Investigating the Effects of Daily Inventory Record Inaccuracy in Multichannel Retailing

• Published date: 01 September 2013
• Author: Anníbal C. Sodero,Mark Barratt,Elliot Rabinovich,Thomas J. Kull
• Date: 01 September 2013
• DOI: http://doi.org/10.1111/jbl.12019

Investigating the Effects of Daily Inventory Record Inaccuracy in

Multichannel Retailing

Thomas J. Kull

1

, Mark Barratt

2

, Anníbal C. Sodero

3

, and Elliot Rabinovich

1

1

Arizona State University

2

Marquette University

3

University of Arkansas

Inventory record inaccuracy (IRI) challenges multichannel retailers in fulfilling both brick-and-mortar and direct channel demands from their

distribution centers. The nature and damaging effects of IRI largely go unnoticed because retailers assume daily IRI remains stable over time

within the replenishment cycle. While research shows that a high level of IRI is damaging, in reality the level of IRI can change every day.

We posit that daily IRI variation increases the uncertainty in the system to negatively affect inventory and service levels. Our research uses data

collected daily from a multichannel retailer to ground a discrete-event simulation experiment. Going beyond testing just the level of IRI, we

evaluate daily IRI variation’s impact on operating performance. What we find in our empirical data challenges extant assumptions regarding the

characteristics of IRI. In addition, our simulation results reveal that daily IRI variation has a paradoxical effect: it increases inventory levels

while also decreasing service levels. Moreover, we also reveal that brick-and-mortar and direct channels are impacted differently. Our findings

show that assumptions and practices that ignore daily IRI variation need revising. For managers, we demonstrate how periods of multiday

counting help assess their daily IRI variation and indicate what the causes may be.

Keywords: inventory record inaccuracy; multichannel retailing; inventory management; distribution center

INTRODUCTION

To manage inventory complexities present across brick-and-mor-

tar and direct (i.e., Internet-based) channels, retailers commonly

use centralized decision support systems to serve both of these

channels from a single distribution center (DC) (Blankley et al.

2008; Galbreth and LeBlanc 2010). These systems assume that

the retailer’ssystem inventory record (SIR) is accurate, even

though widespread inventory record inaccuracy (IRI) occurs in

practice (Raman et al. 2001; Rekik 2011). Research shows that

IRI, which is the relative discrepancy between the SIR and the

actual inventory on-hand of a stock-keeping unit (SKU), is dam-

aging to operating performance (Waller et al. 2006; DeHoratius

and Raman 2008). However, because the level of IRI is continu-

ally changing, managers cannot make one-time adjustments to

account for it, but must assess and manage the variability of IRI

to maintain DC operating performance and predictability (Barratt

et al. 2010).

In this study, we examine the operating performance effects of

daily IRI variation within replenishment cycles (i.e., the interval

between successive supplier deliveries) of a multichannel retailer

DC. Daily IRI variation is the degree to which SIR and actual

inventory differ on a day-to-day basis. In practice, because multi-

channel managers realize IRI damages channel performance

(Metters and Walton 2007), they perform cycle counts and audits

at single points in time to identify and correct IRI (Neeley

1983). This practice assumes that these occasional, periodic point

estimates represent the true level of IRI, that is, there is stability

during the replenishment cycle and between cycle counts. In this

paper, we test this assumption and investigate daily IRI variation

and its effects on operating performance under different demand

patterns (i.e., order size and frequency) within multichannel

retailing. Our study aims to answer the following main questions:

In the context of a multichannel retailer’s DC, to what extent

does daily IRI variation affect performance? Do channel demand

patterns alter the effects of daily IRI variation?

Our research employs a multimethod approach (Mentzer and

Flint 1997; Sanders and Wagner 2011), using both empirical and

simulated daily IRI data to answer these questions. First, we

review the impact that daily IRI variation has within the replen-

ishment cycle and propose a paradoxical effect: that daily IRI

variation increases inventory levels while also decreasing service

levels (i.e., order fill rate). We also propose that direct channel

demand patterns are more sensitive to the uncertainty created by

this phenomenon.

Second, our research uses empirical, multiday IRI data from a

multichannel retailer to investigate common assumptions of daily

IRI variation. We also decouple SIR errors into those that are

transaction-dependent (TD) and transaction-independent (TI)

while accounting for the effects of inventory policies on perfor-

mance. Finally, similar to recent logistics research (e.g., Shapiro

and Wagner 2009; Torres and Maltz 2010), we use our data to

ground a simulation with daily IRI variation and test the hypoth-

esized effects on DC performance (i.e., customer service and

inventory level). Simulation is appropriate because it: (1) incor-

porates a high level of detail regarding the factors of interest, (2)

accommodates nonlinearities essential to IRI research (e.g., fre-

quencies in cycle counting and record corrections), and (3)

accounts for stochastic elements in lead times, SIR errors, and

demand across channels (Bowersox and Closs 1989).

We find that daily IRI variation increases inventory levels but

decreases service levels, and that an interaction exists between

daily IRI variation and channel demand patterns. Also, TI errors

Corresponding author:

Thomas J. Kull, Department of Supply Chain Management, W. P.

Carey School of Business, Arizona State University, PO Box

874706, Tempe, AZ 85287-4706, USA; E-mail: thomas.kull@asu.

edu

Journal of Business Logistics, 2013, 34(3): 189–208

© Council of Supply Chain Management Professionals

seem to be more damaging throughout the inventory system than

TD errors. This reveals to managers the importance of knowing

what type of errors induce daily IRI variation, so they may

reduce the problem instead of using inventory to buffer against

the problem.

