activity should be
allowed to take. Finally, the slots for the initiatives are
laid out in the calendar in advance. When the time slot
of a certain initiative approaches, its actual content is
decided on the basis of what is most valuable for the
company at that point in time.
The most common interpretation of portfolio
management in software development follows the
definition from NPD literature; portfolio management
is the activity of making resourcing decisions across a
portfolio of planned and ongoing projects. Shalloway’s
concept of Lean Portfolio Management [25] means
deciding on a frequent basis across a portfolio of
projects how the development resources are allocated
for delivering the minimum marketable features or
business features that at the moment provide the most
business value. Pichler [20] agrees with Shalloway’s
idea and recommends that “competing backlogs”
should be dealt with in a similar fashion.
Larman and Vodde [14, 15] are along the same
lines. They note that in organizations with less than
100 people, prioritizing on the level of the portfolio of
products and services offered tends to lead to local
optimization. Instead, portfolio management can be
more effectively carried out by merging the backlogs
of different product/service offerings into a single
backlog, and then performing backlog management on
that single backlog. This view is shared by Krebs [12],
who advocates that the ongoing and planned projects
should be kept in a list called the “project portfolio
backlog”. Decisions on which projects should
continue, be put on hold, be launched, or be killed, are
then made on a per-sprint basis. A similar approach is
also mentioned by Rothman [24]. These approaches
seem to assume that the iterations are synchronized for
the whole organization and that all projects use an agile
software process. An off-synch portfolio of projects
can indeed lead to problems, as pointed out in [33].
Only a few experience reports seem to exist to date.
Karlström and Runeson [10] report experiences on how
two large software systems projects attempted to use
eXtreme Programming in the context of a phase-gate
model. Tengshe and Noble [30] report how the
Portfolio Management Office (PMO) helped balance
the demand on Capital One Auto Finance’s resources
from multiple and sometimes inter-dependent projects.
Kalliney [9] reports how Ultimate Software
transitioned from agile development to an agile
enterprise, setting up portfolio management practices
to solve some of the problems related to the transition.
Thomas and Baker [31] report on what challenges they
faced when applying agile methods to IT investment
funding, change management, and governance, and
how they managed their projects as a portfolio. Steindl
[26] reports how IBM manages agility on three levels;
project, portfolio, and business. Laanti [13] describes
how a large organization implemented portfolio
management as an “outer control loop” on top of the
“inner scrum control loop” according to the Sashimi
model from [18].
3. Methodology
3.1. Research methods
This study is a longitudinal, descriptive case study
[35], with elements of action research [27]. The first
and third author and other researchers have observed
the portfolio management process at Eget and later
Paf.com from the beginning of 2008. The second
author has worked in the company since August 2008
and he has been responsible for defining the Paf model
for projects (Pamp) presented in Section 4, and for
further refining the portfolio management approach.
The researchers arranged a survey and interviews
with selected personnel in March 2008 to investigate
the state-of-practice of portfolio management at the
company. The survey contained 56 statements related
to different aspects of portfolio management and it was
answered by 21 respondents. The sampling was
purposeful. Key personnel from a diverse assortment
of roles and responsibilities (developer and tester
representatives, R&D management, business
representatives, Product Owners (PO’s), Scrum
Masters (SM’s)) were selected to participate in the
survey. Of those 21 people, 7 were further interviewed,
using the survey statements as an interview guide. The
results of the survey and interview were disseminated
to the whole company two weeks after the interviews.
After the dissemination of the results of the survey and
interviews, the researchers participated as observers in
different meetings relating to the development process
and later the portfolio management process of the
company. They also reviewed documents related to the
existing and planned processes and provided feedback.
In September 2010 the survey was re-run, with 28
respondents of which 6 had also answered on the first
round. However, new interviews based on the second
survey round were not made before the publication of
this paper.
