Journal of the Operational Research Society (2013) 64, 1194–1204 © 2013 Operational Research Society Ltd.

iety Ltd. All rights reserved. 0160-5682/13

www.palgrave-journals.com/jors/

Demand uncertainty in construction supply

chains: a discrete event simulation study
C Vidalakis  , JE Tookey and J Sommerville
1 2 3
1 2
University of the West of England, Bristol, UK; Auckland University of Technology, Auckland,
3
New Zealand; and Glasgow Caledonian University, Glasgow, UK
The delivery of construction projects is typically an assembly operation involving a high number of
subassemblies and materials brought on site by the supply chain. However, although supply chain
management in construction has attracted significant attention, paradoxically little focus has been
placed on construction supply networks and operations. This paper places emphasis on supply chain
operations by looking at the logistics function of construction material suppliers. Specifically, the paper
examines the impact of demand uncertainty on supply chain performance in order to assess the capacity
of material distribution companies to provide a timely and cost-efficient service to the construction
industry. The study adopts a discrete event simulation approach to assess the impact of demand
fluctuations on two crucial logistics performance measures; lead time and cost efficiency. The results
show that lead times are particularly sensitive to fluctuations under conditions of low demand. The
findings also reveal that, although transport is a significant cost element for lower demand levels, higher
inventory costs result in a negative exponential relationship between increasing demand and cost efficiency.
Journal of the Operational Research Society (2013) 64, 1194–1204. doi:10.1057/jors.2012.156
Published online 20 March 2013

Keywords: demand uncertainty; construction; logistics; distribution; simulation

Introduction on the upstream supply chain highlighting the need for

focusing on supply chain operations. However, this need
Supply Chain Management (SCM) was introduced in the has yet to be addressed. The function of material suppliers
1990s to describe the evolution of logistics from a and supply chain operations in construction remains a
corporate viewpoint to a holistic and strategic approach largely overlooked area of research while the fundamental
to materials and logistics management (Cooper et al, 1997; implicit assumption of logistics management expertise
Tan et al, 2002). SCM in construction has attracted inherent within SCM literature is basically ignored
significant attention particularly in the last two decades (Vrijhoef and Koskela, 2000; London, 2008; Vidalakis
and now stands out as a mainstream field of construction et al, 2011a). The literature reveals that inefficiencies in the
management research (Azambuja and O’Brien, 2008). distribution of construction materials inevitably result in
Adopting a strategic perspective, the industry has used the disruption of construction process and the increase of
SCM to address endemic industry characteristics related to construction cost. Indeed, unavailability of materials and
the highly fragmented nature of the sector. Thus, SCM unreliable deliveries has been highlighted as one of the
implementation in construction has mainly supported main causes of delays in construction projects resulting in
collaborative procurement systems, long-term relationships additional costs due to express deliveries (Tommelein,
and partnering (Akintoye et al, 2000; Saad et al, 2002). 1998; Assaf and Al-Hejji, 2006). Thus, Kagioglou et al
From an operational viewpoint, the adoption of SCM in (2001) underline that construction efficiency can be cor-
construction has been largely based on the concepts of lean related to the same extent with both higher productivity on
and agile construction. Although many differences and site and reduced suppliers’ costs. Furthermore, the
similarities can be identified between lean and agile, both Strategic Forum for Construction (SFC, 2005) suggests
systems suggest the operational dependence of constructors that potential savings from improving logistics could be
substantial and range between 10 and 30% saving in

Correspondence: C Vidalakis, Department of Construction and Property,
construction costs, largely due to reduced costs associated
University of the West of England, Coldharbour Lane, Bristol, BS16 1QY, with the transportation and handling of materials. This
UK. suggestion is further supported by research undertaken by
 C Vidalakis et al—Demand uncertainty in construction supply chains 1195

