Managing subcontractor supply chain for quality in construction

The Authors

Khalid Karim, School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia

Marton Marosszeky, School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia

Steven Davis, School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia

Acknowledgements

The financial and in-kind support of Australand, Barclay Mowlem, Baulderstone Hornibrook, Farrell Management, Hastie, John Holland, Kone, Multiplex, Reed, Star Electrical, and Walter Construction is gratefully acknowledged. Part of the data used in this paper was collected by Shane A. Prior (2002) and presented in his undergraduate thesis, which was funded as part of the research project. The authors are grateful to the reviewers of this article for their extremely useful comments and suggestions.

Abstract

Purpose – To provide a decision support tool for long-term management of subcontractor supply chain for achieving better quality in construction.

Design/methodology/approach – Defects on three construction projects are investigated through direct observation, document analysis, and interviews. A method to analyse and present defects information as an aid to decision making in relation to subcontractor supply chain management is proposed, and its application is illustrated by using the information obtained from the study.

Findings – The importance of managing the subcontractor supply chain to reduce the incidence of defects is established. The nature and extent of the defects, along with what caused them, are discussed. Method for conversion of raw data into a decision support tool is presented.

Research limitations/implications – The data collection method was limited in that it was not based on continuous full-time observation. Such presence by independent observers for full duration of the project would provide more accurate information.

Practical implications – A useful tool for the construction industry in view of the fact that most of the construction work is now done by subcontractors and the head contractors primarily act as project managers.

Originality/value – The concept of using Pareto analysis/histogram for managing quality has been around for a considerable period of time. This paper converts it into a more efficient and useful decision support tool.

Article Type:

Research paper

Keyword(s):

Construction works; Pareto analysis; Subcontracting; Supply chain management; Decision support systems; Operations and production management.

Journal:

Engineering, Construction and Architectural Management

Volume:

Number:

Year:

2006

pp:

27-42

Copyright ©

Emerald Group Publishing Limited

ISSN:

0969-9988

Introduction

Quality remains a critical issue for the construction industry. The extent of the problem, particularly its cost implications, are well documented. The cost of quality rectification problems is of the same magnitude, 3.4 per cent to 6.2 per cent, as the profitability of organisations in the sector (Thomas et al., 2002). Other researchers have put the cost of rework as high as 12 per cent (Burati et al., 1992). Consequently, researchers have focussed on a range of strategies to ameliorate the problem. These include studies identifying causes, magnitude and cost of defects (Love, 2002; Josephson and Hammarlund, 1999); performance measurement and benchmarking as summarised in Karim et al. (2003); management of quality using self-assessment, planning, and control systems including ISO 9000 (Abdelhamid, 2003; Quazi et al., 2002; Dissanayaka et al., 2001); better design and use of technology including design management and control, quality function deployment, constructability, and premanufacture, etc. (Chiang and Tang, 2003; Pheng and Abeyegoonasekera, 2001; Abdul-Rahman et al., 1999); human issues such as leadership, communication, motivation and organisational culture (Bossink, 2002; Marosszeky et al., 2002; Thomas et al., 2002); and management of construction supply chain (Love et al., 2002).

The first of the aforementioned categories of study aims to gather and analyse information at a fundamental level, in order to get to the root of the problem, by taking into account various perspectives such as the type of defects, their frequency of occurrence, cost of rectification and their origin or cause. However, despite the importance of this type of investigation in tackling quality problems, studies to establish a comprehensive taxonomy of defects and related causes have been quite limited. Atkinson (1999) investigated the role of human error in construction defects in the UK. This research involved a general survey of construction industry practitioners, and three studies limited to speculative house-building. Olubodun and Mole (1999) researched factors influencing defects in public housing in UK. This study was based only on a survey of building inspectors and within this context investigated the influence of design, construction, standards, vandalism, and age of building. Love and Sohal (2003) studied two projects in Australia to determine the causal structure of rework. The focus of this study was on rework, wherein rework was the set with various elements such as variations etc. Defective work was an element of the set and was not treated and analysed as a separate subset. It would appear that the only in-depth study into the causes of defects in construction was carried out in Sweden. Josephson and Hammarlund (1999), following up on an earlier study of their own, investigated seven different types of building projects. In this paper, they report root causes of defects as being stability in the client organisation, client's project control, user involvement, time pressure, composition of project organisation, cost pressure (including its relationship with procurement practices), support to the site organisation and motivation of the workers.