3.2. Case background
Paf (Ålands Penningautomatförening), founded in
1966, is a public association that operates gaming
activities on the autonomous Åland Islands, onboard
Ålandic and Finnish ships and on the Internet. Gaming
began on 1
st
January 1967 in collaboration with the
member associations behind Paf: the public health
service on Åland, the Åland branch of Save the
Children, the Finnish Red Cross and the child welfare
foundation Stiftelsen Dagens Barn. This study is
focusing on the software development of the Internet
gaming activities. Internet gaming on www.paf.fi was
launched on 3
rd
December 1999. The first form of
gambling was betting. Today, the business also
encompasses gaming machines, casino gaming, bingo
and lotteries, poker, and skill games.
Paf founded the subsidiary Eget focusing on
internet gambling in 1999. Paf held the majority of
Eget shares. Eget developed and operated Internet
monetary games for Paf and other companies. In the
spring of 2008 Paf and Eget merged but the setup of
buyer and supplier was still around in the minds of the
employees. Later, in the spring of 2009, most of the
former Eget was to be known as Paf.com, the Internet
department at Paf.
Scrum was introduced at Eget around 2006 to solve
problems related to quality, release planning and ways
of distributing work. Scrum was introduced bottom-up
starting with one team. In the fall of 2008 most teams
used Scrum “by the book”. At that time the visibility
into teams’ backlogs was established, but the backlogs
and other documentation of 10+ geographically
distributed teams were spread all over the company
wiki and in Excel files with no structure. However, the
main problem was the visibility into what PO’s and
development teams actually did, which was a result of
unclear and shifting priorities from the Marketing and
Sales department due to congestion of the development
pipeline. For example, if the capacity of the
development pipeline was 5 projects, 5 projects were
started, but before these projects were ready and the
deliverables were released into production, 5 new
projects were started, and so on.
This congestion and thus shifting priorities caused
thrashing, which resulted in long time-to-market. This
could also be seen as Development-Business
disconnect and inefficient and uncontrolled portfolio-level decision making. These had been identified as the
two most pressing challenges in a survey conducted by
the first author and his researcher colleagues in the
spring of 2008, when the collaboration with the
company and researchers began.
In a value chain mapping exercise performed by the
developers at Eget, the outcome was that even if a
game took 5 months to develop on average, the worst
case scenario was that it would go into production in a
total of 24 months with 19 months of shelf time and
handovers. It was also proven that a simple game can
be put out into production in 3 months when expedited
by C-level managers. The situation in the fall of 2008
was that releases were planned five weeks ahead as a
release train [16] and the practice of joint release
planning [8, 16] had been taken into use, to make the
teams’ plans visible to all the other teams and
stakeholders and to reduce handovers.
The problems found are similar to symptoms that
have been identified in literature to be associated with
inadequate portfolio management [32]. Therefore
setting up practices for portfolio management was
selected as one of the improvement paths.
Thrashing was not the only reason for long time-to-market. A monolithic architecture that had evolved
over time was also a major contributing factor. Due to
the architecture and history, Paf.com operates with
component teams: Slot team, Lotteries & Bingo team,
Integration team, BackOffice team, Core team, Report
team, CRM team, Web management team and so on.
Build automation was introduced and further
developed to mitigate the effect of the complex
architecture and help coordinate the work of the 10+
geographically distributed teams. However, the impact
of architecture and build automation or changes in
team structures are excluded from this study that
concentrates on the initial experiences of introducing
portfolio management practices to alleviate the
problem of unclear and shifting priorities and
thrashing.
4. Introducing portfolio management
4.1. First steps towards portfolio management
The first steps towards portfolio management were
taken when the content of every release was internally
prioritized according to business value. In every
release there were 3-5 upgrades or new components
that were committed to be ready to be put into
production by the teams. Some effort was spent on
training the Marketing and Sales department in the
ways of agile planning and prioritizing, i.e. the most
important things need to be completed first, so that less
important things can be scoped out at the end of the
release time box, if need be. Gradually this began to
work but then 4 new problems were discovered:
• lack of visibility on “projects” about to enter
the development pipeline
• the maturity of the planning of the “projects”
entering the development pipeline
• prioritisation of projects was ad-hoc
• rogue, “business critical” projects were
induced “under the hood” by some managers
These problems were discovered in the fall of 2008,
when the newly created Project Management Office
(PMO) made an inventory of all the active projects in
Paf development and technical operations (later
Paf.com). As a result, 214 ”projects” were found,
including duplicates and sub-projects. To get visibility
into the situation, the PMO started to issue unique
project ID’s to existing and new projects. However, it
was difficult to get grasp of the existing projects, since
they were unstructured and very little documentation
was available.