Building Research Establishment (BRE), which considers facilitate better decision making. Indeed, the literature
transport of construction materials to account for 10–20% reveals that previous studies in construction SCM have
of total construction costs (BRE, 2003). utilised various types of optimisation models, including
This paper focuses on supply chain operations by deterministic analytical (for example Ganapathy et al,
looking at the logistics function of construction material 2003; Georgy and Basily, 2008), stochastic analytical (for
suppliers. Given that demand uncertainty is one of the example, Caron et al, 1998; Ng et al, 2008), economic
most important factors affecting supply chain operations (for example, London, 2004; Sobotka and Czarnigowska,
(Fildes and Kingsman, 2011; Taylor, 2011), the paper aims 2005) and simulation models (for example; Tommelein and
to explore the effects of demand uncertainty on logistics Weissenberger, 1999; Vidalakis et al, 2011b).
responsiveness and cost efficiency. The next section of the The main criterion for the selection of an appropriate
paper discusses how demand uncertainty in construction modelling approach to be adopted in this paper was
can add significant complexity to supply chain operations that this should not be an obstacle in testing alternative
and how modelling can be used in order to address this. demand scenarios and providing an understanding of
The rest of the paper describes a simulation study on the supply chain logistics operations. This objective indicated
logistics of two material distribution companies placing that the use of both descriptive and optimisation modelling
particular emphasis on model development and validation, approaches could be useful. Indeed, a descriptive model
data collection, distribution fitting and experimentation can be extended to an optimisation model through
using simulation. The results are discussed in relation to sensitivity analysis, a process perfectly accommodated by
practices that could potentially improve supply chain a simulation model (Greasley, 2004; Kleijnen, 2005).
operational efficiency in terms of lead time and cost Thus, it was decided that simulation modelling would be
effectiveness. the most appropriate approach to assess the effect of
demand uncertainty on logistics performance. Specifically,
in accordance with the problem discussed in this paper, the
use of simulation would help to:
Demand uncertainty and supply chain modelling
Demand uncertainty is a crucial determinant of supply K describe logistical processes and allow exploration of
chain performance with resultant inefficiencies including specific performance areas, such as time and cost
insufficient or excessive inventories, poor product fore- efficiency;
casts, uncertain lead times and poor production planning K incorporate structural uncertainties in the design of
(Lee et al, 1997; Chaharsooghi and Heydari, 2010). In logistics networks such as the location of supplier depots
construction demand uncertainty is often seen as one of the and construction sites;
key features of the industry (Naim and Barlow, 2003; K experiment with different levels of demand, supplier
Green et al, 2005; Ala-Risku and Kärkkäinen, 2006). This capacity and product characteristics;
uncertainty, which ultimately translates to uncertainty in K investigate operational changes and provide a risk-free
the demand of construction materials, emanates primarily environment for applying alternative operational
from the unique nature of construction projects, wide strategies.
range of material requirements, uncertain location of
construction sites and workload seasonality (Arbulu et al, Simulation modelling was facilitated by Simul8, a
2003; Caniato et al, 2011; Vidalakis et al, 2011b). These custom-tailored software, which was used both for
industry and project features can hinder suppliers’ descriptive and analytical purposes. The essential elements
structural decisions in terms of warehouse location, of the study are discussed in the following section.
number and capacity of facilities, trading products and
transportation mode (Silver et al, 1998). At the same time,
they can also impose significant constraints on operational Simulation
aspects such as inventory replenishment planning, delivery
Objectives and background of the study
scheduling, load consolidation and backhauling and thus
reduce significantly transport efficiency (Shakantu et al, The objective of simulation in this study is to measure the
2008). effects of varying demand of construction materials on two
From the above discussion it could be concluded that crucial logistics performance measures that is, lead time
demand uncertainty in construction can add significant and cost efficiency. The study was carried out in
difficulty in the analysis of supply chain operations. Due to cooperation with two material distribution companies; a
the increased complexity involved, a simplified description builders’ merchant (BM) and a construction materials
of supply chains would be particularly useful. This has supplier (MS), both based in the outskirts of Glasgow,
been accommodated by modelling and particularly opti- UK. The BM has two 2-axle-rigid body vehicles with a
misation approaches, which can enable forecasting and 17-tonne gross vehicle weight while the MS has four
1196 Journal of the Operational Research Society Vol. 64, No. 8

