Going Agile – A Case Study

1 | P a g e

Going Agile – A Case Study

Dwayne Read

Software Process Consultant

Strategic Systems

dwayne@ss.com.au

Grey Properjohn

Systems Analyst

Snowden Technologies

gproperjohn@snowdengroup.com

Abstract

This case study examines the approach undertaken by Snowden Technologies in adopting a

range of agile techniques into their product and custom software solutions for the mining

industry. The catalyst to adopt various agile techniques stemmed from the growth of the

development team (30+) and the need to further integrate the development activities across

multiple offices. This required a clearer and more consistent approach to the software

development process.

Snowden Technologies involved the users (Developers, Project Managers, Team Leaders,

Geologists, etc) upfront and proritised the project’s scope resulting in the selection of specific

techniques from a range of agile methodologies (XP, FDD, Scrum, Crystal, DSDM, RAD/JAD).

Implementation of the techniques paralleled various agile principles through the use of scoped

releases, development iterations and feedback through regular reflections. The selected agile

techniques (e.g. iterative development, Domain Object Models, code inspection, automated

testing) were incorporated into a consistent software process with various “value add”

techniques (e.g. CUT complete, code coverage, code analysis) that integrated with the

development tools.

The key lessons learnt from the this approach were that to incrementally introduce agile

techniques was very effective; embedding the process elements into the development tools

helped to reinforce the techniques; and it proved that the agile techniques and PRINCE 2 for

project management could be customised to collaborate into an effective solution. The latter

required some compromises from the “pure” agile view such as to allow a project schedule to

define all Work Packages (Iterations) at the start of each Stage (Releases).

1. Keywords

Agile, methodology, software process, change management, project management, brain-

storming, reflection, iterative development, JAD, Domain Object Model, architecture, code

inspection, automated testing, Team Foundation Server, TFS, PRINCE 2.

2. Background

Snowden Technologies is a division of the Snowden Group that provides global consultancy

services and software solutions to the mining industry, and has a development capacity of over

thirty developers located in Perth (head office), Brisbane and Johannesburg. The reliance on

delivering quality software solutions to its clients has continued to increase over recent years to

the point that being able to rapidly respond to the needs and unique/specific requirements of

clients is a critical success factor to the business.

2 | P a g e

3. Motivation to Go Agile

The driving force to initiate and implement the new agile techniques originated from

expansion of Snowden’s development capacity which presented the necessity for a clearer

defined and (multi) team oriented approach to its software development activities. Key areas

identified by Snowden developers and managers that were targeted as part of this approach

were:

1. Specification and scope definition

2. Estimation (essentially reducing the non-billable hours)

3. Minimising defects and instigating code quality reporting

4. Establishing standard and consistent practices across teams and offices

5. Improving system deployment and build processes

6. Providing feedback mechanisms (within and across teams)

7. Initiating mentoring programs (sharing skill sets)

The need to address these key software development areas were being compounded by

the ongoing growth of the team (in a time of very competitive recruitment) and the need

to improve the Project Management process. Four separate improvement projects had

been identified by the business as focus areas, one of which was the Agile Project (the

other three were Project Management (PRINCE 2), SPI (the Snowden Project Index

reporting tool), and Team Foundation Server (TFS)). The intention was that the Agile

Project would be able to address the majority of the identified key areas by selecting and

tailoring the techniques from a range of industry proven agile methodologies. The

techniques would define the “Snowden Technologies Agile” methodology, herein referred

to as STAg.

4. Approach

The starting point was a brainstorming session (based on the typical Cause-Effect diagram

technique) conducted with all developers to examine the causes behind the identified

problems. As a facilitated brainstorming session, all participants contributed their ideas,

of which, 99 tentative causes were identified in the session. The “top 10” causes were then

examined in more detail in terms of the impact, influence, cost, effort and duration.

Figure 1: Brainstorming Output with Highest Priority Causes Highlighted

3 | P a g e

From this, a prioritised list of the “top 10” was determined and the higher priority items became

the basis of the selection of the agile techniques (from the likes of eXtreme Programming (XP)

[1], Scrum [2], FDD [3], Crystal [4], DSDM [5] and even RAD/JAD [6]).

