Integrating and Exploiting Information

and Communication Technologies

(ICT) in Restaurant Operations:
Implications
for Restaurant Productivity
Marianna Sigala

ABSTRACT. Despite the continuous increase of investment in Infor-

mation and Communication Technologies (ICT), empirical studies have
not persuasively established corresponding increases in productivity. In
contrast, many studies investigating the ICT impact have found no sig-
nificant relationships between productivity and ICT. However, several
shortcomings have been identified in past studies, e.g., measurement er-
rors, redistribution of impacts and mismanagement of ICT. This study
proposes a methodology for assessing the ICT productivity impact that
overcomes these shortcomings. The methodology is tested in a dataset
from hotel restaurants in the UK by using a non-parametric technique
called Data Envelopment Analysis (DEA). Empirical findings revealed
that productivity gains do not accrue from ICT investments per se, but by
exploiting the ICT informate and networking capabilities for rede-
signing and streamlining operations. Suggestions for enhancing the ICT
productivity impact are provided. [Article copies available for a fee from The
Haworth Document Delivery Service: 1-800-HAWORTH. E-mail address:
<docdelivery@haworthpress.com> Website: <http://www.HaworthPress.com>
2003 by The Haworth Press, Inc. All rights reserved.]

Marianna Sigala, PhD, is Lecturer in Operations and Production Management,

Business Administration Department, University of the Aegean, 8 Michalon Street,
Chios, Greece (E-mail: M.Sigala@strath.ac.uk).
Journal of Foodservice Business Research, Vol. 6(3) 2003
http://www.haworthpress.com/web/JFBR
2003 by The Haworth Press, Inc. All rights reserved.
10.1300/J369v06n03_05 55
56 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

KEYWORDS. IT, restaurants, information, communication, productiv-

ity technology

INTRODUCTION

Despite the increasing investments in ICT, research findings investigating

the ICT productivity impact have always led to contradictory and/or question-
able results. Robert Solow, a Nobel winning economist, is supposed to have
said that PCs are showing up all over the place, except in productivity statis-
tics, (in Lucas, 1993: 8), while Brynjolfsson (1993) first referred to the con-
cept of the IT productivity paradox, i.e., the fact that the benefits of IT
spending have not shown up in aggregate output statistics. However, as sev-
eral methodological shortcomings have been identified in past studies (e.g.,
Brynjolfsson, 1993), new IT evaluation methodologies are required to lessen
or eliminate these. Coupled with the increasing spending and importance of
ICT in the tourism and hospitality sectors, it is apparent that an investigation
into the impact of ICT on productivity is warranted. This paper aims to investi-
gate the ICT productivity paradox by developing a methodology that is over-
coming the previous studies methodological shortcomings. To that end, after
analyzing the latter, a framework for measuring the productivity gains of ICT
investments based on Data Envelopment Analysis (DEA) is proposed. The
framework is empirically tested by using data gathered from four star hotel
restaurants in the UK. A model of ICT applications and configurations for en-
hancing ICT productivity impact is developed.

Investigating the ICT Productivity Paradox

The seemingly obvious, yet elusive, relationship between ICT and produc-
tivity has accumulated a great body of research exploring the productivity im-
pact of ICT investments on four different levels (i.e., on the economy, industry
and at the firm- and process-specific levels). Several authors summarise an ex-
tensive number of studies investigating the relationship between ICT and pro-
ductivity (e.g., Brynjolfsson, 1993; Hitt and Brynjolfsson, 1996; Lucas,
1993). However, results from studies are plagued with ambiguities and incon-
sistencies. Some researchers reported no relationship between ICT investment
and improvements in productivity (e.g., Strassmann, 1990; Dos Santos et al.,
1993; Byrd and Marshall, 1997), some others provided evidence that such re-
lationship does exist (e.g., Bender, 1986; Brynjolfsson, 1993; Roach, 1988).
Few studies shown negative/dysfunctional ICT productivity effects (e.g., Weill,
1992). Research within the hospitality sector has been limited, but it draws to
 Marianna Sigala 57

similar conclusions (Sigala, 2002). Thus, non-conclusive evidence for the ICT
productivity impact is provided. Nevertheless, as past studies have been ques-
tioned on methodological grounds, findings reporting a negative ICT impact
on productivity are claimed to be statistical artefacts. Methodological issues
affecting research quality on the ICT-productivity relation are analyzed as fol-
lows, but nonetheless, the productivity paradox is due to a combination of all
these factors.
The Quality of the Data Used and Analyzed. A few studies relied on ques-
tionable data gathered for other purposes, while others did not control for con-
textual factors (Byrd and Marshall, 1997). Cron and Sobol (1983) and
Strassmann (1990) also suggested that ICT have an amplifier effect meaning
that the introduction of ICT into poorly run firms does not increase productiv-
ity, whereas the introduction of ICT into well-run firms pay-off. Previous re-
search that simply incorporated ICT as an input factor of productivity
functions did not consider this issue. Thus, future research should firstly iden-
tify high and low performers and then investigate the impact of ICT on each
one separately.

The Metrics Measuring Productivity

Because of the difficulties in measuring and isolating the soft/qualitative

dimensions of productivity, there has been a misconception that productivity
metrics cannot capture the full impact of ICT (e.g., quality increases, avoid-
ance of competitive disadvantage). However, such methodological issues and
concerns are depended on and closely interlinked with the problems related to
productivity definition and measurement (Witt and Witt, 1989). Indeed, a
widely accepted productivity definition cannot be found in the literature
(Brown and Dev, 1999; Johns et al, 1997). Productivity has been approached
as an umbrella concept including efficiency, effectiveness, quality, predict-
ability and other performance dimensions, as well as a narrower concept re-
flecting only production efficiency (Sigala, 2004b). Consequently, confusion
and disagreement over the concept and definition of productivity create diffi-
culties in productivity measurement. For example, in the restaurant context,
some productivity measurements relate to production efficiency (e.g., material
costs, number of covers served to number of staff, sales per employee, faster
table turns), other measurements relate to outcomes (e.g., sales, profit per
available seat hour, customer satisfaction) reflecting effectiveness. While effi-
ciency measures show whether an organization is doing things in the right
way, they may not indicate effectiveness and hence whether the organization
is doing the right things. These problems are encountered when attempting
to define precisely what are the outputs and inputs of a given industry and what
58 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