2-axle-rigid body vehicles with a 17-tonne gross revenue volumetric quantities of construction materials. This
weight and one similar vehicle with a trailer, which has a approach was also useful in order to reduce the total
combined capacity of 29 tonnes. Both companies employ number of variables and thus, make experimentation with
less than 50 staff and serve mainly urban and suburban simulation feasible. However, it is acknowledged that
customers. According to the Department for Transport different types of material can have significant impact on
(DfT, 2006) both businesses represent typical material specific logistical parameters such as handling requirements
distribution companies where journey planning is based and inventory costs, the latter to be addressed later in
very much on the personal knowledge of staff as to local the paper.
routes and location of customers.
Operational model building
Model conceptualisation The computerised version of the model, as designed in
Simul8, is presented in Figure 2. Similarly to the
Generally, the aim of a conceptual simulation model is to
provide a description of the system to be modelled conceptual model, the system comprises of active and
(Robinson, 2004). Thus, the conceptual model for the passive states represented by two types of simulation
needs of this study, illustrated in Figure 1, was based on a objects; work centres ( ) and storage bins ( ), which
process mapping exercise undertaken together with are dynamic decision agents and queues, respectively.
depot and transport managers, warehouse personnel Table 1 describes the function of each simulation object
illustrated in Figure 2 and links the operational to the
and vehicle drivers. Figure 1 can be divided in the MS-
conceptual representation of the simulation model.
BM, on the left, and the BM-contractor activity cycles on
Notionally, Figure 2 is divided horizontally in two parts.
the right part of the figure. The cycles consist of active and
passive stages, represented by rectangles and circles, The upper part represents the MS’s model while the lower
respectively. The product is introduced into the system in the MS’s system. The timing data TWCB, C, 4–8 for work
the ‘Supplier’s inventory’ node at the MS-BM activity centres WCB, C and WC4-8 was calculated as follows:
cycle. Then, it is forwarded to the BM-contractor cycle TWCB;C;48 ¼ ½ðNumber of deliveries per vehicle movement þ 1Þ
and delivered to the ‘Site inventory’. The model would  Distance travelled between stops=Travel speed
be applicable for single- or multiple-product types by þ Number of deliveries per vehicle movement
replicating inventories as many times as the number of
 Turnaround time
different products. However, in this study the system has
been envisaged as a generic logistics model that does not Timing information for the remaining work centres
account for specific product properties but rather for was set by the use of Visual Logics, Simul8’s internal

Figure 1 Conceptual model of intermediaries’ logistics system.

 C Vidalakis et al—Demand uncertainty in construction supply chains 1197

Figure 2 Operational model in Simul8.

Table 1 Function of Simul8 objects SBD-H and SB9–17 defining the number of deliveries per
vehicle movement in the timing data equation presented
Simulation object Main functions
above. Thus, the number of storage bins linked to each of
WEntr K Decide the use of BM or MS model these work centres indicates the number of interim stops
K Define demand distribution along a vehicle route.
WCA, and WC2, 3 K Forward loads to vehicles
Simulation assumptions
WCB,C and WC4–8 K Transport loads to sites
K Define time for vehicle being away The main simulation assumptions were related to lead
from the warehouse (includes time, demand, system costs and simulation properties.
travelling and turnaround times)

WC9D K Move delivered loads to results

collection point Lead time. Activities to be considered in the calculation
of lead time include order processing, load preparation,
SBA and SB1 K Receive orders loading/unloading, travelling and interim drops. How-
K Define time for orders process, ever, the time required for the first four activities was
preparation and loading
considered to be deterministic. Also, to remove variation
SB2, 3 K Hold load ready for dispatch in drivers’ driving attitude and externalities such as traffic,
weather conditions and speed limits, travelling speed
SBB,C and SB4–8 K Hold load before loading was also considered to be deterministic. Therefore, these
variables were included in the simulation model as
SBD-H and SB9–17 K Receive deliveries
K Define the number of deliveries per constants, but were not allowed to vary, in order to focus
vehicle movement on the subject under consideration, that is demand
variation. However, it has to be noted that although
WComp K Collect loads travel speed was assumed constant, transportation times
did vary since driving distances were considered to be
stochastic. Finally, unavailability of materials has not
been addressed by the developed model in order to avoid
programming language, which allows for building logic assumptions related to inventory decisions depending on
according to the system requirements and input parameters the costs of carrying inventory, attitudes towards the risk
defined later in Table 2. of carrying inventory and replenishment cycles.
It should be noted that the storage bins SBD-H and
SB9–17 indicated in Figure 2 do not represent individual
construction sites but rather serve to define timing data for Demand. Demand was set equal to the quantity of
work centres WCB, C and WC4–8. This is achieved by materials actually delivered within the data collection
1198 Journal of the Operational Research Society Vol. 64, No. 8

