Analyzing Contingency Estimation for Residential Turnkey Projects in Saudi Arabia: A Neural Network Approach

Abstract

Utilizing a turnkey approach to deliver a construction project entails significant risks from the contractor’s perspective. Essentially, the owner awaits project completion without commitments regarding additional expenditures incurred by the contractor during the project’s duration. This paper specifically focuses on estimating and analyzing the contingency value for residential turnkey projects in Saudi Arabia. The contingency value across the project’s life cycle is estimated using six Artificial Neural Network (ANN) models, which are compared to identify the best-trained network according to project complexity, contingency factor, and contingency impact during the project phases. The output layer provides the contingency factor percentages for each project phase. A 13-story reinforced concrete (RC) residential building established in one of Saudi Arabia’s cities was selected to implement the developed methodology. The contingency estimation, performed using @Risk 7.5 and NeuralTools 7.5, was determined to be 11.34% and was distributed across the five phases of the project’s life cycle: 0.30% for predesign, 0.99% for design, 2.61% for preconstruction, 6.33% for construction, and 1.12% for postconstruction. Furthermore, it was found that the estimated contingency varies based on project complexity, which is 7.20% for low complexity, 8.16% for medium complexity, 9.41% for complicated, and 11.34% for very complicated projects. Historical data and peer review approaches are employed to validate the results, both of which are endorsed by professionals in this field. This paper highlights two main contributions: Firstly, it significantly enhances risk management by facilitating a comprehensive understanding and systematic analysis of risks, thus improving the contractors’ ability to mitigate potential negative impacts on projects. Secondly, it supports more informed decision-making through the use of advanced techniques to estimate and analyze contingency values. These contributions are critical for contractors engaged in Saudi construction projects, particularly those involving residential buildings.

1. Introduction

The residential construction sector in Saudi Arabia holds significant value, with an estimated worth of around USD 37 billion and a projected Compound Annual Growth Rate (CAGR) of over 6% until 2029 [1]. In addition, projections indicate the completion of approximately 300,000 residential units by the end of 2025 [2]. This emphasizes the burgeoning opportunities for contractors, developers, investors, and various other stakeholders in the construction and real estate sectors. It is a dynamic landscape, influenced by factors such as demographic shifts, technological advancements in construction, and evolving consumer preferences. This development warrants further attention, particularly when undertaken by the private sector, which heavily relies on turnkey approaches to deliver the majority of these projects. Turnkey delivery is preferred when owners seek minimal involvement in the project. Hence, the contractor assumes responsibility for navigating the challenges inherent in the project’s activities. Typically, addressing these challenges incurs additional costs; failure to do so may result in repercussions extending beyond the specific activities, potentially impacting other aspects of the project with increased costs and delays, or even halting progress altogether. Given the alarming frequency of accidents on a global scale, construction projects place significant emphasis on managing risks throughout their life cycle. However, effectively eliminating or mitigating risks requires the active involvement of stakeholders in adhering to a comprehensive risk response plan. Such a plan should be developed during the initial planning stage and continuously updated throughout the project’s progression. It comprises various subplans aimed at implementing necessary measures to address risks throughout the project’s lifespan. Implementing effective and suitable risk responses has the potential to diminish both individual and collective project risks while also enhancing individual and collective opportunities [3]. A risk response plan delineates the tactics designed to minimize adverse project risks, which are divided into two categories, including known-unknown risks (contingency) and unknown-unknown risks [4]. For the current study, an extensive review of relevant journal papers, technical documents, and direct interviews with specialists in the fields of cost estimation and risk management were conducted. This comprehensive approach helped identify a crucial problem prevalent among contractors in Saudi Arabia: accurately managing the project’s budget with respect to contingency costs during various project phases. The goal was to find a practical solution to this pervasive issue.

2. Objectives

With the construction industry experiencing significant growth in Saudi Arabia, many construction companies in the region are encountering challenges in delivering their projects with the anticipated profits and within the scheduled deadlines. The primary challenge lies in accurately estimating the costs to be submitted for bidding, which involves accounting for both direct and indirect costs along with the markup. It is essential to add extra costs to the markup to account for profit and to mitigate the risks and contingencies that may arise during the construction phase. Estimating the associated risks is particularly challenging as they often fall under the category of “unknown unknowns” and may potentially be covered by insurance or warranties. Therefore, focusing on contingency planning becomes crucial for addressing the anticipated additional costs related to both direct and indirect expenses. The prevailing method of determining contingency involves either applying a fixed percentage of the total cost, which is flawed, or allocating a fixed percentage to individual activities. This approach results in inaccurate estimations due to various factors, including overlaps between activities sharing the same contingency factors and the dynamic nature of these factors throughout the project lifecycle. The main objective of the current research is to estimate the contingency value of the residential construction project with respect to the turnkey delivery method, as a percentage of the total cost during the planning stage. Other subobjectives are as follows:

(1)

Determine the primary factors contributing to project contingency.