In the spring of 2009 the management of Paf was
still not happy with the inefficiency and handovers in
the organization inherited from the merger of Paf and
Eget. Cooperation negotiations were held to get a
competence shift towards the Internet business. The
guiding star of the new organization was less
handovers. In the new organization the departments
System & Services, Games, Sales, Customer
Experience and Customer Relations were formed into
Paf.com with responsibility for the Internet gaming
business. The component/development teams were
divided under the appropriate departments. The idea
was that problems and opportunities discovered in a
department could be handled by its development
teams. Problems and opportunities that are dependent
on resources outside the own department were to be
handled in a structured manner according to the
suggested new project framework Pamp, which is
discussed in the next section.
4.2. Paf model for projects (Pamp)
In the spring of 2009, the PMO and QA proposed a
model for managing the portfolio of projects in a
structured way. It was called Pamp and it combined
elements from Stage-Gate models [7], the Open
Unified Process (http://epf.eclipse.org/wikis/openup/),
and PMBoK [21].
The main goal of Pamp was to clarify a business
need and increase the visibility for each project in
Paf.com. The idea was to tackle the problems
identified earlier and described in the previous section.
Pamp was not to include guidelines for the actual
project execution process, which was left for the
team(s) and project manager to decide. Typically, the
execution was done using the existing Scrum process.
Figure 1 shows an overview of Pamp. Pamp
controls and monitors project planning, prioritization,
and execution through 4 project demos (Proposal,
Planning, Design, Closing), where the person having
the role of project manager presents the status of
planning and execution. The time units in Figure 1 are
indications of calendar time (the size of which may
differ, depending on project size); most of the effort is
done in the Execution phase by the development
team(s). The level of required documentation and
planning is governed by a complexity classification of
the projects into complex, normal, or simple. The
demos are held as a review and workshop with
representatives from the project organization and
senior project and process managers reviewing the
documentation. The emphasis is on clarifying risks,
dependencies, stakeholders, and business value.
In the project proposal demo, the project manager
presents, e.g. one structured PowerPoint slide including
information on:
• Business issue/opportunity
• Primary project deliverable(s) or Epic
statement
• Business benefits of the project outcome
• Project overview
• Main risks and dependencies
If the proposal is accepted, the project is put into
the Paf.com backlog, which is further explained in the
following section. The first draft of the project plan is
prepared and presented in the project planning demo,
addressing the work that needs to be undertaken:
Figure 1 Paf model for projects (Pamp)
Proposal
• Project Proposal
• Project Charter
• Epic statement
Proposal
Demo
Planning
• Who?
• Why?
• What?
• When?
Design
• Estimation
• Roadmap
• Dependencies
• Mock up
Execution
• Development
• Integration
• Te sting
• Release
Closing
• Retrospective
Final report
in 0 to 6
months
Time to Market Production &Maintenance
10 time units 20 time units 70 time units
Planning
Demo
Design
Demo
Closing
Demo
WHAT WHAT
WHAT
HOW
HOW
Chosen Process
e.g. Agile Framework, Business Case,
Construction process, Summer vacation project
• Who will primarily be doing the work
• What are the main deliverables and project
scope
• When is the target release date
• Why the work is done, including more refined
business and other benefits compared to the
project proposal
After the planning demo, project preparations move
on to the design demo, where the plan for how the
project will be executed is presented, including:
• Roadmap of preliminary epics and sprints
• Initial estimations of the epics
• Resource needs
• Dependencies
• Mock-up to show technical feasibility in more
complex projects
The final, closing demo is arranged after the project
has ended. The project’s success is reviewed and a
retrospective is held.
