Teams, Processes and Performance in Software Development

1
Teams, Processes and Performance in Software
Development

• Steve Sawyer
• Associate Professor,
• School of Information Sciences Technology,
• The Pennsylvania State University,
• University Park, Pennsylvania USA 16802
• 814-65-4450 (o)
• 814-865-5604 (f)
• sawyer_at_ist.psu.edu
• Affiliate appointments with
• Department of Labor Studies Industrial
  Relations,
• College of the Liberal Arts
• Management and Organizations Department,

American Univ., 10 Dec, 2003
2
Credits

• Support for this research provided by IBM,
  Boeing, and Syracuse University.
• Some of the data used here collected as part of a
  research project with Erran Carmel!
• Analysis done by Dr. Sophia Cho, now with World
  Bank, Phillipines.

3
Today

• Research questions (1)
• Contributions (1)
• Studying software development (2)
• Research approach (2)
• Analysis (3)
• Findings (5)
• Discussion (1)
• And, so?(1)

4
Research questions

• Do different forms of software development have
  different processes?
• Is conflict a predictor of performance in
  software development teams?
• Does using ICT affect the performance of software
  development teams?

5
Contributions

• Contrasting software development teams and the
  role of conflict in
• Spec. built
• Large package and
• Small package
• Conceptualizing functional view of ICT as
• Production
• Coordination
• Control

6
Why study

• Software development
• It is the center of information systems and
  why we use computers.
• Teams
• Are locus of making software (and conflict
  is an issue)
• Processes
• Are how software is made
• Performance
• To effect change .

7
How to study?

• Technical aspects such as tools, languages,
  environments, methods
• Individual differences and contributions
• Social aspects (teams and team-level issues)
• Context (domain, org, .)

8
Research approach

• Combine three field studies
• Common design and goals
• Survey(s) with common elements, observations,
  interviews secondary data.
• Total of 116 teams
• 62 doing spec. development
• 28 doing large packages
• 26 doing small packages
• - no ICT data on these

9
Research model
Team structure
Team process
Team Perform.
Team conflict
This is a(n) path model factor model
atemporal model
Uses of ICT for Production Coordination Control
10
Analysis

• Exploratory so MRC
• Power (for 6-7 items) ok
• Issues w/combining data
• Different qs for ICT uses
• Focus on comparisons
• Still, exploratory factor analyses load cleanly
  and explain 60.7 of variance
• But, Chronbachs alpha of items range from .53 to
  .91 .

11
Correlations for general model
Variables Mean SD 1 2 3
1. Existing Conflict 3.69 1.86
2. Team Structure 4.94 .84 -.33
3. Team Process 4.81 .94 -.29 .74
4. Team Performance 5.28 .72 -.08 .30 .42
N 116. p lt .05 p lt .01.
12
Correlations for ICT use model
Variables Mean SD 1 2 3 4 5 6
1. Existing Conflict 3.25 1.79
2. Team Structure 4.86 .82 -.68
3. Production ICT 3.01 2.11 .58 -.41
4. Coordination ICT 3.11 1.89 .63 -.44 .85
5. Control ICT 2.91 2.03 .59 -.41 .86 .86
6. Team Process 4.72 .91 -.68 .67 -.45 -.47 -.39
7. Team Performance 5.25 .71 -.21 .29 -.05 -.05 -.08 .44
N 90. p lt .05 p lt .01.
13
Path Models (no ICT)
N 116
N 62
General (adj. R2 16)
Spec. (adj. R2 18)
Small pack (adj. R2 13)
Large pack (adj. R2 18)
N 26
N 28
14
RQ1

• Do different forms of software development have
  different processes?
• Yes -- models differ
• Team process and team structure important (save
  for large package teams)
• Do not predict performance for small package
  teams
• 80 of variance not accounted for here
• Unexpected direct effects

15
RQ2

• Is conflict a predictor of performance in
  software development teams?
• Yes save in general model (!?)
• Either direct or indirect effect
• Accounts for half of all explained variance.
• Negatively correlated with team structure, team
  process and team performance (more is bad).

16
RQ3

• Does using ICT affect the performance of software
  development teams?
• Depends on which teams
• Yes for spec. teams
• Control and Coordination have direct and indirect
  effects
• They account for 9 of explained variance.
• No for large package teams.
• ALL teams use ICT!

17
Uses of ICT in Spec. Large Package teams.
N 62
Spec. (adj. R2 26)
Large pack (adj. R2 18)
No ICT?
N 28
18
Discussion

• Team processes vary by the type of software.
• But, teams level of conflict affects
  performance.
• Value of ICT use complex
• Ubiquitous but
• Varies by type of software
• Varies by ICTs function
• With production less important than control and
  coordination.

19
And, so?

• Social processes, individual actions and uses of
  ICT are related. Suggests one-size does not fit
  all in making software?
• Lots of unexplained variance. Maybe it is all
  about a few good people?
• Small package development is not understood
  (other than it is important, entrepreneurial and
  idiosyncratic). Is this OK?