A key aspect missing from these studies is a lack of specific focus on subcontractors. Most of the work is done by subcontractors as the main contractors rely on a large number of subcontractors. As much as 90 per cent of the construction work is carried out by a variety of subcontractors while the main contractor tends to focus on management and coordination. To make matters worse, a large number of the subcontracting firms are small because the contractors tend to employ pyramid subcontracting by using multiple tiers of subcontractors. In Australia, for example, 94 per cent of construction trades employ fewer than five people, and less than 1 per cent employ more than 20 people (Commonwealth of Australia, 1999). As a consequence, most of these firms simply do not have the resources to adopt modern principles of quality management although their smaller size does provide them the flexibility to be able to adopt innovative methods. Further complications arise because the subcontractors are contractually obligated to the head contractor, whereas the construction process flows from one subcontractor to the other, as shown in Figure 1.

Consequently, the subcontractors do not recognise their following trades as customers (Rethinking Construction, 1998), despite the fact that defective work by a preceding trade has a cost and time impact on the following trade. These problems, combined with the fact that the head contractor has the ultimate responsibility to deliver quality at a competitive cost, make it necessary for the head contractor to manage the supply chain effectively. Approaches such as partnering etc. have been introduced to overcome these problems. Unfortunately, though, the uptake of these concepts by the construction industry has been quite slow and is still in its early stages (Love et al., 2004). As a result, the development and use of basic tools such as objective methods of performance measurement can be valuable for both short-term problem resolution as well as long-term strategic decision making. To this end, collection, analysis, and appropriate presentation of defect related data is essential. It was, therefore, proposed that an investigation of defects on construction sites in Australia was needed because:

It will provide the level of detail essential for effective management of the subcontractor supply chain.
Despite the general nature of construction industry being the same, there are unique country related issues that require country specific study.
The large proportion of the work done by subcontractors creates the need for examining how data can be collected and used for decision-making purposes in relation to management and coordination activities by the main contractor.

In order to achieve the above objectives, a study was carried out on three construction projects in Sydney, Australia to examine the nature, extent and causes of defects and to develop a strategy for using this information for the purposes of better managing the subcontractor supply chain. This paper illustrates how the information pertaining to the type of defects and their rectification costs may be transformed into a decision support tool. For the purposes of this study, defect was defined as “a non-conformance with the contractual documentation”.

Background to the case study

In recognition of the continuous existence of quality issues in construction, ten leading construction firms joined with the University of New South Wales, Australia to support the first phase of a long-term project to improve quality of the constructed product. While the ultimate objective of the project is to develop and implement proactive strategies for defect avoidance, the first phase was designed to examine the causes and cultural issues associated with defective work and develop some preliminary tools and techniques to overcome these problems. There were two main parts to the study. One investigated defects while the other investigated the organisational cultures existing on construction projects. This paper pertains to the former. A total of three projects were selected for inclusion in study, mainly because of their timely availability for the study period and their location to provide easy accessibility for observation and collection of data.

The common feature of all the three projects was the construction of new facilities in heritage listed areas. This meant that the outer skin of the building could not be demolished despite extensive structural changes, additions, and alterations. Furthermore, in view of the complex nature of the work, all the projects were carried out under D&C contracts.

Data collected on all the projects essentially related to the fitout and finishing stages because the industry partners identified these as the critical stages where a greater proportion of defects was likely to occur.

Projects description and data collection

Project 1 involved re-development of an old two-level building into a four-level block of luxury apartments, costing A$24 million. The project was in the completion phase when the study commenced. However, the main contractor had maintained a defects database for second half of the project period, that corresponded to a little over 60 per cent of the construction work. Information contained in this database was used for the study and was supplemented by informal interviews with the contract manager.

Project 2 involved construction of mixed use (residential apartment/offices) buildings by converting and adding to old wharf buildings. The total contract value of the works included in the study was A$40 million. Three types of defects lists were being maintained in relation to this work by the main parties to the contract – the architect, contractor, and the client. After some deliberation, it was decided to use the architect's list in order to avoid any duplication or repetitive analysis. Ideally, all three lists should have been analysed and, after eliminating duplications, a comprehensive list should have been prepared to ensure that each and every defect was included. However, given that there were thousands of items in each list, it would have been a daunting task requiring time input that was not available. It was decided that the number of items on one list, namely 15,000 in this case, was sufficiently representative for the purposes of analysis.