(2)

Estimate the contingency percentage for each project phase based on the project’s complexity and the importance level of the contingency factors.

(3)

Analyze the results to comprehend the impact of the contingency factors on the project throughout its life cycle.

3. Background

3.1. Turnkey

The turnkey delivery method is widely used in construction projects, particularly when the project scope is well defined, and the associated risks can be effectively managed from the contractor’s perspective. However, contractors often encounter challenges in construction projects, including but not limited to the following [5]:

(1)

The absence of owner involvement: Owners of turnkey projects frequently take a passive role compared to the contractor, who handles every element of the project. This absence of owner supervision may result in misunderstandings, poor communication, and unhappiness with the way the project turns out.

(2)

Ambiguity in scope characterization: Effective project execution depends on a precise characterization of the project’s scope. However, in turnkey projects, there may be disagreements or ambiguities in the scope of work, which could result in arguments about deliverables, responsibilities, and project requirements between the contractor and the owner.

(3)

Risk allocation: In turnkey projects, the contractor bears the majority of project risks, with the owner bearing the remainder. Accurately evaluating and controlling project risks can be difficult for contractors, particularly if unanticipated problems crop up while the project is being built. This may result in disagreements about who bears the responsibility for risk-related problems, delays, and cost overruns.

(4)

Cost overruns and schedule delays: Unexpected site circumstances, design modifications, supply chain interruptions, or contractor faults can all cause cost overruns and schedule delays on turnkey projects. It can be difficult for contractors to predict project prices and schedules with enough accuracy, which can result in losses of money and damaged relationships with owners.

3.2. Contingency

During the early stages of the project, a residential cost estimate using the parametric estimate technique is more suitable than other techniques, due to the availability of data of the previous projects [6]. With the historical data availability, contingency magnitudes decrease to 3 to 5% and are added to the bid to cover the allowances [7]. In addition, proper evaluation and management of risk contingencies can significantly impact cost certainty [8]. A study by [9] identified contingency as an event that one must be prepared for, comprising the contingency amount in terms of cost or against the budgeted amount and a contingency plan to be enacted when the contingency occurs. The authors in [10] highlighted that contingency is a crucial component that should be included in the base estimate of the project to account for uncertainty or known unknowns. It also serves to accommodate changes that may arise during the project’s lifecycle. It is emphasized by [11] that contingency may be employed to tackle unforeseen conditions associated with design, change orders, and other risky tasks. Consequently, the reserved amount for contingency is allocated to fund these conditions. In contingency cost analysis, to determine critical work packages, a straightforward approach is to employ the criticality rule. According to this rule, an item is deemed critical if its cost overrun or underrun is equal to or greater than 0.5% of the project’s total target cost. Conversely, items costing less than 0.5% of the total target cost may be considered noncritical, as fluctuations in their prices are unlikely to significantly affect the overall target cost. Noncritical cost items can then be grouped together into a single category and held constant throughout the contingency cost analysis [12]. The evaluation of perceived ratings of cost and schedule impact, along with the relative importance of identified risk drivers on cost contingency amounts, was conducted using stepwise regression modeling [13]. From the owner’s perspective, the authors in [14] proposed a methodology for estimating and allocating the required contingency cost among the project’s activities based on historical data, site conditions, time constraints, constructability issues, and project scope. Across nine construction projects, the average total contingency was found to be 3.86%. The maximum contingency utilized was 21.37%, while the minimum was −7.59%. The negative contingency value represents the reserved contingency value that is not used, and the project is underbudget. A known-unknown contingency plan is established during the preconstruction phase by multiplying the probability by the consequences of the risk factors. Consequently, it is evident that most of the construction projects under study exhibit conservative contingency and risk values [4]. Most previous literature reviews and practices [6,7,8,9,10,11,12,13,14] have largely overlooked the incorporation of project phase and complexity into their contingency estimation. A contractor managing a turnkey project must carefully distribute the contingency cost in accordance with the expected contingency factors during each project phase (predesign, design, preconstruction, construction, and postconstruction). Such strategic allocation enables the contractor to minimize expenses and effectively control the project budget to attain the anticipated profit. The current study addresses this critical gap by examining the contingency factors and their impact on project activities for each phase, thereby providing invaluable support to construction contractors, particularly those engaged in turnkey residential projects.