are the appropriate units for measuring the latter. Such difficulties are particu-
larly difficult in the hospitality industry whose outputs/inputs are intangible
and highly heterogeneous and whose production process requires good labor
and process scheduling due to the demand and consistency variations. How-
ever, research in the hospitality industry (Sigala, 2002) revealed that for se-
lecting appropriate inputs/outputs and their measurement units two issues
need to be firstly addressed: the approach to productivity definition, namely
partial or total approach; the identification of the level and/or unit of analysis.
The implication of the first issue on productivity measurement is discussed be-
low, while the implications of the level of analysis are analyzed in the follow-
ing section.
The distinction between total and partial productivity metrics is based on
whether (Heap, 1992; Rimmington and Clark, 1996): (a) all or only one input
is considered; (b) tangible/hard and/or intangible/soft (management style,
staff flair and expertise) aspects of inputs/outputs are considered; (c) other fac-
tors affecting productivity are included (e.g., hotel design). Heap (1992) and
Rimmington and Clark (1996) were some of the strong advocates of a total ap-
proach to productivity that supported that productivity measurement should
include all factors (soft and hard) affecting productivity, namely, the top-line
factors. However, although the importance of using a total approach to pro-
ductivity has been highlighted, authors have simultaneously stressed the diffi-
culty to encompass all inputs/outputs and/or their tangible and intangible
elements within one productivity metric. At the same time, it is also widely
agreed that different measurement units imply the inclusion or not of soft
and/or hard productivity dimensions. Specifically, it is generally agreed that
quantitative physical measures reflect a quantitative approach to productivity
that equates productivity with production efficiency only (Andersson, 1996),
while a total factor approach would require more sophisticated and qualitative
measures. However, there are arguments supporting the view that the truly
quantitative, aggregate, broad measures (e.g., profit, sales) implicitly encap-
sulate intangible qualitative performance (Rimmington & Clark, 1996; Johns &
Wheeler, 1991; Gummeson, 1998; Ball, 1993). This is for two reasons. First,
only if the intangibles are as they should be will customer levels be sustained
and income earned. Secondly, only if the tangibles are as they should be will
income and costs be controlled in such a way that profit is produced at the re-
quired rate in relation to the capital employed.
Moreover, irrespective of concerns regarding productivity measurement,
Jurison (1996) highlighted that the ICT productivity paradox is due to bad
management and not to mismeasurement of the ICT productivity benefits. In
other words, the problems lies not on how productivity is measured but rather
on firms failure to translate intermediate ICT productivity benefits (e.g.,
 Marianna Sigala 59

better customer service) into final outcomes (e.g., charge higher prices). Res-
taurants ability to materialize and translate ICT productivity benefits is also
limited to the high labor turnover faced in the industry. Frequent changes in
the labor force do not allow restaurants to climb in the technology learning
curve, while the industry has also been characterized by a lack of skills and
competencies on its use and exploitation of ICT (Sigala et al., 2001). Finally,
awareness on ICT tools and capabilities also leads to a strategic myopia
whereby hospitality operators make ICT investment decisions based on imme-
diate ICT benefits and return on investment. However, as ICT benefits may
take some years to materialize, making ICT decisions and evaluations on short
term rather than long term benefits is strategically dangerous and may hinder
the gaining of competitive advantages in the long term.

THE METRICS MEASURING ICT

This is the major criticism of previous studies, since most of them give a
disproportionate emphasis on the cost of ICT equipment and expenditures.
ICT budgets are the most frequently used metric of computerisation as these
figures are readily available and reasonably objective, but crucial concerns ex-
ist regarding the reliability and validity of these metrics. ICT budgets do not
distinguish between different ICT tools and applications, while different ICT
applications can lead to different results (Lucas, 1993; Strassmann, 1990). The
use of ICT financial metrics for comparing ICT intensiveness across firms also
suffers from: fluctuations over time (ICT budgets depend on the firms accu-
mulated ICT assets and on ICT cost which is continually decreasing); waste of
ICT expenses; and different ways of financing and measuring ICT expendi-
tures (e.g., when ICT applications are outsourced through Application Service
Providers, ICT costs are considered as day-to-day expenses and not invest-
ments, but the firms outsourcing ICT will have the same ICT functionality as
firms that decided to buy/develop ICT in-house). ICT budgets neglect two im-
portant facets of ICT namely their deployment and their evolving capabilities
and features (Strassmann, 1990; Willcocks et al., 1998), and so, they fail to il-
lustrate how ICT provide business value. Others (e.g., Weill, 1992) argued
that ICT effectiveness (i.e., effective ICT implementation on time and budget,
ICT user friendliness and functionality, ICT acceptance and use by staff)
should also be measured, however, it is generally agreed (e.g., Willcocks et al.,
1998; Lucas, 1993) that this is an inaccurate and incomplete metric for mea-
suring the ICT productivity impact as it is a mediating factor of the ICT-pro-
ductivity relationship. Within the context of the hospitality industry, several
authors (e.g., Sigala et al., 2001; Werthner and Klein, 1999) argued that the re-
60 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

lationship between ICT and value is not a direct one, but ICT give value when
they are being used to redefine, differentiate and informationalise prod-
uct/services, streamline, rationalise and support value added processes. ICT
mismeasurement is also argued to lead to ICT mismanagement problems, i.e.,
inability to identify and exploit ICT applications and capabilities that can lead
to productivity gains (Sigala et al., 2001).

