Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo...

20
Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a) Nitin R. Joglekar (b) Stephen R. Rosenthal (b) (a) Universidade Nova de Lisboa (b) Boston University INFORMS 2006 Conference Pittsburg, PA November 6, 2006

Transcript of Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo...

Page 1: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

Learning while Sourcing: Evidence of Productivity Gains in Coordinating

Software Development

Paulo J. Gomes (a) Nitin R. Joglekar (b) Stephen R. Rosenthal (b)

(a) Universidade Nova de Lisboa(b) Boston University

INFORMS 2006 Conference Pittsburg, PA

November 6, 2006

Page 2: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 2

Motivation

• Emerging relevance of inter-organizational networks in software development, either through collaborative or outsourcing arrangements

• Coordination is one of the central problems of organizational research (March and Simon, 1958; Thompson, 1967) and in extended settings coordination is expected to drive a larger fraction of overall costs (Malone et al. 1987).

• Interest in systematic studies of learning by doing during the creation, retention, and transfer of knowledge (Argote, McEvily and Reagans, 2003).

• Can we actually observe improvements in coordination of development partners through experience?

Page 3: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 3

Research Objective

• How does one measure organizational learning in terms of coordinating outsourced software development?

• Which task characteristics facilitate or deter ‘learning by doing’ during the sourcing process?

Page 4: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 4

Literature on Strategic SourcingCorbett, Blackburn and van Wassenhove 1999, Quélin and Duhamel

2003, Ratten 2004

• contract crafting and management• negotiation• decision coordination• measuring relationship performance • information exchange • learn and innovate

Page 5: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 5

Related Learning Literature

It is often assumed that cultivating reciprocity and trust are competences that firms “learn by doing”

Organizations build partnering skills and negotiating skills that can be tapped in crafting better collaborative arrangements

There is some empirical evidence supporting this argument: • Experience in network formation drives success in the

launch of new network arrangements (Gulati and Gargiulo 1999).

• Learning in production tasks (Sobrero and Roberts, 2001; Uzzi and Lancaster, 2003, Macher and Mowery 2003)

• Learning to contract with external partners (Mayer and Argyres, 2004).

Page 6: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 6

Measuring Organizational Learning

• Coordination productivity – units of technical problem solving per unit of coordination required

Learning Model

Ln [Coordination Productivity it] = + 1 * ln (Time t) + 2* ln (Task Scale i) + 3* ln (Task

Duration i) + 4 * (Outsourcing Ratio i)+ it

Page 7: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 7

Learning Hypothesis

H1: Coordination productivity is positively associated with accumulated experience, holding all else constant.

Page 8: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 8

Task-level Determinants

H2A: Coordination productivity is positively associated with task scale.

H2B: Coordination productivity is negatively associated with task duration.

H2C: Coordination productivity is negatively associated with fraction of task outsourced.

Page 9: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 9

Moderator Hypothesis

H3A: Task scale moderates the association of experience and coordination productivity; larger tasks are associated with less learning.

H3B: Task duration moderates the association of experience and coordination productivity; tasks with greater duration are associated with less learning.

Page 10: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 10

Empirical Model

Fraction Outsource

Project Complexity

Scale

Duration

PM Experience

Cumulative Task

Experience (Time)

Inverse Coordination

ProductivityH1

H2B

H2C

H3

Task Characteristics

FixedFactors

ScaleH2A

Page 11: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 11

Internal versus External Coordination

• Internal coordination is defined as management of dependencies with related project teams and functional areas within the organization (communication with other software projects, hardware development, test stands, marketing, and customer care)

– Internal Coordination Productivity • External coordination is defined as management of dependencies with

the technology suppliers across the boundary of the focal firm (time spent managing partners under contract and establishing rules for addressing software bugs or reviewing engineering releases).

– External Coordination Productivity

Page 12: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 12

Data and Methods

– Software for Data Management

– Largely Outsourced Work

– 7 projects, 37 tasks over 2 years

– structured interviews with key managers

– review of documents reporting the product development effort

– field observations

Page 13: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 13

Description of Project Sequence

Name Time (days)

Description

P1 1 Software developed for windows (PC), which provides storage and analysis of user data uploaded from hardware devices.

