team affiliation and spatial networks
TRANSCRIPT
Networks, Teams and Space June 2015
Team Affiliation and Spatial Networks
A Comparative Analysis of Organisation, Space and
Network Structure
Dr Kerstin Sailer
Space Syntax Laboratory, Bartlett School of Architecture, University College London, UK
XXXV Sunbelt Conference of the International Network for Social Network Analysis, 23-28 June 2015
@kerstinsailer
Networks, Teams and Space June 2015
Introduction
“How to Collaborate”
British Airways Business
Life Magazine
March 2015
Networks, Teams and Space June 2015
Introduction
Advertising Agency, Frankfurt
Very frequent face-to-face encounter
(several times a week)
Colour of nodes: Teams
Shape of nodes: Floor
To which degree do
organisational and
spatial barriers hinder
interactions?
Networks, Teams and Space June 2015
Introduction
ORGANISATION
Intra-organisational
networks of face-to-face
interaction in the workplace
Proximity
Shared
paths
Shared
workspace
Job roles
Reporting lines
Organisational
cultures
IMPACT IMPACT
Networks, Teams and Space June 2015
Introduction
ORGANISATION
Attribute:
Team affiliation
E-I index:
Comparing numbers of ties within
groups and between groups (Krackhardt and Stern 1988)
Attribute: floor where
desk is located
Networks, Teams and Space June 2015
Introduction
WEEKLY INTERACTION DAILY INTERACTION
organisation team internal floor internal team internal floor internal
University School pre 42% 63% 65% 91%
University School post 47% 61% 54% 86%
Research Institute 48% 59% 64% 71%
Publisher C pre 32% 60% 37% 77%
Some results for a small sample of organisations (based on earlier work presented at 5th
UKSNA conference in 2009 and published in Sailer 2010):
(Based on E-I index calculations of face-to-face interaction networks)
→ But how do we control for intervening variables such as structure of an organisation?
Networks, Teams and Space June 2015
Research Problem
Organisation Structure A
100 staff, N=10 teams of S=10
5050
1010
1010
1010
1010
1010
Organisation Structure B
100 staff, N=2 teams of S=50
Maximum number of internal and external ties vary depending on number and size of
subgroups (Krackhardt and Stern 1988)
E∗ = S2𝑁 (𝑁−1)
2and I∗ =
𝑁𝑆 (𝑆−1)
2
→ E*= 4500; I*= 450 → E*= 2500; I*= 2450
→ How can we compare across organisations and understand the degree of team
cohesion and structural embedding in the light of diverse organisational structures?
Networks, Teams and Space June 2015
Case Study Overview
21 knowledge-intensive organisations across different sectors (creative industry, information
business, retail, legal, technology, media, NGO) in the UK, all studied separately between
2007 and 2015 as part of workplace consultancy undertaken by Spacelab
Ranges:
Organisation size: 67 ↔ 1377 staff
Numbers of teams: 5 ↔ 83 teams
Average team size: 8.5 ↔ 32.5 staff
Office building: 1 ↔ 18 floors
Average size of floor plate: 200 ↔ 2800 sqm
Organisation Size
Average Team Size
0
200
400
600
800
1000
1200
1400
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
Networks, Teams and Space June 2015
Methodology
SNA:
Online survey of each organisation; survey
distributed to all staff members; return quote: 49%
(lowest) to 90% (highest);
Asked each participant to name top 25 contacts
and indicate frequency of face-to-face encounter;
Analysis of network of strong ties (daily
encounter);
Network attributes: team affiliation, floor where
desk is
Calculating E-I index, Expected E-I index, Yule’s Q
Spatial Analysis:
Anaysis of spatial configuration using VGA on eye
level (visibility) [average mean depth];
Networks, Teams and Space June 2015
Results
Calculation and analysis of various metrics for each organisation (using UCINET):
• Percentage of internal links as calculated by E-I index routine (%INT)
• Internal – external preference, i.e.