3.3. The Applications of Artificial Neural Networks (ANNs) in the Construction Industry

Artificial Neural Networks (ANNs) are increasingly utilized in the construction industry, capitalizing on their capabilities in pattern recognition, predictive modeling, and decision-making support. Numerous researchers have employed ANNs to estimate project costs and facilitate budget allocation [15,16,17,18,19,20,21,22,23], as well as to enhance project scheduling [24,25,26,27,28]. The process of designing, training, and using Artificial Neural Networks (ANNs) involves several key stages, each contributing to the network’s ability to perform tasks such as classification, regression, prediction, and more. The process is divided into the following stages: problem definition, data collection, data preprocessing, designing the network architecture, initialization of weights, model training, model evaluation, parameter tuning and regularization, testing and deployment, and monitoring and updating [29]. The structure of an Artificial Neural Network (ANN) is organized into three major layers: input, hidden, and output. To enhance the accuracy of the output, the hidden layer may comprise multiple layers. The effectiveness of the process depends on having sufficient data for training, testing, and predicting outcomes. Consequently, specific software is required to carry out these procedures. Currently, there are numerous commercially available software options designed to facilitate ANN applications.

4. Research Methodology

Figure 1 illustrates the project methodology, which comprises two primary stages: the contingency model for estimating bidding costs (or proposed cost for private projects) and the life-cycle analysis of associated contingencies. In the first stage, the contingency value is determined using an ANN after identifying the associated turnkey contingency factors of residential buildings. The second stage involves the distribution of the contingency value across project phases to establish the contingency baseline of the project.

4.1. Contingency Model

The contingency percentages of individual factors for each project phase are determined using an Artificial Neural Network (ANN) [30], as illustrated in Figure 2. To achieve reliable accuracy, it is essential to provide the network with more comprehensive data from construction projects. The proposed model includes projects (k) with their features [complexity (x), phases (j), and the contingency factors (i) that affect the project throughout its life], while the output is the contingency percentage (C_kxji). For the current paper, the output (C_1xji) represents the contingency percentages of the new four projects (C_{11ji-low complexity}, C_{12ji-medium complexity}, C_{13ji-complicated complexity}, and C_{14ji-very complicated complexity}), which are then required to further predict their contingency percentages. For instance, (C_11ji) represents the contingency percentage of the future project with low complexity, therefore it is necessary to aggregate all predicted contingency percentages across the five phases (j) according to the related contingency factors (i). Similarly, the contingency percentages of the other three complexity levels (C_{12ji-medium complexity}, C_{13ji-complicated complexity}, and C_{14ji-very complicated complexity}) can be obtained.

Table 1. Complexity level (input #2).

Category	Complexity Level
	Very Complicated	Complicated	Medium	Low
(1) Number of stories	more than 9	6–9	3–5	Less than 3
(2) Design	Very High (Iconic or landmark projects with highly intricate architectural designs and structural systems)	High (Large-scale projects with complex architectural features and structural requirements)	Medium (Moderate-sized projects with some architectural complexity)	Low (Simple and straightforward designs with standard building components)
(3) Sophisticated machinery and new technology	Very High (Advanced construction equipment and tools)	High (Automated and robotic machinery and tools)	Medium (Specialized construction machinery and tools)	Low (Heavy equipment and tools)
(4) Safety requirements	Continuous Improvement and Innovation	Regulatory Compliance	Site-specific safety plan	Basic safety requirements

can assist in assigning the complexity category, offering complexity levels determined by four criteria: the number of stories, design complexity, utilization of equipment and technology, and safety requirements.

A decision maker can select the complexity level based on the highest criteria level. For example, if the number of stories exceeds 10 and the design level is low, the complexity level is classified as “very complicated”. This cautious approach is necessary to determine the final contingency value, covering the uncertainty of the project complexity. The third input is the project phases, which include predesign, design, preconstruction, construction, and postconstruction. In each phase, a contingency factor may (or may not) contribute to the total contingency factor for several reasons, such as the impact duration and the consequences of the specific factor. Some factors contribute to the project from the start to the end, such as inflation, while others affect the project for a limited time, such as permissions. Hence, this input is essential for classifying the effect of the factor on the project’s phases. The fourth input is the contingency factors detailed in

Table 2. Contingency factors.

Financial	Legal	Technical	Site Condition	Stakeholders
Inflation	Dispute with owner	Design issues	Weather	Subcontractor changes
Funding	Social impacts	Constructability problems	Safety	Human errors
Bankruptcies	Environmental impacts	Soil problems	Site layout	Conflicts
Warranty	Permissions from authorities	Incorrect scope	Poor logistic	Low skills
Insurance	Change in law	Quality issues	Accessibility
Market fluctuation		Materials	Security
		Machines

, which presents the contingency categories and their factors, identified through input from experts, insights gained from the lessons learned, and a comprehensive literature review [31,32,33,34,35,36,37]. These factors can be adjusted to include more or fewer elements based on the project’s requirements.