Project teams were formed for all four improvement projects, with the Agile and TFS project

teams merging within the first month due to the overlap of process and tool. There was a total of

four staff involved in the Agile Project (one lead ~50% to 80% of his time, and 3 “consulting”

staff, one being an external agile consultant).

Consistency of concepts and terminology between the Agile Project and the Project

Management/PRINCE 2 project was deemed important to ensure a clear message was

communicated to all. As such, close and regular communication between the projects (with two

individuals common to both) was ensured. As a result, the Agile Project used the PM/PRINCE

2 concept and terminology for “Stages” (Releases from XP [1]) and “Work Packages”

(Iterations from XP).

Following the Agile principle [7] of “satisfy the customer through early and continuous

delivery” (plus wanting to avoid any “big bang”) the definition, templating and training of

techniques were undertaken as (2-4 week) Work Packages (Iterations), with introduction of

these techniques primarily undertaken through pilot projects. This was the vehicle to trial the

selected techniques and mentor the handful of developers and team leads involved, who

eventually would become “seeds” of this knowledge in the next project they worked on.

Some techniques (such as code inspection) were rolled out across all active developments using

in-house workshops, where the immediate benefits were more obvious. Several in-house

training sessions were also conducted to provide the developers with overviews of the “Best of

Agile” techniques (including a customised version for Snowden’s other development offices),

plus an overview of automated build and

test processes that formed part of the

second stage of STAg

Following the agile principle [7] of

“business people and developer must

work together”, working groups of 3-5

employees were used for various

techniques to ensure that a suitable

cross-section of inputs and ideas were

brought together. There were working

groups for the Code Inspection, Coding

Standards and Estimation.

Focused feedback from the pilot

project team members was sought, via

Figure 2: Stages and Work Packages Define the Fundamental Scope and Schedule Controls

(other activities are derived from the PRINCE 2 Project Management methodology)

Figure 3: Reflection Sessions Proved Useful to Adjust the

Direction and for Positive Reinforcement

4 | P a g e

Reflection workshops, on a monthly basis, with global feedback from all developers sought via

the weekly Tec meeting with the Agile Project as an agenda item, and also through direct

communications with the Agile Project team members. These feedback mechanisms were the

start of establishing a continuous improvement philosophy.

5. Process Flow

A “road-map” of the development process showing the relationship of the selected

techniques was produced in the first Work Package for the project. This is seen as the

equivalent of the “logical architecture” model and also captures the high level

requirements of the proposed implemented methodology. This agile view, Figure 4, also

served as a planning and progress “information radiator” (as per the Crystal

methodology [4]), by highlighting the techniques that were being introduced for a given

Stage of the Agile Project. Figure 4 shows the process road-map at the time of writing this

paper. This includes a handful of techniques from the Project Management / PRINCE 2

process (e.g. High Level Requirements, Project Brief, Authorise Project/Stage, Stage Test

Plan and Stage Acceptance Testing) which serves to give further context in the

developers’ view.

The STAg methodology process road-map captures the complete suite of activities,

which were prioritised and implemented, mostly via pilot projects incrementally. The

implementation of each activity included the creation of “cheat sheet(s)” which are one

page summaries of each technique(s) employed with examples of the output, and

templates where applicable. The effectiveness of the activities introduced to date are

listed below in the order of implementation.

