Thirsty in an Ocean of Data? Pitfalls and Practical Strategies When Partnering With Industry on Big Data Supply Chain Research

• Author: Rod Franklin,A. Michael Knemeyer,Kevin B. Smyth,Keely L. Croxton
• DOI: http://doi.org/10.1111/jbl.12187
• Published date: 01 September 2018
• Date: 01 September 2018

Thirsty in an Ocean of Data? Pitfalls and Practical Strategies

When Partnering With Industry on Big Data Supply Chain

Research

Kevin B. Smyth

1

, Keely L. Croxton

1

, Rod Franklin

2

, and A. Michael Knemeyer

1

1

The Ohio State University

2

K€

uhne Logistics University

Increased volume, velocity, and variety of data provides new opportunities for businesses to take advantage of data science techniques, pre-

dictive analytics, and big data. However, firms are struggling to make use of their disjointed and unintegrated data streams. Despite this, aca-

demics with the analytic tools and training to pursue such research often face difficulty gaining access to corporate data. We explore the

divergent goals of practitioners and academics and how the gap that exists between the communities can be overcome to derive mutual value

from big data. We describe a practical roadmap for collaboration between academics and practitioners pursuing big data research. Then we

detail a case example of how, by following this roadmap, researchers can provide insight to a firm on a specific supply chain problem while

developing a replicable template for effective analysis of big data. In our case study, we demonstrate the value of effectively pairing manage-

ment theory with big data exploration, describe unique challenges involved in big data research, and develop a novel and replicable hierarchical

regression-based process for analyzing big data.

Keywords: big data; data science; practitioner engagement; governance form

INTRODUCTION

Many of today’s supply chain managers feel like the ancient

mariner in the famous poem by Samuel Taylor Coleridge. But

rather than being surrounded by undrinkable water, supply chain

managers are more likely to be heard saying “Data, data, every-

where, nor any information to use.”An estimated 1,000 exabytes

(10

21

bytes) of data are generated each year (Yin and Kaynak

2015), and U.S. firms store on average more than 200 terabytes,

with a 40% expected annual compounded growth in that figure

(McKinsey Global Institute 2011). Often referred to as “big data”

(McAfee and Brynjolfsson 2012), this rapid expansion of data

creation and storage represents a potential source of competitive

advantage (Waller and Fawcett 2013b) and has businesses scram-

bling to extract value from it (Sanders 2016). Although many

definitions of big data are in common use, we specify big data

as an increase in volume, variety, and velocity of information

such that traditional analytic methods are stretched beyond their

limits (Megahed and Jones-Farmer 2015). This definition cap-

tures the multidimensionality of the concept, as “bigness”of data

is relative to the constantly expanding capacity to store and pro-

cess information (Guarnieri 2016), as well as to the intended util-

ity (McKinsey Global Institute 2011).

Despite its massive potential, firms struggle to derive meaningful

value from big data. Wieland et al. (2016, 207) note that compa-

nies are finding themselves in a situation of “big data, but small

math,”and exploitation of this growing data resource requires more

than accumulation and storage. A complementary issueis that busi-

nesses are careful of the level of access granted to sensitive internal

data, which drive trepidation in enlisting the help of outside experts

who may be a valuable resource for translating this data into

actionable information. This is especially true if the value of col-

laboration is unclear to decision makers within the firm. Despite

growing interest and multiple calls by editors to pursue practically

and theoretically relevant big data research (Waller and Fawcett

2013a,b), academics with the analytic tools and training to pursue

such research often face difficulty gaining access to corporate data.

This manifestation of the gap between research and practi-

tioner communities (Hutt and Walker 2015), stems from logical,

temporal, and incentive-based differences in how the two parties

approach problems in their work. Management practitioners are

concerned with urgent practical problems, expect results quickly,

and are more often rewarded based on short-term measurable

goals (Hutt 2008), whereas academics have tended to prefer

research on more comprehensive (although oftentime less urgent)

underlying causes for observed phenomena. This methodical

approach inevitably delays results, as research entails an exami-

nation not only of the practical phenomena but also of the exist-

ing body of knowledge pertinent to the subject and all possible

contributing factors. Researchers are often rewarded more for this

epistemological “truth”than for the managerial relevance of their

work (Bartunek and Rynes 2014).

Successfully bridging the research–practice gap requires practi-

tioners and academics to find common ground. Van de Ven

(2007), Avenier and Cajaiba (2012), and Hutt and Walker (2015)

all propose an iterative two-way dialog between the parties to

forge mutually beneficial research questions. This dialog often

initiated by and facilitated through an ongoing corporate research

partnership program such as those that exist at many research-

oriented business schools (Hutt 2008), aims to identify recurring

challenges and opportunities of practical relevance that may also

contribute to an increased understanding of more generalizable

management phenomena. Researchers in applied fields seeking to

gain access to “big”data sets to advance epistemological under-

standing of business phenomena must engage in this dialog to

Corresponding author:

Kevin B. Smyth, Department of Marketing and Logistics, Fisher

College of Business, The Ohio State University, 2100 Neil Ave,

Columbus, OH 43210, USA; E-mail: smyth.43@osu.edu

Journal of Business Logistics, 2018, 39(3): 203–219 doi: 10.1111/jbl.12187

© 2018 Council of Supply Chain Management Professionals

identify and balance the obligations they hold to both practice

and the academy (Cotteleer and Wan 2016).