Project 3 was an old hospital that was being converted into an apartment building through major renovation and new construction at the cost of A$22 million. A total of 19 apartment units were included in the study. The researcher visited the site once a week for a period of nine months, and collected defects data from three sources. These include a non-conformance report prepared by the head contractor, defect lists prepared by the head contractor and a defects report prepared by an independent third party on achievement of practical completion.

Nature and extent of defects and their causes

The total cost of defects ranged from 2 per cent to 6.22 per cent of the project value. This includes the direct costs (labour and material) as well as the indirect costs (organisation and supervision) but does not include preliminaries. Also, the cost of delays due to defective work and any associated penalties have not been included. All the projects were similar in terms of the end product being residential/office buildings. The highest cost of defects (6.44 per cent of project value) occurred on Project 1, which is quite substantial as compared to profit margins. As noted earlier, this relates to the latter 60 per cent of the work that constituted primarily the fit out stage. Since the number of defects was very large, a defect survey of the type used on Project 1 could not be implemented. Analysis of the list provided by the contractor revealed defects ranging from minor paint marks to very serious matters such as incorrect reinforcement. The list also included incomplete works, but these were not taken into account when computing the cost of defects.

It would appear that bathrooms and kitchens are the two main areas of defective work in comparative terms, followed by bedrooms, laundries and other areas. While this may be expected given the fact that bathrooms and kitchens have many services and fittings unlike other rooms, the nature of defects did indicate an attitude of carelessness at the least and negligence at worst.

The cost of defects on Projects 2 and 3 was found to be, respectively, 2 per cent and 4 per cent of the contract value, respectively. On Project 2, the highest total cost of defects was in the kitchen area followed by tiling, painting, plumbing, and carpentry. On Project 3, the highest total cost of defects pertained to joinery work, followed by tiling, windows, painting, and formwork.

Many authors have discussed the various causes of defective work in construction but an analysis of their discussions points to a few key categories. These include design or documentation deficiencies/changes as well as variations by clients leading to such changes/errors and damage (Love and Sohal, 2003), planning and administration (Josephson, 1998), workmanship (Hall and Tomkin, 2001), and materials, machines and equipment (Josephson, 1988).

Table I shows the classification of the causes of defects on these projects. It is interesting to note that the predominant cause of defects emerging on the projects in this study was workmanship. While a number of authors have rightly noted design and documentation as a major issue, it does not feature prominently in this case because the study focussed on later stages of the project. Design problems tend to have been resolved by this time.

On Project 1, workmanship caused 92 per cent of the defects, while on Projects 2 and 3 it was responsible for 88 per cent and 73.6 per cent of the defects, respectively. The parallel investigation of organisational culture revealed that the problem of workmanship was being driven by lack of motivation (Thomas et al., 2003). This was particularly evident when variation in quality was observed in different locations with the same group of people performing the same type of work. Since most of the work is done by subcontractors, motivation of workers is almost exclusively related to better management of and by subcontractors. This imparts even greater importance to the need for better management of subcontractors by the head contractor.

Data analysis and presentation for decision making in relation to subcontractor management

The data obtained in this study can be used in many ways. The defects can be seen in terms of their locations, trades, subcontract packages and in relation to cost and numbers. Figure 2 shows an example of defects related to the general description of work area, which may involve one or more trades.

Alternatively, it may be used for comparison between trades (Figure 3). The inter-trade comparison may also be based on the number or cost of defects. Further, subcontract packages within each trade can also be compared in the same manner. However, these need to be normalised for a valid comparison, to be comparable, e.g. as a proportion of the contract value (Figure 4).

The most important consideration in choosing the method(s) of data analysis and presentation is the utility of the chosen method as a management and decision-making tool. The use of data for going past the superficial and converting it into a useful management tool is not uncommon in other industries, though previous studies investigating defects in construction have not focussed on how the data can be analysed and presented for management of subcontractors by the head contractors. Evraert and Riahi-Belkaoui (1998) discuss the benefits of value added reporting for US corporations by extending the reporting to include a larger group of stakeholders. In the context of construction, that would mean value addition through good quality performance by subcontractors and value degradation through bad performance. Lin et al. (2004) treat quality management practices as inputs to organisational performance. Mills et al. (2003) note that good fortune would be the only alternative to a resource and competence-sensitive strategy for firms wishing to achieve sustainable advantage, their definition of resource including the use of third parties (the equivalent of subcontractors in construction). While such examples highlight the importance of use of quality information, an appropriate format to aid decision making also needs to be determined. Craft and Leake (2002) argue in favour of simple approaches such as use of the Pareto principle. So and Smith (2003) analysed methods for presentation of multivariate information for management decision making and found graphical/pictorial methods to be the most useful. The method of data analysis and presentation discussed below is based on similar principles to provide quality information in a simple and effective manner.