Figure 4: Snowden Technologies’ Agile (STAg) Process Road-map

5 | P a g e

Technique

Priority

(Stage-

Wrk Pkg)

Number of: Effectiveness

Notes

Cheat

Sheets Templates Quality Mgnt

Daily Brief 1-2 1 – Medium High Ideal communication

Reflection 1-2 1 1 Medium High Good for feedback and

morale

Code

Inspection

1-3 1 2 V. High Medium Immediate roll-out,

progress indicator

Code

Analysis

1-4 1 – High Low A high number of issues

initially

Check In 1-4 1 – Medium Low More of a “given”

Coding

Standards

1-4 2 2 Medium Low To ensure consistency

Estimation 1-5 2 1 Medium High Project and Stage

estimates using

Wideband Delphi

CUT

Complete

1-6 1 1 High High Ensures Code Insp, auto-

test, code analysis, etc

Test

Harness

2-1 1 1 V. High Medium Automated testing is

seen as a must

6. The People Side

The highest priority cause of development issues identified in the initial brainstorming

session was “communication”. Snowden’s main driver for establishing clear

communication channels derived from the agile principle that the focus should be on

“individuals and interactions over processes and tools” from the Agile Manifesto [7]. As

such, two key aspects were focused on to improve the communication:

1. All developers, project managers and domain experts were involved throughout the Agile Project. This includes working groups, feedback sessions, training

sessions, weekly Tec meetings and “Seeds” roles within project teams)

2. Agile techniques that actually facilitate (or at least prompt for) focused communications (e.g. JAD sessions, Work Package prioritisation, design sessions,

code inspections, daily briefs and Domain Object Model walkthrough) were

selected to form part of the STAg.

The implementation of this communication approach was conducted in the Stage One

Work Package named “Communication”, and its deliverables included rollout of the Daily

Brief, Reflection meetings and Role Identification (including the “Seed” role). Outlining

the project roles and their lines of communication through the Role Identification

deliverable helped to reinforce each of the related activities.

The established project roles and their relationships are shown in Figure 5, with the

grayed/italic roles proposed for later adoption.

6 | P a g e

After the initial roll-out of the techniques, a focus group was established to help

facilitate the introduction and mentoring of subsequent techniques across the three

development sites. This group was named the Technical Operations Management Group

or “TOM Group”

7. Tools

Initially, there was a separate improvement project focused on the primary

development tool set – Microsoft’s Team Foundation Server (TFS). It soon became

obvious that the process needed to be the primary focus and that the tools needed to

support, automate and reinforce elements of this. As such, the two improvement projects

were merged with the emphasis to integrate the applicable process elements into the tools.

The following activities (from Figure 4) relate to the use of the tools in the context of the

STAg methodology:

• Check In and Check Out

• Test Harness (automated) – TFS has an NUnit style

automated test framework

• Code Analysis

• Code Coverage

• CUT Complete

• Continuous Integration Build

• Email Notification of Failure

• Nightly Build

• Weekly Build

Project Manager

Technical Coordinator

Team

Coordinator

Developer

Technical Writer Configuration &

Build Coordinator

Architect

Chief Architect

Developer Quality Assurance Manager

Quality Assurance Officer

Divisional

Manager

End User

To

Divisional Manager

Shared Specialists

Internal Project TeamClient Project Team

To End User

Legend

Current Role

Reports To

Communicates With

Communicates Heavily

To Architect

To Architect

To Domain Expert

Planned Role

Quality Assurance

Domain Expert

Client Internal

Executive

Client Internal

Seed

Figure 5: Team Structure – Focused Roles/Responsibilities Helped to Reinforce the Activities

Figure 6: CUT Complete Implemented as an

Extension to TFS

TFS provided Snowden the ability to modify

incorporating and reinforcing the

and Unit Test) implemented in Stage One

tools with consistency in naming conventions, tailored training, etc to help reinforce the related

activities.

8. Lessons Learnt

In qualifying the lessons learnt for this paper, we have

Reflection on both the approach

the STAg methodology. As such, the “Keep/Problem/Try”

used to “reflect” on the lessons learnt:

8.1. Keep

8.1.1 Implementing Agile Techniques:

• Involvement of all developers, experts are a must.

• Focus on the gradual introduction of techniques 4 weeks duration) and pilot

• Establish a clear “road

• Develop cheat sheets example(s).

• Reflection sessions

• A clear (and simple) role definition to identify resourcing

8.1.2 Applied Agile Techniques:

• Code Inspection – and communication tool

quality of the code with minimal effort (we

found an additional

maintenance defects in the first month).