The Level of Analysis at Which Research Is Undertaken

This refers to the level of productivity and ICT measurement. Regarding

the level of productivity analysis, it is widely recognized that aggregated met-
rics of inputs/outputs tend to obscure information, while partial metrics tend to
hide information, trade offs and complementarities among other productivity
dimensions that have to do with either other business departments or other re-
sources/inputs. To avoid this, researchers attempt to consider partial metrics
simultaneously, but this is very laborious and sometimes may lead to conflict-
ing results (Baker and Riley, 1994). In addition, when partial and aggregated
productivity metrics are used in ICTproductivity studies, then the limitations
and issues related with the productivity metrics are also replicated and added
up to the problems related to the level of ICT measurement. Specifically, stud-
ies measuring productivity at the economy and industry level are limited be-
cause macro data do not capture firm level phenomena and they hide
displacement effects, i.e., ICT benefits in one industry balanced out by bene-
fits in another industry (Brynjolfsson, 1993). Menon (2000) argued that the
best level of analysis is at the organisational level, because substitution, syn-
ergy and complementarities between resources, inputs and factors affecting
productivity can be captured and that process-level analyses suffer from the
following serious problems: difficulty in data collection and insufficient sam-
ple size, since a significant number of firms with similar processes should be
found; difficulty in separating ICT effects from non-ICT effects within a pro-
cess; the validity in the generalisation of results arising from the difficulty of
finding similar processes performed with and without ICT across firms. Stud-
ies measuring the impact of specific ICT applications on business processes
are also limited because they ignore (Lucas, 1993): any ICT impact on other
processes (ICT applications can have dysfunctional effects on other pro-
cesses); impact on final outcomes (intermediate effects on processes may not
translate to final business outcomes); and synergy between different ICT ap-
plications. Indeed, an organisational level of analysis also seems the most ap-
propriate when considering that the ICT costs and benefits span multiple
departments and areas and so, one needs to include these multi-dimensional
effects and synergies/trades off.
 Marianna Sigala 61

To summarize these arguments, Dos Santos et al. (1993) classified ICT pro-
ductivity studies into two streams/categories namely, functional and aggre-
gate. Functional studies are limited in terms of the lack of a direct link between
the observed, functionally-related productivity effects and measurable pro-
ductivity in terms of overall organizational goals. On the other hand, studies
using aggregate data are important because they provide evidence that such in-
vestments could affect firms ultimate outcomes. However, their usefulness is
limited, because of their reliance on aggregate measures that do not allow the
investigation of the relationship between specific ICT applications and func-
tional productivity. Overall, Dos Santos et al. (1993: 521) concluded that re-
search is required that would overcome the limitations of the two previous
research streams by making a distinction between different IT investments and
taking into consideration that activity based results are not always translated
into firm level outcomes. To achieve that, the methodology of this study is a
mixed one, combining the strengths of these two streams of research.

The Statistical Method Used to Relate It with Productivity Metrics

The majority of studies have used regression and ratio analysis, which are
limited since they can simultaneously consider only a limited number of vari-
ables. For example, a productivity metric of revenue to number of employees
does not consider other factors of production, while aggregate productivity
metrics, e.g., total revenue to total expenses does not distinguish the produc-
tivity impact of different productivity inputs/outputs. Regression is also lim-
ited in investigating the effect of one input (or output) to multiple outputs (or
inputs). These techniques also assume away inefficiency in production, which
production functions are capable of modeling. Production function techniques
also consider multiple inputs and outputs simultaneously and so they have
been extensively used in ICT productivity studies. However, production func-
tions are parametric techniques meaning that a functional form for the technol-
ogy transforming inputs into outputs is assumed and so they can suffer from
specification error.
Because of that, a non-parametric, multivariate technique called Data En-
velopment Analysis (DEA) is used in this study. DEA benchmarks units by
comparing their ratios of multiple inputs to produce multiple outputs at the
same time and by using the concept of the performance frontier (Charnes,
Cooper and Rhodes, 1978; Charnes, Cooper, Lewin and Seiford, 1994). DEA
shares the advantages of production function, but it is specification error free
because it does not assume a functional form. Instead, DEA involves the esti-
mation of the best practice frontier from the sample data. In using DEA, the
productivity score of any unit is computed as the maximum of a ratio of
62 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

weighted outputs to weighted inputs, subject to the condition that for all other
units of the dataset, similar ratios are less than or equal to one. The productiv-
ity of a hotel can be obtained by solving the following model (M1) (Charnes,
Cooper and Rhodes, 1978):
t

U Y r rjo
Max h o = r =1
m
(M1)
V X
i =1
i ijo

subject to
t

U Y
r =1
r rj

m
1 for all j = 1, ...n.
V X
i =1
t ij

U r Vi > 0; r = 1, ... s; i = 1, ... m

The model has t output variables, m input variables and n production units.
Yrj and Xij are the amount of the rth output and the ith input for the jth unit, and
Ur and Vi are the weights to be estimated by the data of all comparable units
that are being used to arrive at the relative productivity for the oth unit. If a unit
is on the frontier isoquant, i.e., among the reference set, the solution will be ho =
1 and the productivity score is 1, which can be described as being 100% produc-
tive as compared with other units of the dataset. Other units, using these inputs
less efficiently, will locate above the frontier isoquant and their productivity
score will be smaller than 1. For example, a unit with a productivity score of
0.75 can be interpreted as being 75% as productive as a unit on the frontier
isoquant.
Other DEA advantages are reported as follows (Sengupta, 1988; Banker
and Morey, 1986). DEA identifies bad from good performers by generating an
overall, easy to interpret efficiency score, it identifies and measures the
amount of inefficiency areas, it is independent of the units measuring inputs
and outputs (giving flexibility in specifying outputs/inputs), and it can manip-
ulate uncontrollable, environmental factors, e.g., demand variation. Indeed,
Avkiran (1999) highlighted that failure to account for environmental factors is
likely to confound the DEA results and lead to unreliable analysis. Norman
and Stoker (1991) argued that DEA models not including demand factors mea-
 Marianna Sigala 63

sure production efficiency, while DEA models including them reflect market
efficiency, i.e., ability to control production efficiency given demand factors.
DEA has been extensively used for productivity measurement in various
industries (e.g., in banking Al-Faraj et al., 1993; for more industry examples
see Avkiran, 1999) the hospitality industry included (Sigala, 2002 and 2004b;
Tarim, 2000; Wber, 2002; Johns et al., 1997), as well as for measuring the
ICT productivity impact (Banker et al., 1990; Paradi et al., 1997; Dasgupta et
al., 1999; Shafer and Byrd, 2000). However, the validity and usefulness of
DEA crucially depends on the inputs and outputs used, and these studies pres-
ent several methodological limitations regarding their use of DEA for investi-
gating the ICT productivity impact: (a) use of very few and aggregated
productivity inputs and outputs; (b) ICT are included in DEA models as inputs
measured in financial terms which does not allow for distinguishing between
low and high performers to eliminate the ICT amplifier effect. Banker et al.
(1990) used DEA for assessing the impact of Electronic Point Of Sales
(EPOS) in a restaurant chain, but one of the contributions of this study is the
expansion of DEA at a macro level, i.e., across firms within the same sector.
The proposed DEA methodology overcomes all previous methodological
problems by also extending previous studies by using a stepwise DEA ap-
proach for constructing robust DEA productivity models.