One of the founding fathers of the post-war quality movement, Juran (1954) proposed that attributes can be examined by using the so called Pareto Principle (a couple of decades later, as cited in Juran (1992), Juran stated that reference to Pareto's work as a universal principle was incorrect but nevertheless the practice continued). The objective was to differentiate between “the vital few and the trivial many” for the purposes of planning and prioritising the use of resources. The technique simply requires the plotting of any one of the attributes in descending order, in the form of a chart of individual values, and/or a cumulative curve. This enables the investigator or the decision maker to select attributes with predominant impact for further investigation or consideration. Figures 2 and 3 can, in fact, also be classified as Pareto histograms because they show the proportion of defects in descending order.

Despite its simplicity, the ‘Pareto Principle’ is quite powerful and has been used widely. For example, Au and Yu (1999) used Pareto analysis to examine the nonconformities to various ISO 9001 clauses to determine critical quality related shortcomings on a construction project. Goh (2000) found the use of Pareto analysis in the context of identifying cost components to design more economical car parking stations. Juran (1992) promotes its use in many different contexts such as prioritising customer requirements, examining the cost of poor quality, and analysis of human errors in relation to quality. Craft and Leake (2002) discuss its use in prioritising information service resources in manufacturing.

Knights (2001), in discussing its use for prioritising equipment maintenance, noted that where an outcome is a product of two factors, a Pareto histogram based on one factor alone cannot determine which factor is dominant in contributing to the problems. Further, by focussing on any one factor to the exclusion of others, it may miss identifying individual events having high costs or time consumption, or those that may consume relatively little average cost or time but have the ability to cause operational disturbances. Lastly, Pareto histograms are not useful for trending comparisons across different time periods.

Pareto has been frequently used as tool for quality management but the problems in relation to its use in managing subcontractors in construction are not dissimilar to those mentioned in relation to equipment maintenance by Knights (2001). Firstly, the overall impact of defective work is a product of the frequency of defects as well as their cost of rectification. Secondly, because construction work comprises a sequence of trades, defective work can and does cause downstream operational disturbances, even though the cost of rectification of a particular defect in itself may be low. Thirdly, strategic long-term management of subcontractors requires objective assessment of their performance over time. Consequently, the technique suggested by Knights (2001) has been adapted and improved as an aid to examine and manage subcontractor performance on construction projects. This approach is demonstrated below by using the data given in Table II. It may be noted here that, since this is only an illustration, the data in Table II pertains to just one project (Project 3). It comprises a stratified random sample taken on the project, with only direct costs of rectification being taken into account. In actual fact, the indirect costs have been found to be of the same order as the direct costs (Marosszeky et al., 2002). Consequently, the repair cost per defect would be twice that used in the example below. However, this will not have any impact on the outcome because application of the Pareto principle is an examination of the relative importance of attributes.

Step 1

The first step is to plot an x-y scatter plot of the total number of defects and their average rectification cost, for each trade (Figure 5). Each point is labelled with the trade code. It will be noted that the cluster on the lower left corner makes it difficult to label the points. The reason for the formation of this cluster is the large range of data points, which is skewed towards the lower end. This problem is easily resolved by using a log-log plot, as shown in Figure 6. The log transformation simplifies identification of data by reducing the skew and making it more symmetric. Furthermore, it enables all points of same total monetary value (i.e. product of number of defects and their average cost) to be connected by a straight line.

The usefulness of the information and its portrayal in this manner is immediately noticeable. The person examining the information can see the incidence as well as its cost impact in three ways. It can be observed in terms of cost ranking alone, or frequency of occurrence alone, or in terms of a combination of these two. In the case of structural steel (ST) for example, there are only two defects but their cost impact is very high. On the other hand, the painter (PT) has a large number of low-cost defects.