%𝐸𝑋𝑇
%𝐼𝑁𝑇%𝐸𝑋𝑇𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑
%𝐼𝑁𝑇𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑
as per E-I index routine (INT-EXT pref)
• Degree of internalisation, i.e. 𝐼𝐿
𝑁
𝑆𝑎𝑣, where IL is the total number of internal links, N is
the total number of nodes in the network and 𝑆𝑎𝑣 is the average size of teams, as per E-
I index routine (INTERNALISATION)
• Yule’s Q as calculated by the Homophily routine and derived from the odds ratio, which
maps perfect homophily (+1) and perfect heterophily (-1) by 𝐼𝐿×𝑁𝐸𝐿−𝐸𝐿×𝑁𝐼𝐿
𝐼𝐿×𝑁𝐸𝐿+𝐸𝐿×𝑁𝐼𝐿, where IL is
the number of internal links, EL the number of external links, NIL the number of non-links
internally and NEL the number of non-links externally (Yule’s Q)
Networks, Teams and Space June 2015
Results
% INT
[team]
INT-EXT pref
[team]
Internalisation
[team]
Yules Q
[team]
% INT
[floor]
Yules Q
[floor]
# Staff 0.007 0.242* 0.114 0.168 0.088 0.059
# Teams 0.012 0.344** 0.003 0.278* 0.056 0.067
Av team size 0.012 0.114 0.703** 0.164 0.033 0.001
# Ties 0.002 0.292* 0.040 0.220* 0.080 0.067
Density 0.050 0.074 0.159 0.040 0.219* 0.188
Testing correlation between metrics and standard descriptors of organisation structure
(number of staff, number of teams, average team size) and network structure (number of
ties, density)
DAILY INTERACTION FREQUENCY
R2 values; significance at p<0.05 marked with * and p<0.01 with **
Note: correlation marked in purple is driven by one outlier;
Networks, Teams and Space June 2015
Results
Two metrics seem (relatively) robust: %INT and Yule’s Q
→ calculating values for both attributes (team, floor) for daily interaction
→ plotting range of cases
Networks, Teams and Space June 2015
Results – Analysing
single casesCASE 7
Post prod
house
BENCHMARK
all org.
CASE 9
large retail
organisation
Percentage of internal ties
[%INT]: depicts patterns of
interaction and degree to which
they span team boundaries and
reach across floors
Yule’s Q [team]: depicts degree
of organisational structure as a
barrier
Yule’s Q [floor]: depicts degree of
spatial structure as a barrier
Networks, Teams and Space June 2015
Results – Exploring the Impact of Spatial Structure
Correlation between Yule’s Q [team] and Maximum Mean Depth
(R2=0.468**, p<0.003) (if outlier case 3 is excluded)
Cas
e 14
–S
trat
egic
vis
ibili
ty in
offi
ce (
clos
enes
s ce
ntra
lity)
→ Offices with higher levels of maximum visibility tend
to host more heterophilous interactions, i.e. allow more
interactions between colleagues of different teamsIntegrated Segregated
Networks, Teams and Space June 2015
Outlook – Where to go from here?
• Growing the data and looking at different metricso Calculate %INT and Yule’s Q for weaker ties of weekly encounter?
o Calculate %INT and Yule’s Q for usefulness ties?
• Control for work flows and organisational purpose (the need to collaborate across team
boundaries)… but how?
o Calculate average In-Degree of usefulness as control variable?
o Group by industry?
o Break down analysis to team level?
o Compare teams with similar tasks, e.g. Sales or Marketing?
• Bring space back in more systematically
o Are smaller or larger floor plates better? No obvious correlation…
o Does the provision / distribution of attractors make a difference?
o Does the average integration of a team workspace make a difference?
Networks, Teams and Space June 2015
Dr Kerstin Sailer
Lecturer in Complex Buildings
Bartlett School of Architecture
University College London
140 Hampstead Road
London NW1 2BX
United Kingdom
Thank you!
@kerstinsailer
http://spaceandorganisation.wordpress.com/
http://tinyurl.com/kerstinsailer