Our study makes three primary contributions. First, it uses lon-

gitudinal evidence to extend the literature beyond current simpli-

fying assumptions of daily IRI variation (cf. K€

ok and Shang

2007; DeHoratius et al. 2008), giving a more accurate descrip-

tion for researchers to use. Second, it demonstrates how daily

IRI variation is a detrimental phenomenon that managers should

be aware of as they seek to avoid stock-outs and improve DC

performance (Rabinovich 2005). Third, it presents an approach

that managers can use to begin assessing and controlling the

problems that cause daily IRI variation. Thus, our research leads

to a better understanding of IRI behavior in DCs and suggests

how managers can respond to further improve DC performance.

In the next two sections, we review the literature to propose

direct and interaction hypotheses about daily IRI variation’s

effect on DC operating performance. Next, we present our empir-

ical method and results, followed by our simulation method and

results. We conclude with a discussion of our findings, contribu-

tions, and future research opportunities. Finally, we refer the

reader to the Appendix.

LITERATURE REVIEW

Investigations concerning IRI

Since the early works of Rinehart (1960), Schrady (1970), and

Iglehart and Morey (1972), the extant literature has focused on

the conspicuous presence of IRI in distinct contexts: manufactur-

ing (Graff 1987; Brown et al. 2001), retailing (Morey 1985;

Raman et al. 2001; Corsten and Gruen 2003; Nachtmann et al.

2010), and the extended supply chain (Delen et al. 2007; Heese

2007; Ukun et al. 2008). Common across this literature is the

attempt by researchers to map the causes and consequences of

IRI and to provide managerial guidance regarding ideal or opti-

mal approaches for tackling the problem with inventory manage-

ment practices.

This literature on IRI may be segregated into two streams of

research: empirical and analytical. The empirical stream, which

primarily investigates correlations between operating conditions

and the presence of IRI, is best represented by DeHoratius and

Raman (2008). In that study, data from a retailer are used to

develop a framework relating several factors that mitigate or

exacerbate IRI, such as auditing practices, product variety, sales

velocity, price, retail stores environment, and distribution struc-

ture. DeHoratius and Raman (2008) argue that these factors

should be incorporated into inventory planning tools to account

for the presence of IRI. While empirical research has shown the

influence and variability of IRI, there remains a lack of empirical

characterizations of how IRI varies over time and of the errors

that drive IRI variation, and the impact that this variation has on

performance.

The analytical stream of IRI literature, by contrast, primarily

considers auditing policies and base-stock levels to minimize

inspection and inventory holding costs when IRI is present (e.g.,

Fleisch and Tellkamp 2005; Delen et al. 2007). Kang and Gersh-

win (2005) use simulation to demonstrate how even small levels

(1%–3%) of IRI during replenishment cycles lead to severe

stock-outs. Camdereli and Swaminathan (2005) describe how IRI

influences optimal replenishment policy decisions and coordinat-

ing contracts in a single-period, single-location system. K€

ok and

Shang (2007) develop a joint inspection and replenishment pol-

icy that minimizes total costs in a finite horizon, while DeHora-

tius et al. (2008) propose replenishment policies that account for

errors using a Bayesian updating of error distribution.

Taken as a whole, analytical research has reinforced the notion

that the replenishment cycle is a crucial process in scheduling

periodic inventory counts to manage IRI. However, it has yet to

show how measuring and managing daily IRI variations within

replenishment cycles can impact DC performance. In addition,

the above research commonly makes several simplifying assump-

tions. First, it assumes that errors inducing daily IRI variation are

identically distributed and independent of demand and of the

channels through which demand arrives. Therefore, this research

assumes that IRI follows a simple random-walk pattern over

time.

1

Second, the research assumes that managers will only

know of inaccuracies as a result of scheduled cycle counts.

Third, it assumes that IRI variation involves only a single SKU,

as opposed to a wider range of SKUs, which is more common in

practice. By contrast, our research uses an empirically grounded

simulation to examine these assumptions while showing how DC

performance is affected by daily IRI variation across multiple

channels. In addition, because managers may look to increase

inventory to protect against IRI, we examine how replenishment

policies interact with IRI-inducing errors to exacerbate the

problem.

Drivers of IRI

Since IRI is the logistics equivalent of a manufacturing defect

(Ernst et al. 1992), the manifestation of IRI is linked to SIR

errors that are akin to unscheduled downtimes in manufacturing

and material handling systems (Banks et al. 2000). Such SIR

errors may be classified into two groups: (1) TD errors and (2)

TI errors (Lee and €

Ozer 2007; Nachtmann et al. 2010). TD

errors are changes in IRI as triggered by replenishments, demand

orders, or product returns (Lee and €

Ozer 2007). Errors in this

category can be due to: (1) incorrect deliveries, (2) misplaced

items, or (3) incorrect picking (Kang and Gershwin 2005). TI

errors are changes in IRI that occur irrespective of transaction

and are influenced by the amount of inventory on-hand (DeHora-

tius and Raman 2008). Such errors are related to: (1) internal

materials movement, and (2) shrinkage from theft, spoilage, or

damage (Kang and Gershwin 2005).

Channel characteristics also affect IRI. In some cases, multi-

channel retailers fulfill both their brick-and-mortar and direct

channel demands from a single-location DC. While this allows

for the pooling of inventory to reduce stock-outs (Ton and

Raman 2010) and facilitates the coordination of operations

(Metters and Walton 2007), there is speculation that the inherent

differences in both channels may affect how IRI occurs and

influences DC operating performance (Agatz et al. 2008). In

1

This can be represented by the designation ARIMA (0, 1, 0).

190 T. J. Kull et al.