Figure 7: Automated Testing = TFS Test Harness + Naming

the ability to modify and extend its behavior, which proved ideal for

incorporating and reinforcing the new STAg activities, especially the “CUT complete” (Code

implemented in Stage One. Other elements simply required the use of

tools with consistency in naming conventions, tailored training, etc to help reinforce the related

In qualifying the lessons learnt for this paper, we have effectively

on both the approach taken and the selection of agile techniques

. As such, the “Keep/Problem/Try” (from Crystal

used to “reflect” on the lessons learnt:

Agile Techniques:

Involvement of all developers, team leaders, project managers

are a must.

radual introduction of techniques – utilising Work Packages (

) and pilot projects.

lear “road-map” of the techniques – ideal information radiator

heat sheets – one page (A5) summary of each technique with

Reflection sessions – focused feedback every 4 to 6 weeks with key staff

r (and simple) role definition to

resourcing short-falls.

Applied Agile Techniques:

– powerful peer review

and communication tool, that improves the

quality of the code with minimal effort (we

found an additional 387 run-time and 664

maintenance defects in the first month).

: Automated Testing = TFS Test Harness + Naming Convention + TFS Result Window

Figure 8: Summary of Some of the

Defects Found in Code Inspections

7 | P a g e

extend its behavior, which proved ideal for

STAg activities, especially the “CUT complete” (Code

ther elements simply required the use of in-built

tools with consistency in naming conventions, tailored training, etc to help reinforce the related

effectively presented a

and the selection of agile techniques employed in

[4]) prompts are

project managers and domain

ork Packages (at 2-

ideal information radiator.

one page (A5) summary of each technique with

focused feedback every 4 to 6 weeks with key staff.

Convention + TFS Result Window

: Summary of Some of the

Defects Found in Code Inspections

8 | P a g e

• Automated build and test – specs the code and provides the best form of regression testing.

• Daily briefs – a vehicle for good communication!

• Work Packages – frequent closure on subset of scope – good for team morale.

• Static Code Analysis – an integrated tool and process providing easy identification of code problems.

• Reflection sessions – positive reinforcement and adjustments where required.

• CUT complete – tool integration and definitive statement of quality and progress.

8.2. Problems (with some explanation)

• Communication and Scope Definition – The implementation of the first Work Package was delayed due to a misunderstanding of the business’ expectations of the

deliverables by the project team.

A couple of elements caused this situation:

a. There was no formal briefing or introduction to the projects’ expectations and

objectives within the team (there was no real “ownership” or “big picture” view

of the project);

b. The project team did not have a clear understanding of the scope of their

deliverables to be able to communicate effectively amongst each other and with

other developers in the business.

This was resolved initially in the first Reflection session by showing concrete

examples of what was required to be produced (templates, cheat sheets, etc) and by

providing positive feedback on the direction. Getting closure on the first work

packages from each project (which delivered the “road-map”) was also an effective

reinforcement of the intended approach. It still took an another 2-3 weeks before the

idea that the group was to actually define the scope of their projects (drawing from

industry techniques and methodologies) – but when they did then the true

“ownership” began.

• Information radiators – The use of Microsoft SharePoint and pin-up boards as information radiators were not as effective as we hoped. They are still useful, but

the regular Tec meetings are by far the most effective forum to share the

information and seek feedback. In addition, reflection sessions complimented this

well. Ultimately, “face-to-face”, albeit involving video-conferencing, is the best.

• Other offices – The understanding and up-take of techniques in Snowden’s other development offices were initially limited. Having only two developers involved in

teleconferencing sessions and video recordings of in-house training is now seen as

inadequate. One senior developer was involved in pilots, yet due to location was

isolated when the sharing of knowledge was paramount. If the “seeds” had been

identified and available for mentoring “face-to-face” the techniques would have

rolled across offices easier.