P2 104 Upgrade of software application that enables uploading of data from hardware device. Integrated with P1 as unique system.

P3 118 Software package developed for windows (PC) for managing data collected from different hardware instruments measuring blood gas. Intended to eliminate bugs and enhance functionality of software product.

P4 370 Development of web based software system for storage, analysis and communication of data made available by the hardware devices. Represents development of similar functionality as the product in P1 on different platform.

P5 422 Active participation in industry consortium effort to develop of a commercially viable standard for data transport and integration to and from a category of medical hardware devices.

P6 544 Development of internal standards in terms of communication protocol.

P7 548 Development of software application to upload hardware device data to the platform used in P4. Some product technology is transferred from P2.

Page 14: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 14

Descriptives

Mean Std. Deviation

Inv. Coordination Productivity .464 .648

Inv. Internal Coordination Productivity

.223 .600

Inv. External Coordination Productivity

.241 .305

Task Duration (days) 32 41

Technical Work (engineering hours)

725.8 1252.2

Outsource Ratio 0.39 0.44

Overall Project Complexity (engineering hours)

6001 6447

Page 15: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 15

Results - Learning Model

Ln (Inv. Coordination Productivity)

I

Ln (Time) -.453*

Ln (Task Duration) .407*

Outsource Ratio -.185

Ln (Technical Work) -.866**

Ln (Overall Project Complexity) -.222

Years of Project Mgt Exp. -.004

Adjusted R2

FNDurbin-Watson

.5146.47**

371.69

** Significant at .01 level, * significant at .05 level, † significant at .10 level

Ln (Inv Iternal Coordination Productivity)

Ln (Inv. External Coordination Productivity)

II III

-.586† -.529*

.165 .317

-.330 -.153

-1.063* -.802**

-.134 -.238

.057 .052

.7186.93**

152.93

.4003.88**

271.62

Page 16: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 16

Results - Learning Model with Interaction

Coordination Internal Coordination

External Coordination

IV V VI

Ln (Time) -.598** -.946* -.626*

Ln (Task Duration) .358† -.163 .335

Ln (Time) x Ln (Task Duration) .368† .527† .291

Outsource Ratio -.281† -.306 -.209

Ln (Technical Work) -.859** -1.158** -.797**

Ln (Overall Project Complexity) -.033 -.057 -.066

Years of Project Mgt Exp. -.053 .032 .028

Adjusted R2

FNDurbin-Watson

.5676.81**

371.809

.8008.97**

152.32

.4123.60*

271.75

** Significant at .01 level, * significant at .05 level, † significant at .10 level

Page 17: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 17

Key Findings

The economics of coordinating outsourced software development is marked by scale effects, learning by doing, and differential learning rates:

– we find significant evidence of learning in terms of coordination productivity, in particular external coordination productivity

– the extent of learning does depend on the characteristics of the development task, namely its duration – there is some support for less learning in tasks with greater duration – schedule compression effect

Page 18: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 18

Managerial Implications

• By inducing administrative competence learning, organizations may improve the governance of outsourcing arrangements. However, the sizing of tasks (e.g. the manner in which they are modularized) can moderate this effect.

• Future Work: Mechanisms behind the learning effects observed in terms of coordination (Adler and Clark 1992). The development of routines that decrease perceived transactional risk, hence coordination needs could be such a learning mechanism.

Page 20: Learning while Sourcing: Evidence of Productivity Gains in Coordinating Software Development Paulo J. Gomes (a)  Nitin R. Joglekar (b)  Stephen R. Rosenthal.

INFORMS 2006 Learning while Sourcing 20

Correlation Matrix for Independent Variables

** significant at the 0.01 level * significant at the 0.05 level

1. 2. 3. 4. 5. 6. 1. Time 1.0 2. Task Duration -.332* 1.0 3. Outsource Ratio .005 .297 1.0 4. Technical Work -.450** .617** .409** 1.0 5. Project Complexity -.357* -.188 .074 0.172 1.0 6. Years of Project Mgt Exp. -.018 .218 -.019 .404* .420** 1.0