Table 2 Data distributions—model input parameters

Input parameters BM MS

Mean Distribution Mean Distribution

Number of deliveries (per movement) 1.96 – Binomial (5., 0.393) 3.97 – Negative binomial (15., 0.791)
(Median: 2) (Median: 4)
Vehicle Initial Efficiency (%) 24.1 Pearson 6 (0., 0.515, 2.38, 6.07) 39.4 Erlang (0., 2., 0.197)
Order size (m3) 2.6 Weibull (0., 0.929, 4.74) 4.2 Empirical (mean 4.2, Std 6.7)
Turnaround times (m) 23.7 Gamma (0., 3.14, 7.58) 18.4 Empirical (mean 18.4, Std 11.4)
Distance travelled (per journey leg) (miles) 36.2 Weibull (0., 2.42, 40.6) 192.3 Gamma (0., 1.7, 113)
Order process time (min) n/a 180 n/a 180
Load preparation and loading time (m) n/a 180 n/a 180
Travel speed (miles/h) n/a 30 n/a 30

period, which was 24 days. To allow for varying vehicle suggested by El-Haik and Al-Aomar (2006). The accep-
efficiency, demand was approximated by order frequency table confidence level was set to 95% ensuring that
and was 1 order per 79.5 min for the BM and 1 order per deviation from the mean did not exceed the mean value by
25.6 min for the MS. These demand figures had virtually 10%. This was achieved by allowing for 30 runs.
no effect on the simulation output since demand was the
manipulated parameter; however, the aforementioned
Verification and validation
values have been used as base demand figures (ie 100%)
and therefore, any percentage changes refer to these The conceptual model was developed in collaboration with
values. In terms of demand distribution approximation, experts who were highly familiar with the real-world system
any parameterised statistical distribution would be and therefore, it is assumed that conceptual validity is high.
particularly unreliable, especially when considering the However, to ensure that the conceptual model was
short duration of the data collection period and thus, a transformed into a computer model with sufficient accuracy
normal distribution with standard deviation set to 25% of operational verification was necessary. This was achieved in
the mean value was selected (Hauge and Paige, 2001). two stages. Initially, a walkthrough procedure was followed
in order to calibrate the intended function of each element of
the model. Then, a white box validation, a process for
System costs. Logistics costs incorporated in the analysis determining that all constituent parts of the model represent
were limited to transport and inventory costs as well as real-world elements with sufficient accuracy (Robinson,
time-related transportation costs (ie driver employment 2004), was carried out in cooperation with the transport
costs, depreciation, licences, insurances, interest on managers. Furthermore, to ensure Visual Logics defined
capital and overhead per vehicle). All other logistics costs algorithms were appropriate, the developed code was also
have been omitted as being either negligible compared to reviewed by a modelling expert with extensive experience on
transport/inventory costs or in order to limit assumptions Simul8. Finally, to consider the overall behaviour of the
and focus on areas of interest. For calculation purposes model and its capacity to produce accurate results a black
figures for both fixed and variable costs were obtained box validation (Robinson, 2004) was undertaken using
from the Road Haulage Association cost tables (RHA, hypothetical but real-world data. The results of the black
2005), derived from annual surveys of RHA members. box validation were reviewed by the transport managers
Inventory carrying costs were calculated based on the who confirmed the validity of the output.
inventory holding time estimated with great accuracy by
Simul8. Calculations were performed for a range of
Data collection and input data modelling
material values and inventory holding cost was estimated
as 20% of the average annual inventory value, a figure The aim of the data collection process was to record
suggested to be the norm when an exact calculation is not material and information flows and entailed indirect
possible (Ganeshan, 1999; Bowersox et al, 2007). observations of vehicle movements between material
suppliers’ depots, BMs’ facilities and construction sites.
For the needs of the study, a data recording system was
Simulation properties. The developed simulation system developed capturing information related to the following
is non-terminating and thus, terminating time was set to parameters:
coincide with the end of the data collection period. The
number of simulation runs to be performed was set so as K Vehicles: Number, type, capacity
to provide results with an acceptable confidence level as K Materials: Volume
 C Vidalakis et al—Demand uncertainty in construction supply chains 1199