During the project execution period, the follow-up
is governed by the approved project plan, which is a
living document that is updated and “re-approved”
throughout the project. While this sounds like a plan-driven approach, it is still similar to the roll-out
planning of Scrum. The extra planning is needed for
coordination when scaling agile to development
projects involving several teams, especially if the
teams are component teams and a switch to feature
teams is not feasible in the short run. The teams still do
the actual value-adding work using Scrum and burn-down charts are used to evaluate progress. If the burn
down shows problems in reaching the desired scope,
corrective actions are taken as deemed appropriate, and
a re-evaluation of project value can be performed. In
this way the ranking of the project might fall (or rise)
in the portfolio review, arranged once a month.
The project manager (PM) is a role that a PO can
have during a project. When a project needs multiple
teams, a project manager is appointed to handle
project-related matters not related to the PO work, such
as budgeting for multiple teams, handling reporting to
the steering group, and facilitating the communication
and problem solving between the PO’s in the project.
For a cross-departmental project the final say is with
the PM unless it is escalated to the project owner
(usually a department head). In a typical setup the PO
that is most central in the project and has the biggest
need to get the project done is made PM for the project
and through the Paf.com backlog prioritisation
(described in the following section) (s)he gets the other
PO’s support and resources. Usually the PO appointed
as PM does not have full knowledge about the domain
that the supporting teams are developing. Therefore the
PO responsible for that domain is handling the backlog
for that supporting team.
The effort of using Pamp for giving visibility to the
work and helping in the planning is small compared to
the situation of not having information for portfolio
management. In general, writing the proposal takes a
couple of hours and preparing the plan for the project a
couple of days. Feedback has shown that it is faster if
the project manager has a clear view on the work that
has to be undertaken. The proposal and the plan for the
project are based on templates that also function as
checklists for planning the work and presenting
possible business value for Paf.com management. The
rule of the thumb is that the project manager has a lot
of freedom in designing the work as long as the highest
risks are mitigated to a certain extent and the way of
working is documented. Statements like “according to
company process X”, “will be clarified in week 5 of
the project” or “according to team estimates” are valid.
The demos take approximately 1 hour or less, if the
project manager has prepared herself properly. The
senior experts invited to the demo ask the project
manager to clarify questions around the main points of
the project plan. This is done to make sure everybody
understands what the project is about, so that the
portfolio decisions are based on the best possible data.
Next we take a look at the Paf.com backlog
process, which is the portfolio management process of
Paf.com.
4.3. Paf.com backlog process
The backlog process was first drafted in 2008 and
in the beginning of summer 2009 the current version
shown in Figure 2 was taken into use. The process
addresses the problem of managers inducing “business
critical” projects “under the hood”. All Pamp projects
are in the Paf.com backlog. Also projects that are not
using Pamp are visible. Projects not using Pamp are,
for instance, new games, security patches, or general
updates, which are done from the team’s or
department’s own backlog. The Paf.com backlog is a
simple table placed in the intranet of Paf.com. It shows
project priority, ID, name, sponsor (Paf.com
management representative), manager (person taking
the role of project manager for that particular project),
status, estimated completion date, and capitalization
information. Finance is monitoring the Paf.com
backlog to follow up on costs from projects and
released value-adding content.
The 5-10 most important projects are prioritized,
out of roughly 30 (which is considerably less than 214
in the fall of 2008), by Paf.com management
(consisting of department heads and the Paf.com
director) on a monthly basis in a portfolio review
meeting, with the PMO owning the prioritization
backlog called Paf.com backlog. The reason for
Paf.com management to prioritize the projects is that
they control the resources and need to commit to
assigning their resources to a specific project. The
person having the role of project manager can and is
obliged to ask for resources from projects ranked lower
in the Paf.com backlog to speed up project completion.
The driving force is to concentrate on a few things at a
time so that projects can be done fast to deliver value
and thus help achieve shorter time-to-market. This is in
accordance with lean principles [23].