RESTAURANT ICT AND PRODUCTIVITY

Electronic-point-of-sale-system (EPOS), devices used to take and manage

customers orders, represent the core catering ICT applications. EPOS per-
form such functions as guest check control, communication between servers
and the kitchen and sales data tabulation. An EPOS system is made up of a
number of terminals that typically interface to a remote central processing unit
and/or back office systems. Back-office systems provide the food cost analy-
sis, labor scheduling and financial and inventory controls required at the res-
taurant level. System interfaces accomplishes the basic objectives of
electronic data handling, reduces errors and saves time. EPOS can improve:
customer service, satisfaction and personalization, e.g., through interfaces
with customer databases; enable staff to be more productive; improve commu-
nications and control of activities among employees in food preparation and
service delivery; reduce and monitor costs; increase revenue per seat-hour
(Buergermeister, 2001; David et al., 1996).
E-purchasing/inventory control systems track the items on order, details of
suppliers, inventory on-hand and minimum levels for automating ordering.
Menu/recipe management/engineering software that creates files for each rec-
64 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

ipe and menu item permit the analysis of the impacts of changes of ingredient
costs, ingredient quantities and price changes. Food costs percentages can be
calculated with these systems that are used for pre-costing menus and events.
Other production support systems can also provide sales forecasting, produc-
tion planning, workforce scheduling. Table management systems (e.g., reser-
vation, floor-plan, waiting list management and table availability) are
designed to track table status to improve timeliness of services and speed
turns.
Although these restaurant functions are frequently served by separate ICT
applications, integration among these ICT systems and with other departmen-
tal ICT (marketing, financial databases) is crucially important for enhancing
operations efficiency and effectiveness. For example, interfaces can stream-
line the whole process by allowing perpetual inventory of food ingredients to
be kept in the following way. When the sale of an item is registered in the
EPOS, its component ingredients can be calculated and transmitted to the in-
ventory where the food inventory amount is subtracted from the quantity on
hand. In multi-unit restaurants, additional interfaces, enabling data sharing be-
tween units and from each unit to head office, allow everybody to benefit from
the others experience, for online consolidation of sales, financial reports and
centralized procurement. Systems integration issues and their business bene-
fits in the restaurant business have been identified and discussed in depth by
Ansel and Dyer (1999).
Overall, integration of ICT in restaurants operations aims to effectively
and efficiently manage and maximize exploitation of each of the four core res-
taurant resources while eliminating time bottlenecks in resources co-ordina-
tion. Table 1 summarizes numerous ICT applications in restaurant applications
by categorizing them into a two dimensional matrix. On the vertical axis, a res-
taurant process analysis is done and ICT applications are clustered depending
on the restaurant processes that they aim to optimize and rationalize. On the hor-
izontal axis, ICT applications are categorized based on the resource, namely,
customers (demand), employees, space/seats/physical capacity and menu/in-
ventory items that ICT aims to manage and co-ordinate.

RESEARCH AIMS AND METHODOLOGY

The study aimed to assess the ICT impact on restaurant productivity and so,
its methodological approach that is argued to overcome the previous identified
limitations is designed as follows. Primary data were gathered from restaurants
within the four star UK hotel sector. By concentrating on a specific sector, con-
textual factors and business operational characteristics that would have im-
 Marianna Sigala 65

TABLE 1. ICT Use for Restaurant Productivity (Management of Four Resources)

Productivity Employees Customers/ Space/seats/ Menu/

enhancement Demand physical Inventory
method capacity items

Reduce Forecasting Forecast Forecast table Forecast raw

uncertainty of required labor demand (68%) availability (23%) goods to order
arrivals/demand levels (54%) Overbooking Manage (48%)
(54%) reservations Inventory
Reduce (table control (69%)
no-shows (43%) configuration Just-in-time pro-
optimization) curement (32%)
(13%)

Shifting/ Improve labor Advanced Differential Differential pric-

managing scheduling reservations pricing for ing for menu
demand: (71%) (63%) floor/room items (82%)
Development and Duration sections (1%)
management of charges (0%) Differential
non-physical pricing for
and physical consuming
fences: space at
different times
(weekday/rush
hours, etc.)
(77%)

Differential / Develop guest Recipe database

personalization of history systems and nutritional
pricing and/or (9%) analysis (0%)
experience Frequent Cost accounting
customer and pricing for-
programs (0%) mulation (89%)

Reduce meal Improve staff Direct Track Menu engineer-

duration communication customer- consumption ing (67%)
uncertainty (75%) order-entry times Order by table
Improve systems (0%) (4%) and time (21%)
bussing (73%) Service status- Track food-prep-
Speed check zone conditions aration (15%)
delivery (69%) (10%) Track meal dura-
Server-station Table-status by tion by meal part
management meal part (3%) (3%)
(42%)

Reduce time Buzzer systems Buzzer systems Table

between (alert staff ) (customer management
customers (65%) paging) (0%) (9%)
Improve Estimate
communication waiting time
(77%) (1%)

e-business e-recruitment e-reservations Mutli-channel e-procurement

benefits (2%) (3%) distribution/ (6%)
webification of e-training (8%) e-customer promotion (13%)
business service (4%)
processes
66 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

TABLE 1 (continued)