K Vehicle movements: Nature, number, departure time, The figures indicate similar shaped lead time curves for
manifest loading both companies. At the outset, lead times are high and
K Deliveries/collections: Number, destination, travel dis- decrease as demand increases. Indeed, the dispatch criterion,
tance, turnaround times set according to Vehicle Initial Efficiency as indicated in
Table 2, will be satisfied sooner as incoming orders become
The data recording system was based on information more frequent. As demand continues to rise, lead times go
available in vehicles’ registration certificates, delivery notes, up displaying an exponential relationship to demand. In
invoices and tachographs. Furthermore, vehicle drivers particular, lead times rise by 2.6 and 2.8 times, respectively,
were asked to capture arrival and turnaround times at for the BM and the MS when demand increases from 100 to
various delivery/collection locations. This information was 200%. Thereafter, lead times continue to increase in a
recorded on individual invoices and delivery notes facil- smoother and steadier fashion. The most interesting findings
itating the establishment of vehicle routes and the from Figures 3 and 4 can be stated as follows:
estimation of travel times and distances between stops
along the routes. The empirical data captured included K Low levels of demand can result in longer lead times.
information associated with 672 deliveries and collections K Demand change from 100 to 200% results in significant
transported 26 655 miles by seven vehicles. lead time increase.
To model randomness, distribution fitting was per- K Demand changes over 200% result in moderate lead
formed with Stat:fit, which uses mainly Kolmogorov– time increase.
Smirnov and Anderson–Darling goodness-of-fit tests to
assess the appropriateness of theoretical distributions to fit Significant increase of lead time indicates a minimum 3
the empirical data. Table 2 shows the mean values and days increase, while moderate increase implies lead time
distributions of the input variables used for configuring the delays of no more than 1 day. Lead times can increase by
model. The variables that were assumed to be deterministic up to 2 days when demand is significantly low. Thus, trying
are demonstrated in Table 2 as constants. When empirical to maintain or increase vehicle efficiency during periods of
data could not be fit efficiently by a theoretical distribution low demand would lead to increased lead times. Similarly,
an empirical distribution was used as recommended by timely delivery of materials would be hindered significantly
Law and Kelton (2000). by small demand increases, indicating that both companies
operate at maximum capacity when demand equals base
demand. However, this is not crucial for the MS who
implements a weekly delivery plan but is particularly
Simulation results and discussion critical for the BM who operates on the verge of the 2 days
Demand effect on lead time delivery promise. Therefore, to avoid a high lead time
increase during periods of low or high demand companies
The simulation model was run for 36 different demand should adjust the dispatch criterion by lowering or
levels varying from 10 to 1000% of the base demand. increasing vehicle efficiency, respectively, according to a
Therefore, 36 Simul8 trials were performed incorporating preferable (not necessarily minimum) lead time.
30 runs each. The effect of varying demand on lead time is
presented in Figures 3 and 4 for the BM and the MS, Demand effect on system costs
respectively. The results are presented as point estimates,
that is, the mean of the confident intervals boundaries at a System costs for four different annual inventory values are
95% confidence level. illustrated in Figures 5 and 6 for the BM and the MS,

Figure 3 Demand effect on lead times (BM). Figure 4 Point estimates of lead time (MS).
1200 Journal of the Operational Research Society Vol. 64, No. 8

Figure 5 Estimates of system costs and profits (BM).