Contingencies are classified into five categories: financial, legal, technical, site conditions, and stakeholders, encompassing 28 factors contributing to the contingency value for turnkey residential buildings in Saudi Arabia. Factor number 29 (Others) can be added to the list if it is required based on the uniqueness of the new project. It is important to note that cost overrun and time delay are not included as separate factors since they are often consequences of the other factors. Given that the current study concentrates on a particular geographic area and specific types of construction and delivery methods, it becomes imperative to pinpoint the contingency factors directly affecting deviations from the project baseline. In this research, a questionnaire survey has been selected as the method to pinpoint the significant factors adversely affecting project performance in terms of cost and time. The Relative Importance Index (RII) serves as the primary tool for assessing the significance of these contingency factors. The Relative Importance Index (RII) is a statistical measure used to assess the relative significance or importance of different factors or variables within a dataset. It is particularly useful in situations where there are multiple factors influencing a particular outcome, and it helps prioritize them based on their relative contributions. A high RII value indicates that the factor contributes significantly to the total project’s contingency and vice versa [38]. The resulting contingency values will serve as a vital component of the contingency estimation for turnkey residential projects, reflecting the specific geographic region under study. The output provides the contingency percentage of each factor for the new project, across each phase. The final contingency percentage of the new project is the sum of all participating contingency factors, which is calculated using Equation (1).

$${Con}_x (\%)={\sum }_{j=1}^m{\sum }_{i=1}^n{Con}_{xji} (\%)$$

(1)

where ${Con}_x$is the contingency percentage for project complexity xth [x = 1,……,4], and ${Con}_{xji}$ is the contingency of factor ith [i = 1……n] according to the project phase jth [j = 1……m]. For the current study, x = 4 (complexity levels), j = 5 (project phases), and i = 28 (contingency factors).

4.2. Contingency Life-Cycle Analysis

The contingency percentages obtained from the first model necessitated further analysis to determine the total contingency for each project phase. This step is crucial for effectively monitoring and controlling contingency expenditure throughout the project’s life cycle. Figure 3 illustrates the methodology for deriving the five percentages corresponding to each project phase. To determine the impact of each contingency factor on each project phase, the Effect Matrix is utilized. A value of ‘1’ is assigned if the factor affects a project phase, and ‘0’ if it does not. Consequently, the final output of the matrix is [0,1] only. In the second stage, a Probability Matrix is constructed by distributing 100% of the contingency factor value among the project phases. For example, if a factor impacts both the construction and post-construction phases, 100% is allocated to these two phases. Both matrices (Effect Matrix and Probability Matrix) are populated and validated using two sources: expert judgment and data from previous projects. In the third stage, the contingency value of each factor is allocated among the five phases based on the percentage determined for each phase in the second stage. For instance, if the contingency value of factor “A” is USD 100, and the factor influences preconstruction, construction, and postconstruction with percentages of 10%, 60%, 30%, the contingency values would be allocated as follows: USD 10 for preconstruction (10% of USD 100), USD 60 for construction (60% of USD 100), and USD 30 for postconstruction (30% of USD 100), as determined using Equation (1).

Thus, the total contingency value for each project phase is calculated as the sum of the contingency values associated with all factors within that phase. This allows for the determination of the percentage of the contingency for each project phase. Lastly, the validation of the developed model and its results is necessary. This validation can be achieved by comparing the results with those of other projects. Acceptable results can then be incorporated into the cost estimation process for new projects. However, if the results are not satisfactory, the model should be revisited, and its parameters verified and validated until the model is deemed accurate and acceptable. Gathering more data on contingency factors, their values, and percentages from various types of residential construction projects is crucial for training the ANN and obtaining precise results.

5. Model Implementation

Currently, a new luxury residential building project is in the planning stage, as illustrated in Figure 4. The project entails the construction of a 13-story reinforced concrete (RC) residential building designed to withstand both gravitational and lateral loads. Located in Haql City, Saudi Arabia, the building will employ a two-way solid slab system for its floor slabs.

The initial cost estimation stands at SAR 80 million, excluding any contingency value. An additional SAR 5 million has been allocated to address potential unknown-unknown risks, pending approval from the company’s senior management. Opting for a turnkey delivery method, the company anticipates substantial profits, given the booming residential market in the region. To account for unforeseen events during the project’s lifecycle, the company seeks to estimate and analyze the project’s contingency value. The proposed methodology outlined in this study will be employed for this purpose, leveraging insights from questionnaire responses, literature reviews, and market reports.

Table 3. Demographic breakdown.

Category	First (Frequency, %)	Second (Frequency, %)	Third (Frequency, %)	Total
Degree	PhD (11, 19%)	Master (12, 21%)	B.Sc. (35, 60%)	58
Position	Manager (6, 10%)	Engineer (28, 48%)	Technician (24, 42%)	58
Experience	>20 years (18, 31%)	10 to 20 years (15, 26%)	<10 years (25, 43%)	58

depicts the demographic breakdown of the respondents, who are currently involved in Saudi construction projects.