Productivity Employees Customers/ Space/seats/ Menu/

enhancement Demand physical Inventory
method capacity items
Multi-unit Centralized IT reservations Multi-unit space Centralized
management training (23%) systems to scheduling (0%) procurement
Centralized direct customers (6%)
staff scheduling to other units (1%)
(5%) Share of customer
databases
histories (1%)

pacted on the ICT-productivity relationship are eliminated, while by obtaining

data from a large sample, findings can be generalized. The ICT productivity im-
pact is investigated at the organizational level, since this is regarded as the best
level of analysis. To overcome limitations relating to the quality of data, produc-
tivity measurement and the statistical methods relating inputs and outputs, the
following steps were undertaken. A total approach to productivity was adopted,
including both production efficiency and effectiveness. Thus, financial, objec-
tive and easily obtainable productivity inputs and outputs that encapsulate both
tangible and intangible ICT impacts were gathered. Demand variability was also
measured (1-9 scale, from low variability to high variability) to consider the
productivity impact of environmental factors. Multiple productivity inputs/out-
puts/factors were used for investigating the multi-dimensional ICT productivity
impact. The latter were considered simultaneously by using DEA (Frontier An-
alyst software package). By using DEA, bad and good performers were also dis-
tinguished and the impact of ICT on each of them separately was later assessed,
i.e., avoiding the ICT amplifier effect. However, because the reliability and va-
lidity of the DEA technique are as good as its inputs/outputs, the following
procedure was used for constructing robust DEA productivity models.
The stepwise approach to DEA was introduced by Sengupta (1988) as a use-
ful way for selecting appropriate inputs/outputs in DEA analysis. Stepwise
DEA is based on stepwise regression and is an iterative procedure in which pro-
ductivity is measured based on the important inputs/outputs identified up to that
step. Correlations between inputs/outputs and productivity metrics as well as
cause and effect judgements are examined for identifying inputs/outputs that
can significantly determine productivity. When productivity determinant in-
puts/outputs are identified, these are incorporated into DEA and the process is
repeated until no further significant correlations emerge. At that stage, a ro-
bust productivity metric accounting for all the identifiable inputs/outputs influ-
 Marianna Sigala 67

encing productivity is constructed. Moreover, by including aggregated inputs/

outputs at the first DEA step and by later disaggregating them into their con-
stituent parts (i.e. the partial productivity metrics that were found to signifi-
cantly affect overall productivity), one can also avoid the productivity
obscuring effects created by aggregate metrics. This application and value of
stepwise DEA for measuring productivity has been recently tested in the hos-
pitality context (Sigala, 2002; Sigala, 2004b; Sigala et al. 2004). Overall, by
using stepwise DEA for selecting appropriate productivity inputs/outputs, ro-
bust productivity metrics are constructed because the construction of the pro-
ductivity metric initially considers all potential productivity determinant
factors but it finally includes only those with a significant productivity effect.
Because of that, productivity differences among units can be attributed to fac-
tors that the stepwise DEA analysis has not so far considered.
As ICT can impact the use of all restaurant resources, data reflecting the
four major restaurant resources/inputs were gathered: number of seats and
banqueting covers (physical capacity); number of full time equivalent em-
ployees (FTEE) and payroll expenses for both full and part time staff (human
resources); material & other (M&O) expenses; management fees; ICT training
costs; demand variability (uncontrollable input); percentage of annual covers
from repeat customers. Data gathered regarding productivity outputs in-
cluded: number of restaurant and banqueting covers served; total revenue.
DEA did not include metrics such as profit since this could have lead to double
calculation of costs (Norman and Stoker, 1991). Since profit is the difference
between sales and costs and costs are already included in inputs, inclusion of
profits in the DEA model would have confounded the results. Moreover, ratio
metrics such as average check and sales per wait person are not included. Only
raw inputs/outputs were included in the DEA model in order to allow the free
investigation and manipulation of economies of scale through the DEA
process.
To overcome limitations relating to ICT measurement, the following analy-
sis was undertaken. Because it is the deployment of ICT tools and capabilities
that leads to productivity gains and not investments in ICT per se, the ICT con-
struct was operationalized by using three metrics: (1) number of ICT systems
(Table 2); (2) number of interfaces among ICT systems (ICT networking capa-
bilities); and (3) use of ICT and their generated information (as classified in
Table 1 (number of activities performed by ICT). Respondents were given a
figure illustrating all possible ICT systems and they were asked to circle those
that they used while draw lines between systems that interfaced. Indeed, sev-
eral authors have highlighted the impact of ICT to foster and support Business
Process Re-engineering (BPR) initiatives, while it has been proven (Willcocks
et al., 1998; Sigala, 2004a) that integration among ICT is a vital condition for
68 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

TABLE 2. CT Availability

ICT system % of respondents

Marketing systems 51%
HRM systems 23%
Finance & Accounting systems 79%
EPOS 68%
Table management systems 22%
Production support systems 54%
Inventory/procurement systems 61%

fostering BPR and enhanced productivity. Integration among ICT results in

great operational and process efficiencies, as it eliminates manual re-entry of
data, facilitates easy retrieval, sharing and search of consolidated databases,
which are also vital actions for informationalizing product/services and
streamlining processes. Sigala (2002) proposed a model for the exploitation of
ICT in the hospitality sector, whereby increased ICT integration and infor-
mate exploitation leads to higher ICT payoffs.
A structured questionnaire was developed and sent to a random sample
(compiled from the AA directory) of 400 four star UK hotels. To ensure consis-
tency, managers were asked to report inputs/outputs data referring to the finan-
cial year ending in 2001. After a follow up, 103 usable responses were gathered.

ANALYSIS OF THE FINDINGS

Respondents Profile

Respondents represent a diversified sample. Indeed, 34.1% and 25% were

independently owned and managed, respectively, with the remaining being
owned and managed respectively by a hotel chain. 48% of respondents were
located in city centres, fewer (22%) in rural and 30% in suburban places. Data
regarding size and nature of respondents operations also reveal a diversified
sample. So, respondents capacity varied from 30 to 300 seats and from 0 to
600 covers regarding banqueting capacity. Statistics regarding FTEE revealed
a similar pattern of size of operations, i.e., minimum FTEE for full time and
part time employees 8 and 5 respectively, maximum 28 to 59, respectively.
Repeat customers represented on average 23.1% of annual covers, while great
demand variability was also reported (average score 7.2). Table 2 provides
data regarding ICT availability, while Table 1 provides more information re-
 Marianna Sigala 69

garding the specific use and activities carried out with the available ICT sys-
tems. Regarding ICT integration, the ratio of the number of ICT interfaces to
the maximum number of potential ICT interfaces was calculated for each re-
spondent. This ratio rather than the raw number of ICT interfaces was used in
order to take into consideration that the number of ICT interfaces is dependent
on the number of available ICT. The average of this ratio among respondents
was 38%, indicating that respondents in general make little exploitation of the
ICT networking capabilities.

