Distributing a Lean Organization: Maintaining Communication While Staying Agile.

Full text

Page 1

Distributing a lean organization:
Maintaining communication while staying agile
(LESS 2010, Helsinki)
Sebastian Meyer, Eric Knauss, Kurt Schneider
{sebastian.meyer,eric.knauss,kurt.schneider}@inf.uni-hannover.de
Software Engineering Group
Leibniz Universität Hannover

Page 2

Outline
• Motivation
• Project Setup
• Our GQM Approach
• Problems
• Conclusions
Sebastian Meyer: Distributing a lean organization (LESS 2010)219.10.2010

Page 3

Motivation
• There are just little few developers who are skilled in every
aspect/technology of software development.
– It is a common practice to build teams consisting of specialists
– Specialists needs to be integrated from outside or inside the company
• Not in all cases a co-located team can be formed
– Integrating these teams can turn into an extreme overhead
• Communication is one of the major keystones in agile
development
• We wanted to understand communication in a (distributed)
agile project to understand which additions are absolutely
necessary for distributing teams
– And how they change the agile character
3Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010

Page 4

Project Setup (I)
• We used our XP-Lab as a base for the observations
– Two 4h Spikes (Tuesdays)
– One get-together to kickoff the project
– Working one 40h week (Tuesday – Friday + Monday)
• For the distribution, we made use of the GloSE-Lab
– A distributed lab between (in this case) two universities
– Part of the Global Software Engineering project in Germany
• We built two teams:
– A dispersed team with 4 students at each site
– A co-located team with 7 students
– All students where in the Master program
4Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010

Page 5

Project Setup (II)
• To connect to the remote site, we used:
– Skype
– A dedicated video conference system
– A distributed MindMap
– A desktop/document sharing tool
• Coaches and on-site customer where located in Hannover
• Both teams worked on Android Projects in Java
– Using Eclipse with integrated Android tools
– SVN for code sharing
5 Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010

Page 6

The GQM Tree
• We used the GQM method with the following goal:
Understand the communication in distributed XP from the
viewpoint of an XP coach and the developers
6Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Understand communication
Q1: How much Coordination
Overhead is needed in a
dispersed XP-Project?
M1.1: Measure how much time it
takes to setup and initiate each
communication channel
M1.2: Measure which activity
uses which channel
M1.3: Count how often each
activity is done
Q2: Does a dispersed project
need more communication than
a co-located one?
M2.1: Count the number of
stand-up meetings during the
project
M2.2: For each stand-up
meeting measure its length.
Q3: In relation to forced rituals:
Do developers contribute in
communication situations and
do they initiate communication
by themselves?
M3.1: Count Skype and video
calls initiated by the developers
Q4: Do dispersed teams
document more information
items per story card than co-
located teams?
M4.1: Count the number of
documented information items
per story card in each team.

Page 7

Question 1
7Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Q1: How much
Coordination
Overhead is needed in
a dispersed XP-
Project?
M1.1: Measure how
much time it takes to
setup and initiate each
communication
channel
M1.2: Measure which
activity uses which
channel
M1.3: Count how often
each activity is done

Page 8

Hypothesis for Question 1
• Hypothesis 1: We assume that communication in a
dispersed XP project needs considerably more technical and
organizational effort than in the co-located case and has
influence on the question, whether the project stays in time
and budget.
Sebastian Meyer: Distributing a lean organization (LESS 2010)819.10.2010

Page 9

Measurements for Question 1
9Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010

Page 10

Question 2
10Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Q2: Does a dispersed
project need more
communication than
a co-located one?
M2.1: Count the
number of stand-up
meetings during the
project
M2.2: For each stand-
up meeting measure
its length.

Page 11

Measurements for Questions 2
• Hypothesis 2: We assume that one stand-up meeting per
day is enough, as in the co-located case but that it will take
more time (approx. 50 % more) compared to a co-located
setting.
11 Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Date
Length
25.06.10
19:48 min
26.06.10
31:29 min
27.06.10
26:00 min
28.06.10
27:32 min
31.06.10
37:52 min
• Average length of dispersed meeting:
• Average length of co-located meeting:
28:32 min
~15 min

Page 12

Question 3
12Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Q3: In relation to forced
rituals: Do developers
contribute in communication
situations and do they
initiate communication by
themselves?
M3.1: Count Skype and video
calls initiated by the
developers

Page 13

Measurements for Question 3
• Hypothesis 3: We assume that the developers contribute
less than in a co-located project. Furthermore, we expect
very few self-initiated communication situations. Rituals like
planning poker or 5-point-evaluation improve the situation
• Skype calls initiated by developers:
• Video conferences initiated by developers:
16
0
• Co-located team and parts of the distributed team that where
on the same location communicated a lot
• Few communication between the distributed locations
13Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010

Page 14

Question 4
14Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Q4: Do dispersed
teams document more
information items per
story card than co-
located teams?
M4.1: Count the
number of documented
information items per
story card in each
team.

Page 15

Measurement for Question 4
• Hypothesis 4: We assume that they document about 50 %
more than in a co-located setup. Video channels decrease
amount of documented information items and allow to value
interactions over documents.
15Sebastian Meyer: Distributing a lean organization (LESS 2010)19.10.2010
Co-Located Team
135
54
81
Dispersed Team
127
28
99
Requirements total
Documented Req.
Not documented Req.
Contradiction!