respectively. The figures also indicate transportation costs vehicles with direct negative effects on lead time and
and sales. It has to be noted that profit has been calculated inventory holding costs. This point reveals the significance
without allowing for potential profit margins. Also, to of the number of vehicles used by a company and is related
highlight the trade-off between cost elements and profits, to the significant capital involved. Although under higher
the graphs display results for demand varying between 10 demand levels inventory costs can be a major element of
and 200% of base demand only. system costs, this is not the case for lower levels of demand.
For both companies system costs and profits show the The more the annual inventory value decreases the more
same trend as demand levels increase. For zero demand, significant transportation costs become. This finding would
system costs are equal to time-related transportation costs have crucial implications for smaller companies, which
since it was assumed that the inventory held by the two need to generate a significant amount of sales in order to
companies was the minimum required so as to avoid cover transport costs. Larger companies, indicated poten-
unavailability of materials. As a result, inventory levels and tially by higher annual inventory, can cover system costs
value are also zero. When demand increases transportation faster; with their operations, however, being at higher risk
and inventory costs also increase causing the profit curve to when demand increases.
fall. Two important points can be noted from the costs and Figures 7 and 8 illustrate how varying demand affects
profits curves illustrated in Figures 5 and 6: system profits according to different inventory values. In
general, maximum profit is achieved when the annual value
K Under high levels of demand system costs rise sharply of inventory is high but demand is low. Interestingly, both
largely due to inventory costs. graphs incorporate a demand level where the profit figure is
K There is a negative exponential relationship between the same notwithstanding the annual inventory value. This
demand and profits. figure is equal to the loss incurred from time-related
transportation expenses when demand is zero and occurs at
Indeed, inventory costs rise sharply when increasing 134 and 137% of base demand levels for the BM and the
demand cannot be satisfied by the number of existing MS, respectively.
 C Vidalakis et al—Demand uncertainty in construction supply chains 1201

Figure 6 Estimates of system costs and profits (MS).

The different profit curves indicate that up to a critical However, these suggestions have to be further consid-
demand level (CDL), the higher the annual inventory value ered under different demand conditions and in accordance
and the lower the demand, the greater the profit made. After with the supplier’s delivery promise, a major determinant
this point, the higher the annual inventory and the higher of customer satisfaction (Ogden and Turner, 1996). As
the demand, the lower the profit made (or the greater the previously discussed, during periods of low demand
loss). In the figures this is illustrated by the profit curves that increasing efficiency would have a negative impact on lead
move anticlockwise until they become horizontal and equal times. Increasing capacity would certainly facilitate deliv-
to the cost for merely maintaining the fleet. Notwithstanding ery to promise but would also require suppliers to bear
the value of materials traded, this cost occurs always at the cost of unutilised capacity. However, this could be
the CDL. Therefore, a company operating under demand avoided if suppliers made any excess capacity available to
beyond the CDL will generate additional loss (further than other suppliers or trading partners in the supply chain.
vehicle maintenance cost) unless it: Thus, suppliers would operate essentially as part of a
collaborative transportation network (CTN), an emerging
K increases the CDL: This implies that the company will concept in the transportation industry aiming to improve
need to explore efficiency improvements either by distribution flexibility and minimise the SCM inefficiencies
increasing vehicle efficiency or purchasing additional (Esper and Williams, 2003; Chan and Zhang, 2011). To
vehicles (increasing capacity). ensure delivery to promise during high demand periods
K outsources demand beyond the CDL: This can be suppliers could utilise services offered by 3PL providers
facilitated by a third-party logistics (3PL) provider or a such as on-demand transportation. Inevitably, this would
business partner who does not face significant demand come at additional cost, which could be termed as the price
increase at the same period. of deliver to promise. Then, it would be up to individual
1202 Journal of the Operational Research Society Vol. 64, No. 8

Figure 7 Profit curves for varying demand (BM).

Figure 8 Profit curves for varying demand (MS).