Productivity Results

Table 3 illustrates the stepwise DEA approach, whereby aggregated metrics

in the first step were disaggregated into productivity-significant determinant
factors in later steps. A robust DEA productivity metric is finally concluded at
step 4. Constant returns to scale were assumed but their validity was tested by
correlating DEA scores taken from all steps with a metric reflecting size of oper-
ation (i.e. number of seats). As no significant correlations (Pearson correlations,
= 0.05) were identified, the assumption of constant returns to scale was main-

TABLE 3. Stepwise DEA

Step 1 Step 2 Step 3 Step 4

(Input min) (Input min) (Input min) (Output max)
Outputs
FB revenue * * * *
Ratio of banqueting to restaurant covers * *
Inputs
FB capacity (banqueting & restaurant * * * *
seats)
Total payroll *
FTEE full time staff * * *
FTEE part time staff * * *
M&O expenses * * * *
Demand variability (uncontrollable *
input)
Other disaggregated Inputs/Outputs correlated with DEA scores
Inputs: FTEE and payroll expenses for full and part time staff; M&O expenses; management
fees; ICT training costs; % of annual covers from repeat customers.
Outputs: number of restaurant, banqueting and total covers served; % of: banqueting to total
covers served, restaurant to total covers served, banqueting to restaurant covers served
70 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

tained. Initially, DEA models assumed input minimization, meaning that res-
taurants aim to maintain at least the same level of outputs (be effective) while
minimising inputs (be efficient). However, because on step 4 an uncontrolla-
ble input (demand variability) was included, input minimization was not ap-
propriate (as management cannot control demand variations) and so, output
maximization was assumed. However, this did not affect the analysis across
steps as constant returns to scale were assumed and under constant returns to
scale input minimisation and output maximisation give the same DEA scores.
The DEA model at step 1 included revenue, FB capacity, total payroll and
M&O expenses (their inclusion was confirmed by an isotonicity test that re-
vealed positive intercorrelations among inputs and outputs). The DEA score
was calculated and correlated with the dissagregated inputs/outputs. Significant
negative correlations between FTEE referring to part time staff and DEA score
revealed that the efficient use of part time staff can significantly impact produc-
tivity. Because of that at step 2, total payroll was replaced by two variables:
FTEE for full time and FTEE for part time staff. After recalculating DEA scores
and conducting the correlations, a significant positive correlation was found be-
tween the ratio banqueting to restaurant covers and DEA scores. This is not sur-
prising since banqueting operations are usually more mass produced, delivered
and streamlined than restaurant covers, which in contrast are more time, mate-
rial and staff demanding processes. Thus, DEA scores were recalculated again
at step 3 which also incorporated this ratio. Correlations at step 3 between DEA
scores and dissagregated metrics revealed that demand variability significantly
affects productivity and so, the former was included in the DEA model at step 4.
DEA scores were recalculated and since no other significant correlations were
found at step 4, the DEA model at step 4 was concluded as a robust restaurant
productivity metric including all productivity determinant factors.
The stepwise DEA was also used for clustering hotels depending on their
type/reason for being productive (Table 4). Specifically, the DEA model in-
cluding business variability (step 4) reflects combined efficiency, i.e., ability
to be productive given the market conditions, while when business variability
is excluded (step 3), the DEA score reflects only operational efficiency. That
is because inefficient restaurants in step 3 that became efficient in step 4 attrib-
ute their efficiency to the fact that they can effectively manage demand vari-
ability (i.e., they are market efficient only), while inefficient hotels in both
step 3 and 4 are both operational and market inefficient.

ICT Productivity Impact

The ICT productivity impact was investigated by relating the three ICT
metrics with the different productivity DEA scores and types. As no signifi-
 Marianna Sigala 71

TABLE 4. MarketOperational Productivity Matrix (Number of Restaurants)

Efficient Market efficiency Combined efficiency

in step 4 only (22)
Market Efficiency

(7)

Inefficient Combined Operational

in step 4 inefficiency efficiency only
(71) (3)

Inefficient in step 3 Efficient in step 3

Operational Efficiency

cant Pearson correlations between DEA scores at step 3 (operational produc-

tivity) and 4 (combined productivity) and number of ICT systems (P = 0.231,
P = 0.173, = 0.05, respectively) were found, it was concluded that simple
ICT availability/investment does not affect productivity levels. ICT availabil-
ity did not also affect market productivity. This is because when conducting a
t-test (df = 101, t = 0.542), it was found that market efficient restaurants (i.e., 7 +
22 = 29) did not have a significantly different number of ICT systems from mar-
ket inefficient (71+3 = 74) restaurants. However, a significant Pearson correla-
tion (P = 0.682, = 0.05) between the ratio of available to potential interfaces
and the DEA scores at step 4 revealed that ICT integration does have a signifi-
cant impact on combined restaurant productivity. In contrast, exploitation of
ICT systems as well as of their generated information can significantly impact
on productivity as: (a) significant positive Pearson correlations were found be-
tween DEA productivity scores at step 3 and 4 and number of productivity im-
provement activities supported by ICT (P = 0.692, P = 0.713, = 0.05
respectively); market efficient restaurants reported to use ICT for a statisti-
cally significant greater number of productivity improvement activities (t =
0.002, = 0.05). ICT activities for multi-unit management were not included
in this analysis because 25% of respondents were not part of a chain.
DEA can also identify and calculate the level of inefficiency for each unit.
To do this, DEA calculates what the performance of each unit for every of its
input/output would have been if the unit had been as efficient as the other units
of its reference set and compares it with the actual input/output performance.
In this way, the inefficiencies of the four inputs for every restaurant at step 4
(combined productivity) were calculated and then correlated with the number
of productivity improvement activities (Table 5). Findings revealed that ICT
supported activities did not have a significant productivity effect on the man-
agement of single resources apart from material/inventory management. Thus,
72 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