Figure 2 Paf.com backlog process
Resourcing can be done by giving work to other
teams or by taking individuals into the team working
with the project with a higher priority. Company-wide
changes to platforms or processes are prioritized in the
Paf.com backlog in the same way so teams know what
to focus on in the company sprint of 5 weeks (release
train) and what is expected from them when the release
is done. Joint company sprint planning is used as a
practice to make it possible to more easily negotiate the
resources for each company sprint. The teams and
individuals can easily block requests that are not
connected to a higher priority project. If there are no
requests, the team can work on their assigned project
even if it is not prioritized. The teams’ sprints
(typically 2 weeks) are not synchronized and every
team has the freedom to plan their work within the
company sprint as they see fit. Only those teams that
have their own component in an upcoming release
participate in the releasing activities. On average
during 2009, 8 out of 16 teams were involved in
making a production release. The smallest amount of
teams participating in making a production release was
4 and at one occasion 12 teams participated.
Figure 3 shows the backlog hierarchy at Paf.com.
Pamp and the Paf.com backlog process are only
concerned with the two left-most backlogs, which
provide input to the teams’ backlogs. You could also
say that the Paf.com backlog and the project backlog
show business priorities and high-level plans, whereas
the teams’ backlogs contain the actual development
work.
Figure 3 The backlog hierarchy
The PMO monitors project progress based on the
teams’ story point estimates and burn down charts, and
reports to Paf.com management as part of the portfolio
follow-up function. Story point estimates reflect the
complexity of the story to be implemented and do not
account for how long it takes to implement the story.
The estimation is done by the teams using standard
planning poker with Fibonacci series numbers. Story
points per sprint, normalized by sprint length gives an
approximation of velocity from which a forecast of
project completion can be derived, when the story
points of unfinished stories are known. The forecast is
calculated based on the mean value of the story points
in the last three completed sprints.
When several teams work on the same project in
different sprint schedules with different velocity and
differently sized backlogs, the forecasts are made on
team level. The team with the completion forecast
furthest away is used as the current forecast for
completion of the project. The forecast gets more
uncertain as the complexity of the project grows, i.e.
more teams, new technology, or complex solutions are
involved. Forecasted date and targeted date are not to
be mixed. The difference is basically how a project
was planned to go and how is it going time wise. The
forecast can be used like a project burn down,
signalling when there is a need to re-scope the project
or extend the schedule of the project.
A web-based backlog management tool was
introduced in early spring 2009 for all the teams to
gradually take into use as their primary reporting tool
to show the team’s stories and their progress in respect
to the release train. In the fall of 2009 all teams were
using the tool. The aim was to provide a unified way of
measuring and visualizing progress of work and effort
left undone in the projects. The information is used as
background in the portfolio review. From a
development project’s viewpoint all necessary data is
available at a glance; amount of story points done and
not done, velocity, and forecast. Additionally, the size
of releases of the release train can be estimated based
on complexity, not time spent on coding. SM’s and
PO’s are the main users of the tool. The PO’s keep
their backlogs in the tool and SM’s update progress
information, taking them an average of about 5 minutes
Paf.com
Backlog
Paf.com management
Portfolio follow-up function
Release train
PMO
Paf.com backlog
process
Project
backlog:
Epics
Paf.com
backlog:
Projects
Team(s’) sprint
backlog(s):
Stories and
tasks
Release
content
per day. The main tool for internal communication and
information is the team’s own Scrum wall in the team
room. One of the main reasons for additional, web-based reporting is that the teams are geographically
dispersed to seven offices in four countries. The
information that is needed, e.g. to plan and coordinate
the release train, cannot thus be collected just by
walking through all the team rooms and reading the
Scrum walls.
4.4. Initial results from introducing portfolio
management
While explicit portfolio management has only just
been introduced at Paf.com and the investigation to its
possible drawbacks and benefits is still ongoing, we
can provide some initial results. As already mentioned,
the amount of ongoing projects has gone down with
one order of magnitude from 214 to 30. Even if the 214
projects included sub-projects and duplicates, the
reduction is considerable and can be attributed to the
introduction of Pamp, the restructuring of the backlog
process, and improved planning practices, which also
have helped in reducing handovers. Pamp is working
as a gate for the paf.com backlog, guarding it from too
many or too immature projects.
