Managing uncertain tasks in technology‐intensive project environments: A multi‐method study of task closure and capacity management decisions

• Published date: 01 April 2020
• Author: Sridhar Balasubramanian,Ying Zhang,Sriram Narayanan,Jayashankar M. Swaminathan
• DOI: http://doi.org/10.1002/joom.1062
• Date: 01 April 2020

RESEARCH ARTICLE

Managing uncertain tasks in technology-intensive project

environments: A multi-method study of task closure and capacity

management decisions

Sriram Narayanan

1

| Sridhar Balasubramanian

2

| Jayashankar M. Swaminathan

3

|

Ying Zhang

4

1

Department of Supply Chain Management,

Eli Broad School of Business, Michigan

State University, East Lansing, Michigan

2

Marketing Area, Kenan-Flagler Business

School, University of North Carolina at

Chapel Hill, Chapel Hill, North Carolina

3

Operations Management Area, Kenan-

Flagler Business School, University of

North Carolina at Chapel Hill, Chapel Hill,

North Carolina

4

Department of Management, College of

Business, Clemson University, Clemson,

South Carolina

Correspondence

Sriram Narayanan, Department of Supply

Chain Management, Eli Broad School of

Business, Michigan State University, East

Lansing, MI 48824.

Email: narayanan@broad.msu.edu

Handling Editor: Gregory Heim

Abstract

Engineers working on tasks in technology-intensive project environments face sub-

stantial task resolution uncertainty. This can result in poor capacity utilization as

they expend effort on tasks that are not successfully resolved. A deeper understand-

ing of task-related uncertainty can help the firm optimize effort allocation across

tasks by implementing well-designed task closure policies that facilitate superior

capacity utilization and capacity planning. In this article, we characterize the empir-

ical distribution of task uncertainty and demonstrate how the resolution process

affects task outcomes in project environments. Using a combination of empirical

estimation, and analytical and simulation modeling, we develop insights into task-

related decision-making and engineer effort allocation. Using real-world task reso-

lution data from a software maintenance setting, we first model and estimate a

beta-geometric survival distribution which indicates that the likelihood of success-

ful task resolution substantially reduces with the time a task remains in the system.

Using analytical and simulation modeling, we then examine the implications of

imposing task closure policies to improve engineer effort allocation and increase

system productivity. We demonstrate that adopting well-designed task closure poli-

cies can significantly improve engineer resource utilization in capacity-constrained

settings without a substantial negative impact on project outcomes. We discuss the

implications of our research for theory and practice.

KEYWORDS

customer service, knowledge worker productivity, project management, resource allocation, system

performance, task closure, technology management

1|INTRODUCTION

Many project activity domains such as software develop-

ment and maintenance, implementing engineering change

orders (ECOs), and technology search during R&D pro-

cesses are characterized by significant uncertainty related to

task resolution efforts and outcomes. Yet, these activities are

critical to providing superior results of interest to stake-

holders and the marketplace. For example, large software

products require ongoing support through dedicated teams

focused on bug resolution. Similarly, teams often work on a

constant flow of ECOs for complex physical products

(Loch & Terwiesch, 1999). Organizations dedicate a sub-

stantial degree of resources to these tasks. Forrester Research

Received: 26 March 2016 Revised: 2 September 2019 Accepted: 14 September 2019

DOI: 10.1002/joom.1062

260 © 2019 Association for Supply Chain Management, Inc. J Oper Manag. 2020;66:260–280.wileyonlinelibrary.com/journal/joom

notes that more than half of software budgets in North

American and European firms are dedicated to software

maintenance (Kisker, Ried, & Shey, 2010). Similarly, ECOs

may comprise 20–50% of product costs (Terwiesch &

Loch, 1999).

Proactive project planning in technology-intensive envi-

ronments in general, and software maintenance operations in

particular, can be challenging. For example, budgeting for

software maintenance jobs is difficult because, even while

organizations allocate significant budgets to this area, sub-

stantial uncertainty exists with respect to problem resolution

likelihood and timeliness (April, Abran, & Dumke, 2004).