it can be concluded that the ICT productivity benefit is maximised when ICT
exploitation is not focused on the management of a single resource but rather
on the efficient co-ordination, combination and management of all resources.
In other words, ICT should aim at maximising processes (combination of re-
sources) rather than resources efficiencies.
This finding is not surprising, as no restaurant activity is an isolated task,
but rather, its efficiency and management should be considered and managed
within its relevant business process. Thus, ICT in restaurants should aim to
have a multi-dimensional and synergic impact amongst restaurant operations,
activities and processes. The networking and communicating tools and capa-
bilities of ICT applications are ideal for enabling process and task integration.
The benefits of integrating restaurant ICT applications amongst each other
and the value added activities that a restaurant can develop are discussed in de-
tail by Ansel and Dyer (1999), while Sigala (2002) also provided evidence of
the ICT integration impact on the synergetic benefits between rooms and res-
taurant division in hotel properties. ICT integrating and informalization capa-
bilities are also important for developing a customer-centric restaurant unit.
To achieve that, ICT systems and integration will provide the digital nervous
system enabling communication among customer, information and material
restaurants processes in such a way that very operation and task will be able to
become personalized and customized depending on the needs, wants and pat-
terns of loyal customers. Moreover, information gathered from loyal custom-
ers can be cloned by restaurant operators for targeting other similar potential
TABLE 5. ICT Productivity Impact per Resource and Type of ICT Supported
Activity (A = 0.05)

Number of ICT Employees Customers/ Space/seats/ Menu/

supported (1) Demand physical Inventory items
productivity (2) capacity (4)
activities (3)
Percentages of
inefficiencies
(1) (P = 0.201)

(2) (P = 0.103)

(3) (P = 0.382)

(4) (P = 0.601)
 Marianna Sigala 73

market segments. Overall, in the knowledge era, the introduction of ICT in

restaurant operations should not only be concerned with maximizing effi-
ciency of isolated activities but it should be coupled and integrated with a cus-
tomer-centric process design and re-organisation analysis.

Conclusions and Suggestions

Despite the increasing ICT investments, the productivity impact of ICT has
been elusive. After reviewing the literature on the ICT productivity paradox
and on productivity measurement, the study proposed and empirically tested a
methodology for assessing the ICT productivity paradox that overcomes pre-
vious identified methodological shortcomings. The methodology was tested
in a dataset of three star hotels in the UK. Thus, findings measuring productiv-
ity and investigating its determinant factors are valid within a specific context.
Future research could investigate whether the same conclusions can be repli-
cated and generalized in different types of restaurant and/or countries. Given
the great product differentiation, operational, environmental and clientele di-
versity of the global restaurant industry, the application of DEA across restau-
rant types and countries can produce interesting results with crucial academic
and managerial implications.
However, findings regarding the ICTproductivity relation revealed that
the ICT productivity impact becomes apparent only when the exploitation of
the ICT networking/integration and informational capabilities are considered.
For optimizing ICT business value, restaurants should adopt a more strategic
approach to ICT implementation and management. Strategic ICT investment
decisions would translate in the full exploitation of three ICT capabilities
namely information, systems integration and architecture with the aim to
streamline, optimize efficiency and effectiveness and informalize processes
rather than individual resources. Within the experience and knowledge era,
restaurants, as all other types of businesses, are advised at creating a cus-
tomer-centric ICT and organizational infrastructure. ICT applications and
their communication and interactive capabilities are considered as the most
appropriate tools for creating an adaptable and responsive organization to cus-
tomers needs and patterns. However, the way with which ICT are being ex-
ploited should be aligned with business strategy and operations. For example,
although fast food outlets might use customer databases for predicting de-
mand and informating their product development practices, fine dining restau-
rants will be using customer databases for better personalizing customer
experiences. In this vein, future research should investigate how businesses in
different sectors can best deploy ICT and align them with corporate strategies.
74 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

Moreover, by conducting cross-sectors studies, one can also further enhance,

refine and test the validity of this studys methodology.
Notwithstanding the previous discussion, this study has one more limita-
tions that needs to be acknowledged, but which also points towards future re-
search avenues. The study adopted a total approach to productivity and used
financial productivity inputs/outputs for encapsulating both tangible and in-
tangible elements. Irrespective of the strength of this argument, such an ap-
proach did not allow the distinction of tangible (efficiency) and intangible
(effectiveness) productivity issues (e.g., customer satisfaction, service qual-
ity) and so, the detailed investigation of potential ICT productivity impacts.
Future research could actually try to develop better metrics for such qualita-
tive dimensions and then apply DEA for investigating any potential soft im-
pacts. Indeed, because DEA can deal with qualitative data it offers a great
potential for redefining service productivity and solving some of the problems
of its measurement. However, when soft data are used, issues of instrument re-
liability and validity become extremely important and so, DEA would need to
be combined with other research approaches and methodologies.

REFERENCES

Andersson T.D. (1996). Techniques for measuring productivity and efficiency in res-
taurants. In Johns, N. (ed), Managing productivity in hospitality and tourism,
Cassell, London, p. 209-226
Ansel, D. and Dyer, C. (1999). A framework for restaurant information technology.
Cornell Hotel and Restaurant Administration Quarterly, June, 7484.
Avkiran, N.K. (1999). Productivity Analysis in the services sector with Data Envelop-
ment Analysis. Camira, Queensland
Ball, S. (1993). Productivity and productivity management within fast-food chainsa
case study of Wimpy International, MPhil dissertation, University of Huddersfield,
UK.
Banker, R. D. and Morey, R. C. (1986). Efficiency analysis for exogenously fixed in-
puts and outputs. Operations Research, 34, 513-521.
Bender, D. (1986). Financial impact of information processing. Journal of Manage-
ment Information Systems, 3 (2), 232-238.
Brown, J.R., and Dev, C. S. (1999). Looking beyond RevPAR; productivity conse-
quences of hotel strategies. Cornell Hotel and Restaurant Administration Quar-
terly, April: 23-33
Brynjolfsson, E. (1993). The Productivity Paradox of Information Technology. Com-
munications of the ACM 36 (12), 67-77
Buergermeister, J. (2000). Point-of-sale technology for foodservice: underutilized
power tool or disappointing over achiever. International Journal of Hospitality In-
formation Technology, 1 (2), 9-20.
 Marianna Sigala 75