Additionally, from those 30 projects, only 5-10 at a
time are indicated as high-priority, making the
priorities clear to the whole organization. This drives
the way work is organized and accepted by the teams,
and the team members have already shown that they
now can refuse to take work of lower priority, instead
of ending up in a thrashing mode of constantly
changing priorities. Otherwise Pamp has not directly
affected the way teams work, and they can still use
Scrum as defined in the company.
The portfolio follow-up function and the Paf.com
backlog provide visibility into what is currently
ongoing at Paf.com. The process enforces a clearer and
more specific definition of project goals and business
value. This helps managers understand what is going
on in other departments and the business value of the
work. New opportunities can still be utilized fast
enough. It takes at the longest 5 weeks to get a new
project accepted into teams’ sprint planning. Even
though the process gives room for expediting, there is
no record of terminating a sprint due to demands from
a higher-ranked project. Making things more visible
has also helped with this, because now managers can
negotiate with each other in preparing a business case
for a new opportunity and they can make an informed
decision based on the facts of all the ongoing projects.
In this way, a less important project can be put on hold
and a new project of higher business value can replace
it in a controlled manner.
The structured process for proposing new projects
has also dramatically reduced the amount of “Do it, it
doesn’t matter what it costs and how much time it
takes, because I want it!” projects in the portfolio.
These “pet projects” could be seen 2 years ago (if one
knew where to look for them), but today none exist as
far as we know.
Still, not everything is run as a project and thus is
not part of the Paf.com backlog or the portfolio
management process. For this, there is a clearly
defined and accepted, company-wide prioritization
based on work types. All production problems, i.e.
problems in the customer-facing, live products, get first
priority. Second priority is on all work that needs to be
done to be able to make the next release. During 2009
10 production releases were made and more than 20
patches, meaning a response to production issues, were
deployed into production.
These two work types are allowed to thrash work of
the third priority, work based on the priority order of
the Paf.com backlog. If time still remains, other work
can be performed. In general, one team is working on
one project backlog with additional stories or tasks
from company-wide improvement work, such as
upgrading or migrating infrastructure or improving the
logging functionality of the component they are
responsible of. If somebody tries to circumvent this
priority order of work, they are stopped or the issue is
escalated. However, as noticed in [2], this could be a
source of problems because the availability of
resources at any moment is unknown, and thus the
situation needs to be monitored in the future.
To summarize the initial positive results, it seems
that the problems that were tackled by introducing
Pamp and portfolio management have been solved
quite satisfactorily. However, it is still too early to
conclusively say how much time-to-market has
improved. The second survey results show
improvement in the areas identified as the biggest
challenges. Also, the areas of over commitment and
multitasking show improvement, which is in line with
the goals for introducing portfolio management.
4.5. Improvement needs and ideas
While the initial feeling and reaction has been
mostly positive, some improvement needs and ideas
have already surfaced. According to the second author,
one of the most challenging things related to the
introduction of portfolio management and Pamp has
been to make people understand the separation of the
value-adding work in a project from the planning of the
project. The conceptual mix of having an overall agile
framework for development work together with more
“traditional” project management practices and
portfolio management for planning the work can
admittedly be confusing. The planning is needed to
help prioritize and coordinate the work, especially
when several teams work on the same project. From an
agile standpoint one could argue that the “right” way to
go would have been to create true feature teams. But
the agile ideal is far from easy to gain in a multi-team
environment. Portfolio management has been shown
both in theory and practice to be one good alternative
for helping in scaling agile. Therefore it was a valid
choice to be tried out in our opinion.
Developing a training package for project
management in the context of Pamp and the Paf.com
backlog process could help mitigate the lack of
understanding. Furthermore, a forum for PO’s is being
started, where they can interact, learn from each other,
and solve problems together. This forum will later be
officially added to the Paf.com backlog process after
the test-run has been completed and the retrospective
has found it useful in practice.