Table 4. Previous construction projects.

Project Complexity	No. of Project	Project No. (k)
Low	6	2, 3, 4, 5, 6, and 7
Medium	4	9, 10, 11, and 12
Complicated	3	14, 15, and 16
Very complicated	3	18, 19, and 20

shows the collected data of the previous Saudi construction projects used in this study. The “k” value represents the project number utilized in this paper. The project numbers 1, 8, 13, and 17 correspond to future projects with low, medium, complicated, and very complicated complexity, respectively. In the current case study, number 17 is assigned to the 13-story building under study.

5.1. Contingency Model

Performing the ANN model posed challenges in determining the contingency percentages of factors that contributed to previous residential projects. The collected data frequently involved combinations of multiple factors rather than individual ones. Consequently, extensive analysis and direct meetings were conducted to categorize the data, in alignment with the current model’s requirements. However, this process highlighted the need for additional data to refine the accuracy of the results obtained from the ANN model. The data of the new project is added to the model to obtain the percentages of the contingency factors. NeuralTools 7.5, a Microsoft Excel neural nets add-in, is used in this study to implement the adopted methodology. This software trains, tests, and predicts the provided data, but it is limited to Generalized Regression Neural Net (GRNN) and Multi-Layer Feedforward Network (MLF) configurations for numeric predictions. Both methods are adopted in this study to determine the contingency percentage of the turnkey construction projects in Saudi Arabia. The number of hidden layers is not specified for either method, and the number of nodes is limited to between 2 to 6 hidden layer nodes for the MLF method only. Users cannot adjust or select a cross-validation method (such as K-fold or holdout validation) or hyperparameters (such as learning rate), as these are automated by the software. The final report that is generated by NeuralTools 7.5 provides limited information, such as the number of trials. On the other hand, this tool is very useful for obtaining results within a few minutes. The developed model includes four variables: project number (k), complexity (x), phase (j), and contingency factor (i). These variables are considered the input of the ANN model. The model utilizes 20 projects, with the data from 16 projects collected (Table 4) to train and test the network. NeuralTools 7.5 defaults to selecting 80% of the data for training and 20% for testing randomly, which is kept with no change in this study. The other four projects, representing the four categories of complexities, are added to the model to predict their contingency percentages, which are the outputs of the model. Figure 5 illustrates the utilization of these twenty projects, which is limited to depicting the developed methodology and generalizing the final output. While twenty projects might not be enough to achieve highly accurate results, collecting data on construction projects is a challenging task. It requires a significant amount of time to gather and refine the necessary data for the currently developed model.

The project complexity (x), the second input of the network, is divided into four categories (L: low; M: medium; C: complicated; VC: very complicated complexity), according to Table 1. For the 13-story building (P17), it falls under Level 4 complexity, denoted as “Very complicated-Level 4”. The project phase (j) is the third input. It is divided into five phases: predesign, design, preconstruction, construction, and postconstruction. The fourth input addresses the contribution of each contingency factor (i) to different phases of the project, as depicted in

Table 5. Contingency-effect matrix.

No	Contingency Factors	Phase
		Predesign	Design	Preconstruction	Construction	Postconstruction
1	Site layout	0	1	1	1	0
2	Soil problems	0	1	1	1	0
3	Dispute with owner	0	0	1	1	1
4	Insurance	0	0	1	1	1
5	Weather	0	0	0	1	0
6	Design issues	0	1	1	1	0
7	Social impacts	0	0	1	1	1
8	Market fluctuation	0	0	1	1	0
9	Funding	1	1	1	1	1
10	Quality issues	0	0	0	1	1
11	Subcontractor changes	0	0	1	1	1
12	Human errors	0	1	1	1	0
13	Material	0	0	0	1	1
14	Environmental impacts	0	0	1	1	1
15	Low skills	0	1	1	1	0
16	Poor logistic	0	0	1	1	1
17	Warranty	0	0	0	1	1
18	Constructability problems	0	0	0	1	1
19	Conflicts	0	0	1	1	1
20	Accessibility	0	0	1	1	1
21	Safety	0	0	1	1	1
22	Permissions from authorities	1	1	1	1	1
23	Change in law	1	1	1	1	1
24	Incorrect scope	0	1	1	1	1
25	Machines	0	0	1	1	0
26	Security	0	0	1	1	0
27	Bankruptcies	1	1	1	1	1
28	Inflation	1	1	1	1	1
Total		5	11	23	28	19

, illustrating the Effect Matrix of the current project. For example, the site layout is a contingency factor that contributes to the design, preconstruction, and postconstruction phases, hence “1” is assigned to those phases. Conversely, the weather factor affects the project exclusively during the construction phase. This input pertains to the significance of the contingency factors.