Byrd, T. and Marshall T. (1997). Relating IT investment to organizational perfor-

mance: A causal model analysis. International Journal of Management Science. 25
(1), 43-56.
Charnes, A.C., Cooper, W.W., and Rhodes, E. (1978). Measuring the efficiency of de-
cision making units. European Journal of Operational Research 2(6): 429-444.
Charnes, A.C., Cooper, W.W., Lewin, A.Y., and Seiford, L.M. (1994). Data Envelop-
ment Analysis: Theory, Methodology and Applications. Boston, MA: Kluwer Aca-
demic Press.
Cron, W. and Sobol, M. (1983), The relationship between computerization and perfor-
mance. Information Management, 6 (1), 171-181.
Dasgupta, S., Sakris, J. and Talluri, S. (1999), Influence of information technology in-
vestment on firm productivity: a cross-sectional study. Logistics Information Man-
agement, 12 (1/2), 120-129.
David, J., Grabski, S. and Kasavana, M. (1996), The productivity paradox of hotel-in-
dustry technology, Cornell Hotel and Restaurant Administration Quarterly, 37(2),
64-70.
Dos Santos, B.L., Peffers G.K., and Mauer D.C. (1993) The impact of Information
Technology investment announcements on the market value of the firm. Informa-
tion Systems Research 4 (1): 1-23.
Gummesson, E. (1998). Productivity, quality, relationship marketing in service organi-
zations. International Journal of Contemporary Hospitality Management, 10 (1),
4-15.
Heap, J.P. (1992). Productivity Management: a fresh approach. London: Cassell.
Hitt, L., Brynjolfsson E. (1996). Productivity, profit and consumer welfare: three dif-
ferent measures of information technology value. MIS Quarterly, 20(2), 121-142.
Johns, N. & Wheeler, K. (1991). Productivity and Performance measurement and mon-
itoring. In Teare, R. (1991) Strategic Hospitality Management, London, Cassell, pp.
45-71.
Johns, N., Howcroft, B., and Drake, L. (1997). The use of Data Envelopment Analysis
to monitor hotel productivity. Progress in Tourism and Hospitality Research 3(2),
119-127.
Jurison, J. (1996). Towards more effective management of information technology
benefits. Journal of Strategic Information Systems, 5, 263-274.
Lucas, H.C. (1993). The business value of Information Technology: a historical per-
spective and thoughts for future research. In Strategic Information Technology
Management (Banker, R., Kauffman, R. and Mahmood, M.A. Eds), p. 102-123,
Idea Publishing, Harrisburg, PA.
Menon, N.M. (2000). The Impact Of Information Technology: Evidence From The
Healthcare Industry. Garland Publishing, New York and London
Norman, M. and Stoker, B. (1991). DEA The Assessment Of Performance. Wiley, Lon-
don.
Paradi, J.C., Reese, D.N. and Rosen, D. (1997) Applications of DEA to measure the ef-
ficiency of software production at two large Canadian banks. Annals of Operations
Research, 73, 91-115.
76 JOURNAL OF FOODSERVICE BUSINESS RESEARCH

Rimmington, M., & Clark G. (1996). Productivity measurement in food service sys-
tems. In N. Johns (Ed), Productivity Management In hospitality And Tourism (pp.
194208). London: Cassell.
Sengupta, J.K. (1988). A robust approach to the measurement of Farrell efficiency. Ap-
plied Economics, 20, 273-283.
Shafer, S. and Byrd (2000). A framework for measuring the efficiency of organiza-
tional investments in information technology using data envelopment analysis.
Omega, The International Journal of Management Science, 28, 125-141.
Sigala (2004a, in press). The Information and Communication Technologies produc-
tivity impact in the UK hotel sector. International Journal of Operations & Produc-
tion Management.
Sigala, M. (2004b). Using Data Envelopment Analysis for measuring and benchmark-
ing productivity in the hotel sector. Journal of Travel and Tourism Marketing,
16(2/3).
Sigala, M. (2002) Assessing the impact of Information and Communication Technolo-
gies (ICT) on productivity in the hotel sector: an operations management approach.
PhD Thesis, University of Surrey, UK.
Sigala, M., Jones, P., Lockwood, A. and Airey, D. (2004, in press). Productivity in ho-
tels: a stepwise Data Envelopment Analysis of hotels rooms division processes.
Services Industries Journal.
Sigala, M., Airey, D., Jones, P. and Lockwood, A. (2001). Investigating the effect of
multimedia technologies on employment patterns in UK tourism and hospitality in-
dustry. In Information & Communication Technologies in Tourism 2001 (Sheldon,
P., Wober, K. and Fesenmaier, D. Eds), p. 205-215, Springer, Vien.
Strassmann, P. (1990). The Business Value Of Computers: An Executives Guide. The
Information Economic Press, New Canaan, Connecticut.
Tarim, S., Dener, H.I., and Tarim, S.A. (2000). Efficiency measurement in the hotel in-
dustry: output factor constrained DEA application. Anatolia: An International
Journal of Tourism and Hospitality Research, 11(2): 111-123.
Weill, P. (1992). The relationship between investment in IT and firm performance: a
study of the valve manufacturing sector. Information Systems Research, 3 (4),
307-333.
Werthner, H. and Klein, S. (1999). Information Technology and Tourism; a challeng-
ing relationship. Springer Verlag, Vein.
Witt, C.A., and Witt, S.F. (1989). Why productivity in the hotel sector is low. Interna-
tional Journal of Contemporary Hospitality Management, 1 (2): 28-33.
Wber, K.W. (2002). Benchmarking in Tourism and Hospitality Industries: The Selec-
tion of Benchmarking Partners. Wallingford: CAB International.