By observing a data point in terms of both the number of defects as well as their average cost, an assessment of their combined impact may also be made. For this purpose, the visual utility of the chart can be further enhanced by using Isoquant lines as shown in Figure 7. Isoquant lines represent points of same total value (i.e. product of number of defects and their average cost), thus making it easier to simultaneously visualise and compare the total cost impact. It is clearly evident from this chart that in terms of total cost impact on this project the joiner is the worst performer, followed by the tiler, window contractor, the painter, and so on.

Step 2

As noted earlier, subcontractors do not tend to recognise the importance of their role in relation to the trades that follow their work. This can have serious implications in terms of the potential disruption of the following trades as a consequence of defective work by the preceding trade or the sequence of preceding trades. Consequently, the head contractor can also use the modified Pareto chart to analyse the quality performance of groups of trades, as shown in Figures 8 and 9.

Figure 8 shows defects' information relating to three trades that precede the painting subcontractor, namely, rendering, bricklaying, and plastering. In this case, the plasterer had caused a large number of defects (113), that may have impacted the painter's work. At the very least, such large number of defects by a preceding trade may result in productivity pressures being applied to the following trade, resulting in further quality problems. Similarly, Figure 9 shows an example of a chain of subcontractors forming a sequence of trades that precede the tiler's work. The renderer and waterproofer directly preceded the tiler. However, the concrete placer, while immediately preceding the tiler, relies on the steel fixer whose work in turn relies on the formworker. The dashed arrows from the renderer and concretor to waterproofer indicate that the type of work may require an alternate sequence, which extends the supply chain. Such sequencing can and does have a cumulative downstream effect, the longer the sequences the greater the magnitude of quality issues and productivity pressures.

Step 3

While Figures 7-9 display information for a single project, there remains the issue of the performance of various trades on different projects over time. This is important in terms of development of quality management strategies and plans, monitoring post-implementation improvements, as well as long-term relationship management. The trending chart (Figure 10) can be used for this purpose. It can be seen from the chart below how the defect related performance of the trades identified as requiring attention can be tracked over time across different projects. It may be noted that due to the availability of data from only one site, the chart values are hypothetical because a factual example would require multiple projects of the same type from the same contractor.

Conclusion and recommendations

There is a significant body of literature on quality in construction. However, only a few studies have focussed on detailed investigation of defects and their outcome is not presented in the context of managing the subcontractor supply chain despite the fact that up to 90 per cent of the work is done by subcontractors. There is a need to articulate the SC supply chain and provide data to examine various interfaces within this supply chain. Information pertaining to defects can be analysed and presented in a manner that helps decision making and prioritisation of resources in this context. Pareto charts have been used quite extensively for this purpose but they have their own limitations. The strategy presented in this paper provides a much greater ability for analysis and decision support by linking cost and frequency as a rational basis for decision making. It is easier to present this information to help decision makers who do not have a lot of time at their disposal. Priorities for subsequent action can be established with the help of this tool. The tool may, however, be further improved by incorporating the cost of defects as a proportion of the total contract value of the work of each trade to get an even better perspective. This information was not available for the projects used as case study in this paper, except to a very limited degree.

A drawback of the information obtained for this study was its reliance on information obtained and provided by head contractors. Consequently, the method of data collection was not uniform across the projects, and even reasonably comprehensive data bank maintained by the contractors did not contain all the variables desirable for such a study. Ideally, this requires continuous presence of observers on sites but that requires extensive financial resources. Consequently, an alternative approach has been adopted for the next phase of this project, which involves the use of a Defect Incident Record (DIR).

Figure 1Conflict between contractual obligation and process flow

Figure 2Comparison by area of work

Figure 3Comparison of trades

Figure 4Comparison of subcontract within trades

Figure 5Scatterplot of number of defects and average cost

Figure 6Log-log scatterplot of defects and cost

Figure 7Isoquant lines

Figure 8Tracking defect of preceding trades and following trade

Figure 9Tracking sequence of preceding trades to following trade

Figure 10Trending chart for trades

Table IProportion of defects according to cause

Table IIFrequency and direct cost of defects for each trade

References

Abdelhamid, T.S. (2003), "Six-sigma in lean construction systems: opportunities and challenges", in Martinez, J.C., Formoso, C.T. (Eds),Proceedings of 11th Annual Conference on Lean Construction, Virginia Polytechnic Institute and State University, Blacksburg, VA, pp.65-81.