suppliers to decide whether this extra cost will be covered suppliers’ logistics operations. The performance criteria
by the increased revenue or passed to the customer, putting included lead time and cost efficiency aiming to assess the
probably customer satisfaction at risk. Although the capacity of material distribution companies to provide a
adoption of CTN and 3PL strategies could result in cost timely and cost-efficient service to the construction
savings and reduced lead times, it has to be acknowledged industry under demand uncertainty, one of the key features
that both policies would require substantial reform of of the sector. Certainly, the study can only claim
construction supply chains and thus, are suggested as areas generalisable results in terms of the participating compa-
of further research. nies. However, it is believed that the dynamics of the
logistics processes and performance revealed by this
research will be applicable to other typical supplier
organisations. Therefore, the paper suggested that the use
Conclusions
of CTN and 3PL strategies could reduce the effects of
The study presented in this paper aimed to balance the varying demand on lead times and system costs and thus,
contractor-centric approach that dominates existing con- should be further explored. This would require suppliers to
struction SCM literature by exploring the function of isolate logistics and transportation costs in order to be able
material distribution companies and the impact of demand to assess the efficiency of alternative operational strategies.
uncertainty on supply chain performance. The paper Also, it would call for the whole industry to consider the
demonstrated the potential of simulation modelling to total acquisition cost of construction materials and
investigate logistics operations within construction supply collaboratively explore how this could be reduced through
chains and more specifically to assess the effect of varying developing supply chain efficiencies. This would result in
demand of construction materials on the performance of significant value creation throughout the supply chain,
 C Vidalakis et al—Demand uncertainty in construction supply chains 1203