Table 6. Contingency factor ranking based on questionnaire results.

No	Contingency Factors	1	2	3	4	5	Total Reviewers	RII
1	Site layout	2	6	12	14	24	58	0.77931
2	Soil problems	2	9	8	24	15	58	0.74138
3	Dispute with owner	3	7	15	18	15	58	0.72069
4	Insurance	8	6	7	18	19	58	0.71724
5	Weather	9	7	7	12	23	58	0.71379
6	Design issues	6	7	12	16	17	58	0.70690
7	Social impacts	8	8	9	12	21	58	0.70345
8	Market fluctuation	8	9	8	12	21	58	0.70000
9	Funding	2	11	18	12	15	58	0.69310
10	Quality issues	8	9	11	24	6	58	0.63793
11	Subcontractor changes	7	13	14	13	11	58	0.62759
12	Human errors	8	11	13	18	8	58	0.62414
13	Material	8	13	14	11	12	58	0.62069
14	Environmental impacts	11	12	14	8	13	58	0.60000
15	Low skills	6	13	22	11	6	58	0.59310
16	Poor logistic	18	19	5	7	9	58	0.49655
17	Warranty	12	18	20	6	2	58	0.48966
18	Constructability problems	15	13	22	6	2	58	0.48621
19	Conflicts	15	15	17	11	0	58	0.48276
20	Accessibility	13	23	14	5	3	58	0.46897
21	Safety	17	18	14	7	2	58	0.45862
22	Permissions from authorities	18	21	13	5	1	58	0.42759
23	Change in law	23	20	5	6	4	58	0.42069
24	Incorrect scope	31	12	6	6	3	58	0.38621
25	Machines	23	24	8	3	0	58	0.36897
26	Security	32	11	8	7	0	58	0.36552
27	Bankruptcies	32	14	10	2	0	58	0.33793
28	Inflation	32	20	3	2	1	58	0.32414

presents the Relative Importance Index (RII) of the 28 contingency factors obtained through a questionnaire survey involving 58 respondents. Notably, site layout emerges as the most significant contingency factor, with the highest RII value (0.77931), while inflation ranks lowest, with a contingency value of 0.32414.

These findings highlight the localized dynamics of the Saudi construction market, particularly in relation to turnkey residential projects. Using NeuralTools 7.5 to predict the contingency percentages of the 28 factors according to the project phase, the results are obtained as depicted in Figure 6. To implement the ANN model using NeuralTools 7.5, a dataset of twenty projects, totaling 1931 entries, is used for testing, training, and predicting, with distributions of 16%, 65%, and 19%, respectively. Figure 6 presents a portion of the predicted results, along with the test and training sets. For network training, the best search selects six configuration methods, detailed in

Table 7. NeuralTools 7.5 report.

Method	RMS Error	Training Time	Reason Training Stopped	Number of Trials	Training (% Bad Predictions)	Testing (% Bad Predictions)
GRNN	0.00	0:00:14	Auto-Stopped	108	5	25
MLFN 2 Nodes	0.00	0:03:00	Auto-Stopped	726,188	75	79
MLFN 3 Nodes	0.00	0:03:00	Auto-Stopped	587,146	76	78
MLFN 4 Nodes	0.00	0:03:00	Auto-Stopped	471,988	75	80
MLFN 5 Nodes	0.00	0:03:00	Auto-Stopped	374,796	75	77
MLFN 6 Nodes	0.00	0:03:00	Auto-Stopped	319,689	72	75

.

Accordingly, the NeuralTools report shows that the best configuration is GRNN, among other methods, due to the % of bad prediction of training and testing. The results of contingency percentages (pink color) of the four projects are directly filled into the empty cells of column “F”, as show in Figure 6, which will be elaborated upon in detail in the subsequent section. Columns “H” to “K” show the second output of the NeuralTool 7.5, identifying the type of each seat used in the ANN process, including the train, test, and predict sets, with the good/bad results of the test set.

5.2. Contingency Life-Cycle Analysis

The contingency percentage of the construction project, as per the results obtained from the preceding section, is calculated by aggregating the contingency percentages of all the factors corresponding to its complexity using Equation 1. Hence, the outcomes are classified into four categories. The contingency factor for residential complexity 1 amounts to 7.20%, for complexity 2 it is 8.16%, for complexity 3 it stands at 9.41%, and for complexity 4 it reaches 11.34%. These percentages can be categorized according to the project phase, as illustrated in

Table 8. Contingency according to complexity and project phase.