Specifically, resolution likelihood is the probability that the

software maintenance task can ultimately be resolved

irrespective of how long an engineer works on the task

(McConnell, 1993). Many maintenance requests in the soft-

ware domain are ultimately closed without resolution

(Aranda & Venolia, 2009; Guo, Zimmermann, Nagappan, &

Murphy, 2010; Zhang, Gong, & Versteeg, 2013). Time to

resolution refers to the length of time an engineer expends

effort on the task before it is resolved. Thus, from a capacity

management standpoint, both resolution likelihood and time

to resolution are important. For example, Jarratt et al. (2011,

p. 111) note that the calendar processing time for ECOs on

large products can vary substantially from “2 days to the

whole life of the product (i.e., the change was never

implemented).”Some insights related to task resolution like-

lihood and timelines can be extrapolated to a new project

from other, similar projects. To the extent that such extrapo-

lation is appropriate, those insights can be incorporated into

the proactive project planning process. At the same time, in

parallel with proactive project planning, the nature of the

projects related to software maintenance also calls for sub-

stantial learning on the go—the “retrospective”development

of project management policies related to capacity and effort

allocation based on insights gained from managing the early,

ongoing flow of tasks. We focus on such retrospective learn-

ing in this article.

An additional, important point is that the nature of uncer-

tainty encountered in software debugging and support envi-

ronments differs from that encountered during traditional

software product development (programming) processes.

Weiss et al. (2007, p. 1) note: “…in contrast to program-

ming, which is a construction process, debugging is a search

process.”Debugging typically requires iterative problem

solving and knowledge discovery cycles (Hale & Haworth,

1991; Jarratt et al., 2011). Given these challenges, managers

in software maintenance environments frequently struggle

with unexpected delays. These delays are often self-

propagating and engender firefighting behavior (Bohn,

2000; Repenning, 2001), ultimately causing systemic bottle-

necks, performance degradation, and poor productivity. As

discussed earlier, such inherent uncertainty is difficult to

manage solely through proactive planning.

Arguing that the effective management of uncertainty in

project settings remains sparsely explored, Böhle et al.

(2016, p. 1386) note that “…the question remains open as to

how project actors handle uncertainty in practice.”Uncer-

tainty in these tasks arises because outcomes are either

unknown or known with limited precision; further, uncer-

tainties become “more known”as additional information is

collected on the project/task over time (Ramasesh &

Browning, 2014). Stated differently, there is “discovery

through analysis.”However, managers often lack an

approach to systematically incorporate emerging task-related

information on a dynamic basis to adjust their task-related

efforts at a micro-level. Ultimately, this capability gap nega-

tively impacts project resource management at a macro-level

as well. The core contribution of our study is to help bridge

this gap by presenting a robust framework for the manage-

ment of task uncertainty within projects that integrates

micro-level task characteristics and macro-level planning

efforts. Our research spans the theoretical development and

empirical verification of the framework, and simulation-

based policy generation related to capacity, effort allocation,

and task closure decisions using the framework. We demon-

strate that by adopting appropriate task closure policies, pro-

ject resource efficiency related to the effort expended on

task resolution can be substantially enhanced while

maintaining high levels of project outcomes that are impor-

tant to customers and other stakeholders.

We draw on a dataset of software maintenance

requests—or bugs—assigned to engineers in an India-based

software services company. For brevity, going forward we

refer to software maintenance requests as “tasks”and indi-

viduals working on these tasks as “engineers.”Engineers

worked on tasks that were allocated to them by managers.

We present an analytically anchored empirical methodology

to (a) develop and empirically validate an approach using

archival data to model the ongoing resolution of uncertainty

as tasks progress and (b) demonstrate how this model can be

leveraged to make effective engineer-level task-effort alloca-

tion and task closure decisions, and system-level capacity

decisions to improve efficiency while delivering on overall

project outcomes.

Specifically, we examine how firms can design policies

to maximize the marginal value of engineer effort when suc-

cessful task resolution is uncertain. We proceed in three

steps. First, we model the process of task resolution using a

beta-geometric survival distribution (or beta-geometric dis-

tribution) and estimate the model parameters for a range of

task classifications. In this step, we show that the beta-

geometric distribution closely estimates the length of time a

task stays in the system. In particular, the successful

NARAYANAN ET AL.261