which would enable more effective delivery of construction Greasley A (2004). Simulation Modelling for Business. Ashgate
projects and ultimately, maximise value for the construc- Publishing Ltd: Aldershot.
tion client. Green SD, Fernie S and Weller S (2005). Making sense of supply
chain management: A comparative study of aerospace
and construction. Construction Management and Economics
23(6): 579–593.
Hauge JW and Paige KN (2001). Learning SIMUL8: The Complete
References Guide. PlainVu Publishers: Bellingham.
Akintoye A, McIntosh G and Fitzgerald E (2000). A survey of Kagioglou M, Cooper R and Aouad G (2001). Performance
supply chain collaboration and management in the UK management in construction: A conceptual framework.
construction industry. European Journal of Purchasing and Construction Management and Economics 19(1): 85–95.
Supply Chain Management 6(3/4): 159–168. Kleijnen JPC (2005). Supply chain simulation tools and techniques:
Ala-Risku T and Kärkkäinen M (2006). Material delivery problems A survey. International Journal of Simulation and Process
in construction projects: A possible solution. International Modeling 1(1): 82–89.
Journal of Production Economics 104(1): 19–29. Law AM and Kelton WD (2000). Simulation Modelling and
Arbulu R, Ballard G and Harper N (2003). Kanban in construc- Analysis, 3rd edn. McGraw-Hill: New York.
tion. Proceedings of the 11th Annual Conference of the Interna- Lee HL, Padmanabhan P and Whang S (1997). Bullwhip effect in a
tional Group for Lean Construction. Blacksburgh: Virginia. supply chain. Sloan Management Review 38(4): 93–102.
Assaf SA and Al-Hejji S (2006). Causes of delay in large London K (2004). Construction supply chain modelling. PhD thesis,
construction projects. International Journal of Project Manage- University of Melbourne.
ment 24(4): 349–357. London K (2008). Construction Supply Chain Economics. Taylor
Azambuja M and O’Brien WJ (2008). Construction supply chain and Francis: London.
modelling: Issues and perspectives. In: O’Brien WJ, Formoso Naim M and Barlow J (2003). An innovative supply chain
CT, Vrijhoef R and London KA (eds). Construction Supply strategy for customized housing. Construction Management and
Chain Management Handbook. CRC Press: Boca Raton, FL. Economics 21(6): 593–602.
Bowersox DJ, Closs DJ and Cooper MB (2007). Supply Chain Ng ST, Fang Y and Ugwu OO (2008). Modelling construction
Logistics Management, 2nd edn. McGraw-Hill: New York. material logistics system with stochastic Petri nets. Construction
BRE (Building Research Establishment) (2003). Construction Site Innovation: Information, Process, Management 8(1): 46–60.
Transport: The Next Big Thing. BRE Ltd: Watford. Ogden HJ and Turner RE (1996). Customer satisfaction with
Caniato F, Kalchschmidt M and Ronchi S (2011). Integrating delivery scheduling. Journal of Marketing Theory and Practice
quantitative and qualitative forecasting approaches: Organiza- 4(2): 79–94.
tional learning in an action research case. Journal of the RHA (Road Haulage Association) (2005). Goods Vehicle Operating
Operational Research Society 62(3): 413–424. Costs Tables. The Road Haulage Association and DFF
Caron F, Marchet G and Perego A (1998). Project logistics: International: Bristol.
Integrating the procurement and construction processes. Robinson S (2004). Simulation: The Practice of Model Development
International Journal of Project Management 16(5): 311–319. and Use. Wiley: New York.
Chaharsooghi SK and Heydari J (2010). LT variance or LT mean Saad M, Jones M and James P (2002). A review of the progress
reduction in supply chain management: Which one has a higher towards the adoption of supply chain management (SCM)
impact on SC performance? International Journal of Production relationships in construction. European Journal of Purchasing &
Economics 124(2): 475–481. Supply Management 8(3): 173–183.
Chan FTS and Zhang T (2011). The impact of collaborative trans- SFC (Strategic Forum for Construction) (2005). Improving
portation management on supply chain performance: A simula- Construction Logistics: Report of the Strategic Forum for
tion approach. Expert Systems with Applications 38(3): 2319–2329. Construction Logistics Group. Strategic Forum for Construction
Cooper MC, Lambert DM and Pagh JD (1997). Supply chain and Construction Products Association: London.
management: More than a new name for logistics. International Shakantu WM, Muya M, Tookey JE and Bowen PA (2008). Flow
Journal of Logistics Management 8(1): 1–14. modelling of construction site materials and waste logistics:
DfT (Department for Transport) (2006). Key Performance Indica- A case study from Cape Town, South Africa. Engineering,
tors for the Builders’ Merchants Sector. HMSO: London. Construction and Architectural Management 15(5): 423–439.
El-Haik B and Al-Aomar R (2006). Simulation Based Lean Silver EA, Pyke DF and Peterson R (1998). Inventory Management
Six-sigma and Design for Six Sigma. Wiley: Hoboken, NJ. and Production Planning and Scheduling, 3rd edn. Wiley:
Esper TL and Williams LR (2003). The value of collaborative New York.
transportation management (CTM): Its relationship to CPFR and Sobotka A and Czarnigowska A (2005). Analysis of supply system
information technology. Transportation Journal 42(4): 55–65. models for planning construction project logistics. Journal of
Fildes R and Kingsman B (2011). Incorporating demand un- Civil Engineering and Management 11(1): 73–82.
certainty and forecast error in supply chain planning models. Tan CT, Lyman SB and Wisner JD (2002). Supply chain
Journal of the Operational Research Society 62(3): 483–500. management: A strategic perspective. International Journal of
Ganapathy S, Narayanan S and Srinivasan K (2003). Simulation Operations & Production Management 22(6): 614–631.
based decision support for supply chain logistics. Proceedings of Taylor JW (2011). Multi-item sales forecasting with total and split
the 2003 Winter Simulation Conference. New Orleans: Louisiana. exponential smoothing. Journal of the Operational Research
Ganeshan R (1999). Managing supply chain inventories: A multiple Society 62(3): 553–563.
retailer, one warehouse, multiple supplier model. International Tommelein ID (1998). Pull-driven scheduling for pipe-spool
Journal of Production Economics 59(1): 341–354. installation: Simulation of lean construction technique. Journal
Georgy M and Basily S (2008). Using genetic algorithms in of Construction Engineering and Management 124(4): 279–288.
optimizing construction material delivery schedules. Construction Tommelein ID and Weissenberger M (1999). More just-in-time:
Innovation: Information, Process, Management 8(1): 23–45. Location of buffers in structural steel supply and construction
1204 Journal of the Operational Research Society Vol. 64, No. 8

processes. Proceedings of the 7th International Group for Lean Vrijhoef R and Koskela L (2000). The four roles of supply chain
Construction Conference. Berkeley: California. management in construction. European Journal of Purchasing
Vidalakis C, Tookey JE and Sommerville J (2011a). The logistics of and Supply Management 6(3/4): 169–178.
construction supply chains: The builders’ merchant perspective.
Engineering, Construction and Architectural Management 18(1): 66–80.
Vidalakis C, Tookey JE and Sommerville J (2011b). Logistics
simulation modelling across construction supply chains. Construc- Received March 2012;
tion Innovation: Information, Process, Management 11(2): 212–228. accepted October 2012 after one revision