Complexity	Predesign	Design	Preconstruction	Construction	Postconstruction	Total
1	0.24%	0.52%	2.10%	3.82%	0.53%	7.20%
2	0.26%	0.35%	2.05%	4.71%	0.79%	8.16%
3	0.08%	0.78%	2.10%	5.38%	1.06%	9.41%
4	0.30%	0.99%	2.61%	6.33%	1.12%	11.34%

using Equation (1).

It is evident that the contingency during the construction phase is the highest among all phases, while the contingency during the predesign phase is the lowest. Figure 7 shows the cumulative contingency percentage based on complexity and project phase, offering a baseline for residential construction projects. In the presented case study, with a complexity level of 4 (13-story building), the estimated contingency is 11.34%. This amounts to SAR 9,072,000 (11.34% of SAR 80,000,000). This value is derived by considering 28 contingency factors, distributed across five project phases.

However, further analysis is warranted to address the uncertainty associated with this value. To delineate the range of uncertainty, @Risk software [30] is utilized, as illustrated in Figure 8. The output of the @Risk software is presented across three charts: (a) a simulation chart, (b) a percentile chart, and (c) a sensitivity chart. The input parameters include the project cost (SAR 80 million) and the contingency percentage for each factor per phase, determined from the first model. The output provides the total contingency of the project, amounting to SAR 9,072,000, as previously discussed. Running the simulation for 100,000 iterations with a ±5% error of the normal distribution, as depicted in Figure 8a, reveals a contingency range of SAR 8.4 million–SAR 9.8 million. This broad range underscores the uncertainty associated with the 28 factors throughout the project’s lifecycle. Figure 8b illustrates the percentile distribution of the contingency value, showing that the range below SAR 8.8 million and above SAR 9.4 million approaches zero. Figure 8c showcases the sensitivity analysis of the top ten contingency factors for each project phase. It reveals that the funding during construction is the most sensitive factor, followed by soil problems during the construction phase, and soil problems during the preconstruction phase, which ranks third.

Finally, Figure 9 depicts the contingency baseline of the current case study, divided into the project’s five phases. The baseline is an important tool for monitoring and controlling the project throughout its life cycle.

6. Findings

This study emphasizes the crucial role of accurately estimating and analyzing contingency values in residential turnkey construction projects in Saudi Arabia. An Artificial Neural Network (ANN) model provides a sophisticated, yet practical solution for estimating contingency values, considering project complexity and impact. The estimated contingency values for various project complexities serve as benchmarks for contractors, who anticipate and manage project risks, enhancing project planning and budgeting. Numerous current practices in Saudi Arabia, as cited in [1,2], could benefit from the findings of this study to update the contingency values for their ongoing and future construction projects.

7. Originality/Value

It underscores the uncertainties surrounding additional expenditures for contractors throughout the project lifecycle by focusing on the estimation and analysis of contingency values in residential turnkey projects in Saudi Arabia. This paper introduces an Artificial Neural Network (ANN) model to estimate the contingency value across the project lifecycle, factoring in elements like project complexity and contingency impact. The resulting insights provide valuable guidance for contractors involved in Saudi construction projects, especially those pertaining to residential developments. Additionally, this study introduces a new concept, termed “contingency baseline”, which has the potential to enrich the field of project management. The contingency baseline, determined according to the project phase, provides an innovative tool for monitoring and controlling the contingency funds of construction projects.

8. Project Validation

To validate the results of the current study, two approaches have been employed: comparison with historical data and peer review. The first approach was utilized to assess the estimated contingency’s alignment with real-world outcomes, based on project complexity. The results of the current study were compared with those of seven construction companies engaged in turnkey residential construction projects in Saudi Arabia. Figure 10 illustrates this comparison, juxtaposing the findings of the current study with those of the aforementioned companies. The results of the current study indicate that contingency increases with complexity, attributable to factors such as safety considerations and the utilization of heavy equipment such as cranes, as explained in Table 1. Additionally, the variance in contingency across projects decreases as complexity rises. While this study’s results lean towards a more conservative estimate compared to other projects, they still fall within the range of the real projects. However, further data are necessary to enhance the network’s training and improve contingency prediction. The second approach was chosen to validate the estimated contingency across the project’s phases. This method was adopted because the developed model is novel and has not been previously employed. Peer review feedback is valuable in identifying potential limitations, biases, or areas for improvement in the study’s methodology and the interpretation of the results. Forty reviewers, who are specialists in construction management and engineering within Saudi Arabia’s construction sector, were selected to review the final results of the developed model.

Table 9. Peer review of contingency results across project phases.