With this perspective in mind, the goals of this study are as

follows: (1) to describe a practical roadmap for collaboration

between researchers and practitioners pursuing big data research

and (2) to detail a case example of how, by following this road-

map, researchers can provide insight into a firm on a specific

supply chain problem while developing a replicable template for

effective analysis of big data. The case example highlights our

experience working on a big data study to find the sources of

replenishment forecast deviation and bias in the quick service

restaurant industry. Key contributions of the current article are to

propose a process of conducting research that is of mutual value

to practitioners and researchers, demonstrate the value of pairing

a priori theorizing with big data exploration as proposed in Wal-

ler and Fawcett (2013a), describe unique challenges involved in

big data research, and develop a novel and replicable hierarchical

regression-based process for analyzing big data.

The remainder of the study is organized as follows: We describe

the origin and context of our big data case study, elucidate unique

technical challenges in collection, manipulation, and traditional

methods of exploration of big data, propose an alternative novel

hierarchical regression-based approach to explore big data, and

finally present a process for future practitioner-academic research

collaborations derived from our case study experience.

RELATIONSHIP DEVELOPMENT

Before describing the proposed facilitating process for big data

research engagement of academics and business practitioners, we

outline our specific experience. The process is organized into

subprocesses of relationship identification, research motivation,

project management, and findings validation. It should be noted

that the description of the findings validation subprocess will

occur after we describe the proposed facilitating process.

Relationship identification

The origins of this study stem from an ongoing relationship with

members of a practitioner–academic research group at a large

Midwestern university. Through the group, the primary fourth-

party logistics (4PL) provider for a major international quick ser-

vice restaurant enlisted our assistance in helping address multiple

supply chain issues their major restaurant customer was experi-

encing. The fact that the 4PL service provider felt comfortable in

approaching the team for assistance was based on their experi-

ence with academic researchers through their participation in the

practitioner–academic research group and other academic rela-

tionships that their organization had developed. The level of

experience a firm has working with academics critically impacts

how researchers generate initial interest and foster a strong ongo-

ing relationship for collaborative research.

Research motivation

Following the guidance of Hutt and Walker (2015), our research

team engaged in a bilateral collaborative dialog to translate the

practitioner’s challenges into mutually beneficial research

questions. One of the issues identified was that the major restau-

rant firm was experiencing significant order deviation by individ-

ual restaurant outlets in their centrally developed replenishment

forecasts, causing costly inefficiencies and exaggerated responses

in multiple levels of the firm’s supply chain.

After this initial problem identification, we visited the head-

quarters of the 4PL provider to understand the work the firm

conducted for the restaurant chain. Through interviews with man-

agers and analysts across multiple divisions in the 4PL, as well

as personnel from one of the five third-party logistics (3PL) com-

panies that provide distribution services for the restaurant chain,

we began to understand the connections between various inter-

ests that affect the focal issue of replenishment forecast

deviation.

The restaurant firm in our research operates almost 15,000

retail outlets domestically, of which more than 80% are con-

tracted to franchise companies. The firm utilizes the aforemen-

tioned 4PL to centrally develop sales forecasts and plan

replenishment for all restaurant outlets. All involved firms—the

restaurant firm, its 4PL provider, its 3PL providers, and the fran-

chise companies—utilize a management information system

(MIS) curated by the 4PL, thus providing a centralized source

for big data across the involved companies. Each entity’s rele-

vant MIS data are visible to at least the adjacent link in the sup-

ply chain. Figure 1 illustrates the relevant data flows within and

between firms.

Driven by data or by theory?

These data that permit exploration of causes of replenishment

forecast deviation clearly represent big data, because for almost

15,000 distributed outlets and 8,100 stock units, there are more

than 120 million potential daily transactions to evaluate. With

daily frequency, this data also had a velocity consistent with

incipient definitions of big data (McAfee and Brynjolfsson 2012;

Kitchin and McArdle 2016). While the variety of information

forms may not span to unstructured data, drawing from multiple

databases for a useful sample and mixing numeric and non-

numeric data are common features of big data analysis (Megahed

and Jones-Farmer 2015).

The generation of useful knowledge from this ocean of data is

increasingly achieved through a synthesis of data science,

1

pre-

dictive analytics,

2

and big data, referred to as DPB (Waller and

Fawcett 2013b). While use of DPB has expanded rapidly in

industry and in practitioner literature, it is still a matter of intense

debate for academics. Several generic approaches exist for big

data mining or exploratory pattern recognition that can identify

correlative relationships and aid immensely in prediction (Hand

et al. 2001; Han et al. 2011; Kuhn and Johnson 2013). However,

they lack the means to explain the causal mechanisms of their

predictions and appear to challenge the paradigm of a priori

theorizing.

1

Data science is defined here as the study of the generalizable

extraction of knowledge from data (Dhar 2013).

2

We define predictive analytics as a broad class of statistical or

analytic techniques used to develop predictions of otherwise

unknown future events or behavior (Nyce 2007).

204 K. B. Smyth et al.