Inventory record inaccuracy: Causes and labor effects

AuthorHoward Hao‐Chun Chuang,Rogelio Oliva
Date01 November 2015
DOIhttp://doi.org/10.1016/j.jom.2015.07.006
Published date01 November 2015
Inventory record inaccuracy: Causes and labor effects
Howard Hao-Chun Chuang
a
, Rogelio Oliva
b
,
*
a
College of Commerce, National Chengchi University, Taipei 11605, Taiwan
b
Mays Business School, Texas A&M University, College Station, TX 77843, USA
article info
Article history:
Available online 31 July 2015
Accepted by Daniel R. Guide
Keywords:
Retail operations
Store execution
Inventory record inaccuracy
System dynamics
Design of experiments
Bayesian inference
Econometrics
abstract
Inventory record inaccuracy (IRI) is a pervasive problem in retailing and causes non-trivial prot loss. In
response to retailersinterest in identifying antecedents and consequences of IRI, we present a study that
comprises multiple modeling initiatives. We rst develop a dynamic simulation model to compare and
contrast impacts of different operational errors in a continuous (Q,R) inventory system through a full-
factorial experimental design. While backroom and shelf shrinkage are found to be predominant
drivers of IRI, the other three errors related to recording and shelving have negligible impacts on IRI.
Next, we empirically assess the relationships between labor availability and IRI using longitudinal data
from ve stores in a global retail chain. After deriving a robust measure of IRI through Bayesian
computation and estimating panel data models, we nd strong evidence that full-time labor reduces IRI
whereas part-time labor fails to alleviate it. Further, we articulate the reinforcing relationships between
labor and IRI by formally assessing the gain of the feedback loop based on our empirical ndings and
analyzing immediate, intermediate, and long-term impacts of IRI on labor availability. The feedback
modeling effort not only integrates ndings from simulation and econometric analysis but also struc-
turally explores the impacts of current practices. We conclude by discussing implications of our ndings
for practitioners and researchers.
©2015 Elsevier B.V. All rights reserved.
1. Introduction
Inventory record inaccuracy (IRI) refers to the discrepancy be-
tween physical and recorded inventory levels, and is a pervasive
problem in retailing. Kok and Shang (2014)conclude that IRI can be
attributed to shrinkage (e.g., spoilage and theft), transaction errors,
and misplacement. Because it is difcult to fully eliminate these
execution errors, IRI becomes a norm rather than an anomaly in the
retail sector. Kang and Gershwin (2005) report that inventory ac-
curacy in a global retailer is on average only 51%. DeHoratius and
Raman (2008) nd 65% of the inventory records at a retail chain
to be inaccurate, and Oliva et al. (2015)observe that more than 60%
of SKUs in a European retail store have IRI. Most surprisingly, in a
retail store that had not even started operating, Raman et al. (2001)
found that the system had incorrect records for 29% of the items
and estimated that IRI reduces a companys total prots by 10%. At
the rm level, IRI can signicantly distort aggregate book value of
inventory and business decisions. At the item level, IRI can delay
ordering decisions because most extant inventory models do not
differentiate between physical and system inventories. IRI also in-
terrupts shelf replenishment even when there is plenty of in-
ventory in the backroom. Consequently, retailers suffer severe out-
of-stock (OOS) and signicant economic loss.
To tackle IRI and associated OOS in retailing environments,
radio-frequency identication (RFID) has been deemed as a
promising solution (Heese, 2007; Lee and Ozer, 2007). However,
issues such as cost, ownership, and privacy/security hinder the full
implementation of RFID at the item-level (Kapoor et al., 2009). Even
when RFID becomes cheap enough to be fully adopted like bar-
coding, the fact that retail operations is a complicated issue
involving people, processes, and technology makes error-free op-
erations extremely difcult to achieve. In order for retailers to
enhance execution quality and data integrity, it is important for
managers to understand the causes of IRI and identify the policy
levers that they can use to reduce it.