Decision	Predesign	Design	Preconstruction	Construction	Postconstruction	Total
Agree	23 (58%)	25 (62%)	30 (74%)	32 (80%)	28 (70%)	138 (69%)
Not sure	8 (20%)	7 (18%)	5 (13%)	3 (8%)	9 (22%)	32 (16%)
Disagree	9 (22%)	8 (20%)	5 (13%)	5 (12%)	3 (8%)	30 (15%)
Total (%)	40 (100%)	40 (100%)	40 (100%)	40 (100%)	40 (100%)	200 (100%)

presents the peer review results, indicating that 69% of the reviewers agree with the findings of the current study, while 16% express uncertainty about the obtained results, and 15% disagree. Among the project phases, the predesign phase has the lowest agreement percentage (58%), likely because many risk factors are still unpredictable at this early stage. Conversely, the construction phase has the highest agreement percentage (80%), as most of the anticipated risks are identified during this stage.

9. Practical Applications

The estimation and analysis of contingency values for residential turnkey projects in Saudi Arabia offer significant practical and managerial implications, enabling contractors to effectively manage projects. Firstly, it enhances risk management for contractors and project stakeholders by fostering a deeper understanding of the inherent risks in turnkey construction projects, particularly within the residential sector. By systematically identifying and analyzing known risks throughout the project lifecycle, stakeholders can refine their risk management strategies to mitigate the potential negative impacts on project outcomes. Secondly, informed decision-making regarding the estimation and analysis of contingency values, utilizing advanced techniques, provides valuable insights into the additional costs required to effectively manage project risks. This facilitates informed decision-making at various stages of project development, from inception to completion. Thirdly, the integration of the contingency percentages determined for different project complexities into the cost control and budgeting processes offers contractors a reliable basis for accurate cost estimation and budget allocation. By incorporating these contingency values into their financial plans, contractors can exercise better control over project costs and ensure adequate resource allocation to address potential risks as they arise. Furthermore, considerations such as improved project planning, competitive advantage, and stakeholder confidence are crucial for maximizing stakeholder benefits and minimizing risks throughout the project lifecycle.

10. Limitations and Future Work

(1)

Data availability: The accuracy and reliability of the estimated contingency values rely heavily on the availability and quality of historical project data. Limited access to comprehensive project data may restrict the generalizability of the findings and the effectiveness of the ANN model in estimating contingency values accurately. The current model relies on only twenty construction projects: sixteen for testing and training and four for predicting. These data might not be enough for machine-learning models, which are typically designed for much larger datasets. A small dataset may result in the undertraining of the machine-learning model. Accordingly, future research stemming from this study might concentrate on integrating data from additional construction projects to enhance the ANN model, thereby improving the accuracy of the contingency value estimates.

(2)

Generalizability: The findings of this study are specific to the residential turnkey projects in Saudi Arabia and may not directly apply to other construction project types. Future research should investigate the applicability of the proposed methodology in diverse geographic locations and construction sectors.

(3)

Model refinement: Further refinement of the ANN model is essential to enhance its predictive accuracy and usability. Future research may involve optimizing the model parameters, exploring alternative machine-learning algorithms, and incorporating additional input variables to capture a broader range of project characteristics.

(4)

Questionnaire results: The applications of the developed models involved a survey partially based on questionnaires, which has inherent limitations that can affect the quality and reliability of the results. Validation, as described in Section 8, was conducted to mitigate these limitations. Future work could reduce the reliance on questionnaire surveys by gathering more specific data from construction projects to validate the results. Additionally, a comprehensive database is required to train the developed ANN model effectively.

11. Conclusions

This study addresses the gap in current practices for accurately estimating the contingency value of turnkey residential buildings in Saudi Arabia, specifically in terms of project phases and the development of a contingency baseline. The methodology developed through this research provides construction contractors with a new management tool to monitor and control contingency values throughout the project’s lifecycle, particularly when implementing the turnkey delivery method. The utilization of a turnkey approach presents significant risks, particularly due to the uncertainty surrounding additional expenditures from project owners. Throughout the project lifecycle, both known and unforeseen risks pose challenges that necessitate careful consideration and proactive mitigation strategies. The development of an Artificial Neural Network (ANN) model offers a sophisticated, yet practical solution for estimating the contingency values for each project phase. By incorporating project complexity, contingency factor importance level, and contingency impact during project phases, the model provides detailed insights into the distribution of contingency percentages across different project stages. The estimated contingency values obtained through this model are 7.581% for low complexity, 8.161% for medium complexity, 9.785% for complicated, and 10.466% for very complicated projects, which serve as invaluable benchmarks for contractors engaged in residential construction projects in Saudi Arabia. These findings enable contractors to better anticipate and prepare for project risks, thereby enhancing project planning, budgeting, and risk management practices. Overall, the methodology proposed in this study contributes to advancing the field of contingency estimation in construction projects, offering practical insights and tools to support informed decision-making and mitigate project risks effectively. Further research and validation efforts are recommended to enhance the robustness and applicability of the proposed methodology in diverse construction contexts and geographic regions.