• Alignment with management practices and tools – The evolutionary development style of iterative development (Work Packages being scoped and planned every 2-3

weeks) was a difficult one for the Project Management planning view to handle. A

compromise was developed that required the Work Packages to be identified at the

start of the Stage for planning purposes (although this is not considered agile). The

(re)prioritization of the later Work Packages would occur to still give the

opportunity to ensure the focus is on the highest risk and most important set of

features for the next Work Package.

9 | P a g e

There is concern that the initial Work Package allocation will bias the subsequent

prioritization session and therefore lose some of the benefit of this approach (but at

least the highest priority elements will still be in the initial Work Packages at the

start of the Stage). There was also the alignment with the in-house project

management system (SPI) and accounting systems through the integration of

reporting requirements and mapping of Stages and Work Packages to a relatively

restrictive (and fixed) structure.

8.3. Try

• Create a defined set of objectives for any improvement program (at a Stage level)

• Establish more working groups – although a trade off, and we see it still needs to be selective, but we should (and will) have more working groups to define and refine

particular techniques. This was done for the Coding Standards, Code Inspection and

Estimation activities and worked well by getting a variety of inputs and ownership.

• Ensure the “Seeds” are more involved in the working groups as they need to “own” and represent the techniques moving forward.

• Increase the communication and visibility to all developers on the progress and direction of the techniques being implemented.

• Establish a continuous improvement environment – this is more an ethos and openness to improve the way we work, but so long as we maintain the Reflection

session (and follow through on the agreed actions) then this should be sustainable.

• An exchange program between the team members. There were a few visits by staff from other offices which facilitated a much deeper exchange of knowledge and

techniques as well as building up a stronger relationship.

9. Summary

Overall, the approach used so far in tailoring and introducing a selection of agile

techniques has worked well. The most fundamental lesson learnt has been to involve as

many of the “users” (developers, team leaders, etc) as possible throughout the entire

improvement project. It’s also been critical that it is done in small, digestible slices (i.e.

Iterations / Work Packages) and to seek and give feedback all the way through.

Having the target areas for improvement provided a clearer view for the adoption of

particular agile and value add techniques. We selected the techniques from a range of

industry agile methodologies, some common, some fairly unique, but all of them to

improve our target areas. What has been selected and tailored to date is working and

working well. Testimony to this is the very successful delivery of the single and multi-

team projects that piloted the initial techniques:

“… it was a very successful trip… big smile on his face and feeling very proud… [the

projects] were within 5% of the budget – both delivered on time!”

Rayleen Riske, Project Manager and Geologist

The STAg diagram is a key information radiator providing an ongoing reference point

for the context of our development activities on a given project as well as identification of

further improvements we wish to incorporate.

The realization that the integration and alignment of our agile software development

methodology with the project management methodology and development tools was

important as the three perspectives need to complement each other.

10 | P a g e

Not only have we established a “continuous improvement” mindset, but we have

reinforced this through explicitly embedding feedback mechanisms throughout the STAg

methodology itself.

10. Acknowledgements

The author’s would like to acknowledge the contributions of the following colleagues

that have been and continue to be involved in the Agile Project (in alphabetical order):

Jeffrey Alexander

Paul Fox

Peter Gibbes

Chris Gilbert

Mark Holst

Rayleen Riske

and all the developers, team leaders and project managers that have been involved.

11. References [1] K. Beck, eXtreme Programming Explained – Embrace Change, 2nd Ed., Addison Wesley, 2004

[2] Advanced Development Methods Inc., “Controlled Chaos: Living on the Edge”, 1996

[3] S. Palmer, and J. Felsing, A Practical Guide to Feature-Driven Development, Prentice Hall, 2002

[4] A. Cockburn, Crystal Clear: A Human-Powered Methodology for Small Teams, Addison Wesley, 2004

[5] DSDM Consortium and J. Stapleton, DSDM: Business Focused Development, 2nd Ed., Addison Wesley, 2003

[6] J. Wood and D. Silver, Joint Application Development, Wiley, 1995

[7] K. Beck, et al., “Manifesto for Agile Software Development”, www.agilemanifesto.org , 2001