While some empirical work has focused on product and store
attributes that affect IRI (e.g., DeHoratius and Raman, 2008), in this
work we explore the impact of store stafng levels and operational
performance on IRI. Our study comprises multiple modeling ini-
tiatives. First, grounded on empirical observations and eld work,
*Corresponding author. Fax: þ1 979 845 5653.
E-mail addresses: chuang@nccu.edu.tw (H.H.-C. Chuang), roliva@tamu.edu
(R. Oliva).
Contents lists available at ScienceDirect
Journal of Operations Management
journal homepage: www.elsevier.com/locate/jom
http://dx.doi.org/10.1016/j.jom.2015.07.006
0272-6963/©2015 Elsevier B.V. All rights reserved.
Journal of Operations Management 39-40 (2015) 63e78
we formulate a dynamic model of continuous review (Q,R) in-
ventory system and explicitly incorporate multiple execution errors
into the model. The (Q,R) policy is often used for fast moving
products and widely adopted by retailers, including numerous
mass merchants that carry a large number of items (Kang, 2004;
Kang and Gershwin, 2005) and stores that we work with. To
compare and contrast the impact of different errors and their in-
teractions, we conducted a full-factorial experimental design. We
nd that backroom shrinkage and shelf shrinkage errors are the
dominant drivers of IRI and that, under-shelving, along with erro-
neous checkout and data capture, have negligible impact on IRI
when compared to shrinkage. We also nd that the interaction
effects between error sources are non-substantial and mostly seem
additive and linear. These primary ndings hold under different
distributional assumptions and parameter settings.
Next, we investigated the relationships between labor avail-
ability and IRI using longitudinal data from ve stores in a global
retail chain. After deriving a robust measure of IRI through Bayesian
computation and estimating panel data models that control for
store-section and time xed effects, we nd strong evidence that
more full-time labor reduces IRI whereas part-time labor fails to
alleviate it. Finally, we articulate the reinforcing relationships be-
tween labor and IRI by formally assessing the gain of the feedback
loop based on our empirical results. We nd that the work pressure
introduced by IRI does further increase IRI, but the gain of feedback
loop is not enough to compound its growth. We also analyze the
intermediate and long-term effects of IRI on labor availability and
use the developed structure to assess the impact of current stafng
practices on performance.
Our paper contributes to practice and theory in four signicant
ways. First, the simulation model has a simple but realistic struc-
ture that addresses the issue that most retail inventory models
ignore ethe dynamics between the retail shelf and the backroom
used for extra storage (Eroglu et al., 2013)eandallows for a joint
assessment of the relative impact of operational errors in IRI. Sec-
ond, despite the abundance of optimization models developed to
tackle IRI, empirical investigations are scant. By econometrically
estimating the effects of labor allocation on IRI, we broaden
empirical knowledge of IRI and develop new research opportu-
nities. Retail managers should be aware of labor effects on data
quality, which is deemed to be an important source for competitive
advantage (Redman, 1995). Third, we articulate the reinforcing re-
lationships between labor and IRI by formally assessing the gain of
the feedback loop based on our empirical ndings and analyzing
immediate, intermediate, and long-term impacts of IRI on labor
availability. The feedback modeling effort not only integrates
ndings from simulation and econometric analysis but also struc-
turally explores the implications of current practices. Last, we
illustrate the utility of a joint use of system dynamics and econo-
metrics. Such a combination widens our ability to answer questions
of what-if and what-is given unobservable factors (i.e., execution
errors) and limited observations of IRI over time. Using dynamic
simulation, Bayesian shrinkage estimation, panel data modeling,
and causal loop modeling enhances our understanding of IRI while
responding to the call for adopting multiple methods (Boyer and
Swink, 2008).