There is still work to be done to get good enough
progress data out of evaluation projects, or spikes in
XP-terminology [1]. Currently the data is a subjective
opinion communicated in progress reports distributed
to the whole organization by e-mail. Demos are also
used, but mainly for the stakeholders of the evaluation
project, which makes it hard to see the impact to other
projects and overall resourcing when making portfolio-level decisions.
5. Contribution and future work
The contribution of this paper is twofold. First, we
describe how Paf.com has introduced portfolio
management to help them scale agile software
development. We provide a rich description of the
introduction of portfolio management and initial
results. While we cannot generalize the results based
on only one case company, the results may still inspire
other companies into considering portfolio
management as an option to help in scaling agile. The
descriptive case study adds to the body of knowledge
of portfolio management
Second, the condensed section on related work,
especially Section 2.2, provides a good starting point
of references for anyone interested in adopting or
improving portfolio management in their agile
organization. While the list of references is by no
means exhaustive, we feel that it is a good sampling of
the current trends in portfolio management, especially
regarding agile ways of working.
Since this work only reports on the initial results
and experiences, we intend to continue monitoring and
improving the portfolio management practices and
process. Part of this work is conducting the interview
for the second round of the survey and analyzing and
disseminating the results.
We are also planning to write a case study on the
biggest project in Paf.com history, ending in the fall
2010. That would also help fill the gap left in this
paper as to describing in more detail the whole
development process, including the project and
portfolio management processes presented in this
paper.
While specific techniques for portfolio
management, such as valuation of projects/epics, were
left out of scope of this paper, there is still a need to
study which of these exist or could be applicable for
portfolio management in a lean or agile context. This
could include, for example, studies of IT investment
analysis, real options, and other IT governance models.
Also, other competing models for scaling agile should
be reviewed and a comparative analysis of different
models would provide value in helping companies
choose suitable scaling strategies and practices. In
general, more case studies and industry experience
reports on portfolio management are needed to help
industrial actors find better ways of working.
6. References
[1] K. Beck, EXtreme Programming eXplained. Addison-Wesley, 2000.
[2] B. S. Blichfeldt and P. Eskerod, “Project portfolio
management – There’s more to it than what management
enacts,” International Journal of Project Management, vol.
26, pp. 357-365, 2008.
[3] A. Cockburn, Agile Software Development. Pearson
Education, 2002.
[4] R. G. Cooper, S. J. Edgett and E. J. Kleinschmidt,
Portfolio Management for New Products. Perseus Books,
2001.
[5] R. G. Cooper, S. J. Edgett and E. J. Kleinschmidt,
“Portfolio management: Fundamental to new product
success,” in The PDMA Toolbook for New Product
Development$ $, P. Belliveau, A. Griffin and S.
Somermeyer, Eds. New York: John Wiley & Sons, Inc, 2002,
pp. 331-364.
[6] R. G. Cooper, Product Leadership: Creating and
Launching Superior New Products. Perseus Books, 1998.
[7] R. G. Cooper, “Perspec tive: Third-Generation New
Product Processes,” JPIM, vol. 11, pp. 3-14, January, 1994.
[8] V. T. Heikkilä, K. Rautiainen and S. Jansen, “A
revelatory case study on scaling agile release planning,” in
Proceedings of the 36th Euromicro/SEAA, Lille, France,
2010, pp. 289-296.
[9] M. Kalliney, “Transitioning from agile development to
enterprise product management agility,” in Proceedings of
AGILE 2009, Nashville, USA, 2009.
[10] D. Karlström and P. Runeson, “Integrating agile
software development into stage-gate managed product
development,” Empirical Software Engineering, vol. 11, pp.
203-225, 2006.
[11] P. Kettunen, “Adopting key lessons from agile
manufacturing to agile software product development—A
comparative study,” Technovation, vol. 29, pp. 408-422, 7,
2009.
[12] J. Krebs, Agile Portfolio Management. Redmond, WA,
USA: Microsoft Press, 2009.
[13] M. Laanti, “Implementing program model with agile
principles in a large software development organization,” in
2008, pp. 1383-1391.
[14] G. Larman and B. Vodde, Practices for Scaling Lean &
Agile Development: Large, Multisite, and Offshore Product
Development with Large-Scale Scrum. Boston, MA, USA:
Addison-Wesley, 2010.
[15] G. Larman and B. Vodde, Scaling Lean & Agile
Development: Thinking and Organizational Tools for Large-Scale Scrum. Boston, MA, USA: Addison-Wesley, 2009.
[16] D. Leffingwell, Scaling Software Agility: Best Practices
for Large Enterprises. USA: Addison-Wesley, 2007.
[17] M. McGrath, “Setting the PACE in product
development,” 1996.
[18] I. Nonaka and H. Takeuchi, “The New New Product
Development Game,” Harv. Bus. Rev., vol. 64, pp. 137-146,
1986.
[19] P. O’Connor, “Spiral-up implementation of NPD
portfolio and pipeline management,” in The PDMA Toolbook
for New Product Development , P. Belliveau, A. Griffin and
S. Somermeyer, Eds. New York: John Wiley & Sons, Inc.,
2004, pp. 461-492.
[20] R. Pichler, Agile Product Management with Scrum:
Creating Products that Customers Love. Boston, MA, USA:
Addison-Wesley, 2010.
[21] PMI, A Guide to the Project Management Body of
Knowledge : PMBOK Guide. Newtown Square, PA, USA:
PMI, 2008.
[22] M. Poppendieck and T. Poppendieck, Leading Lean
Software Development: Results are Not the Point. Boston,
MA, USA: Addison-Wesley, 2010.
[23] M. Poppendieck and T. Poppendieck, Implementing
Lean Software Development: From Concept to Cash. Boston,
MA, USA: Addison-Wesley, 2007.
[24] J. Rothman, Manage Your Project Portfolio: Increase
Your Capacity and Finish More Projects. USA: Pragmatic
Bookshelf, 2009.
[25] A. Shalloway, G. Beaver and J. R. Trott, Lean-Agile
Software Development: Achieving Enterprise Agility. Boston,
MA, USA: Addison-Wesley, 2010.
[26] C. Steindl, “From agile software development to agile
business,” in Proceedings of 31st Euromicro/SEAA, Porto,
Portugal, 2005, pp. 258-265.
[27] G. I. Susman and R. D. Evered, “An Assessment of the
Scientific Merits of Action Research,” ASQ, vol. 23, pp. 582-603, December, 1978.
[28] J. Sutherland, G. Schoo nheim and M. Rijk, “Fully
distributed scrum: Replicating local productivity and quality
with offshore teams,” in Proceedings of the 42nd Hawaii
International Conference on System Sciences, Hawaii, USA,
2009.
[29] J. Sutherland, A. Viktor ov, J. Blount and N. Puntikov,
“Distributed scrum: Agile project management with
outsourced development teams,” in Proceedings of the 40th
Hawaii International Conference on System Sciences,
Hawaii, USA, 2007.
[30] A. Tengshe and S. Noble, “Establishing the agile PMO:
Managing variability across projects and portfolios,” in
Proceedings of AGILE 2007, Washington D.C., USA, 2007.
[31] J. C. Thomas and S. W. Baker, “Establishing an agile
portfolio to align IT investments with business needs,” in
Proceedings of Agile 2008, Toronto, Ontario, Canada, 2008,
pp. 252-258.
[32] J. Vähäniitty, K. Rautiainen and C. Lassenius, “Small
software organisations need explicit project portfolio
management,” IBM J. RES. & DEV., vol. 54, pp. 1-12,
March/April, 2010.
[33] J. Vähäniitty and K. Rautiainen, “Towards an approach
for managing the development portfolio in small product-oriented software companies,” in Proceedings of the 38th
Hawaii International Conference on System Sciences,
Hawaii, USA, 2005.
[34] S. C. Wheelwright and K. B. Clark, Revolutionizing
Product Development. New York: The Free Press, 1992.
[35] R. K. Yin, Case Study Research: Design and Methods.
London: SAGE Publications, 1994.
Place this order with us and get 18% discount now! to earn your discount enter this code: special18 If you need assistance chat with us now by clicking the live chat button.