The rest of our article is organized as follows. Section 2briey
reviews relevant literature to frame our contribution. Section 3
presents a continuous-time simulation analysis that enables us to
identify the main drivers of IRI. We then postulate and articulate
how those drivers of IRI are associated to store labor. Section 4
shows econometrical estimation results of labor availability on
IRI. Section 5presents feedback loops and behavioral dynamics
associated with the impact of IRI on labor availability. We conclude
by discussing managerial and theoretical implications of our
ndings.
2. Literature review
A signicant number of studies have attempted to analyze
causes and effects of IRI in recent years (e.g., Fleisch and Tellkamp,
2005; DeHoratius and Raman, 2008). Due to the randomness of
errors that cause IRI and uncertainties in the distribution of IRI,
simulation has been widely adopted to assess the effect of IRI on a
retail supply chain (Fleisch and Tellkamp, 2005) or a retail outlet
(Nachtmann et al., 2010). Among simulation studies on IRI, the
continuous review (Q,R) system has been the focus of investigation.
Kang and Gershwin (2005) analyzed how stock loss (shrinkage)
causes IRI and severe OOS. They found that OOS increases mono-
tonically in stock loss. Thiel et al. (2010)simulated the impact of IRI
on service level and in contrast with Kang and Gershwin (2005),
they observed that OOS is not a monotonic function of IRI when
error rate is symmetric with a zero mean. Following Kang and
Gershwin (2005),Agrawal and Sharda (2012) concentrated on IRI
attributed to stock loss, and examined how the frequency of in-
ventory audit affects OOS and average inventory. Similarly, in the
rst part of our paper we develop a dynamic simulation model of
the (Q,R) inventory system. Our model differs from the afore-
mentioned studies in two ways. First, while most models address a
single source of error (Sahin and Dallery, 2009), Lee and Ozer
(2007) point out that modeling efforts are needed to articulate
the joint effect of multiple errors. Our model takes into account
multiple errors (both operational and information-related) simul-
taneously. Although Fleisch and Tellkamp (2005) also assessed the
impact of several errors using stochastic simulation, we analyze
operations inside a retail store instead of ows in a three-echelon
supply chain. Second, while existing simulation studies on IRI
stress the consequences (e.g., inventory level, ll rate) of poor data
quality (Nachtmann et al., 2010), our analysis focuses on the im-
pacts of different antecedents of IRI.
While simulation analysis enhances our understanding about
antecedents and consequences of IRI, there is still limited empirical
knowledge about IRI due to the low availability of data. Few studies
empirically investigate IRI through analyzing actual data on in-
ventory discrepancies. Sheppard and Brown (1993) presented the
rst analysis to empirically assess how product-related factors
affect IRI within a manufacturing plant. In retail stores, Raman
(2000) and Oliva et al. (2015) both found that more than 60% of
items had inaccurate records. Using data from a single store, Oliva
et al. (2015) derived empirical estimates of an aggregate model
that characterizes inventory information decay. The estimated
functional form is further incorporated into inspection policy
design. To our knowledge, the only cross-store econometric anal-
ysis of IRI is by DeHoratius and Raman (2008). They collected cross-
sectional data on IRI from a retail chain to empirically examine IRI.
The econometric analysis performed in the second part of our paper
differs from DeHoratius and Raman (2008) in three important
ways. First, expanding their efforts on examining how product, and
store, related attributes affect IRI, we assess the association be-
tween labor decisions and IRI in each product sector. Second, we
obtain longitudinal observations of IRI and labor decisions, which
allow us to test labor effects while tackling unobserved factors.
Third, our econometric estimation focuses on developing an oper-
ational functional form for the impact of labor on IRI (Richmond,
1993), as opposed to a correlational study to test hypotheses.
Finally, our work is also informed by system dynamics
(Forrester,1958; Sterman, 2000) efforts to assess the impact of la-
bor and stafng levels on operational performance (e.g., Anderson,
2001; Oliva and Sterman, 2001; Lyneis and Ford, 2007). While we
adopt from these articles the feedback perspective on stafng
H.H.-C. Chuang, R. Oliva / Journal of Operations Management 39-40 (2015) 63e7864

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT