Estimation of ideal construction duration in tender preparation stage for housing projects

Construction projects are considered successful if they are completed on time and satisfy other criteria such as cost, quality and stakeholder satisfaction (Chan and Kumaraswamy 2002; Majid 2006). The ideal construction duration must be estimated to avoid common delays, increase the efficiency of organisations and benchmark the execution of projects in the construction industry (Chan and Kumaraswamy 2002; Mensah et al. 2016). Effective planning and scheduling prevent delays, enhance construction project performance, and yield time and cost savings (Gibson et al. 2006; Luu et al. 2009; Ismail 2013; Lines et al. 2014; Tunç and Özsaraç 2015), whereas poor planning and scheduling negatively affect clients, contractors and owners (Ndekugri et al. 2008). However, few studies have been performed on planning and scheduling to estimate the construction duration during the tender preparation stage.

Time planning in the tender preparation stage is regulated by relevant laws and authorised organisations (International Federation of Consulting Engineers [FIDIC] 2017). However, a different state of affairs is observed in the public construction industry in Turkey (Yitmen and Dikbaş 2002; Birgonul et al. 2007; Erbaş and Çıracı 2013; Akkaynak 2014; Usta 2014), wherein legal regulations disregard planning and prioritize only the lowest tendered price (Türesoy 1989; Karapinar 2005; Tokalakoglu 2010; Kaplan 2012). Lower costs should not be the only basis for selecting contractors; time and performance factors should also be considered (Obodo et al. 2021). Moreover, better tender offer systems and effective plans focusing on development have not been implemented (DPT 2001, 2007; Turkish Republic Presidency of Strategy and Budget 2019). Although stakeholders in the construction industry emphasise construction duration, many contractors do not complete projects’ work within the specified deadlines (Lin et al. 2011). Evidently, the on-time completion of construction projects is a prevalent issue, and projects in Turkey lack time planning.

In Turkey, the most significant factor affecting housing-project duration is cost (Odabasi 2009; Baltaci 2012). Arditi et al. (1985) surveyed public agencies and contractors involved in public construction projects in Turkey and classified the causes of delays as ‘those influenced by national policies’ and ‘those that can be controlled by public agencies and contractors’. They found that the overall state economy influenced the construction duration. Sonmez (2019) emphasized that the total construction duration that affected the cost of housing projects could not be calculated using a single mathematical equation and that time-planning failures depended on the characteristics of the housing project. In practice, most methods for estimating construction duration depend on the ‘subjective skill and cognition of the estimators and planners, rather than on objective assessment’ (Lin et al. 2011). In Turkey, the Supreme Court adjudicates many legal disputes concerning housing-project delays (TOKI v Beneficiary1 2010, 2011, 2013, 2014, 2015). This has also been indicated by previous studies (Al-Khalil and Al-Ghafly 1999; Aibinu and Jagboro 2002; Odabasi 2009), suggesting that clients use modern planning and scheduling techniques in place of data obtained from previously completed housing projects. Therefore, a reliable estimation tool is needed owing to the complex nature of housing projects.

As mentioned previously, methods for calculating the construction duration during the tender preparation stage of public housing projects have not been extensively studied. Similar to other developing countries such as Nigeria (Mansfield et al. 1994; Elinwa and Joshua 2001; Ubani et al. 2013), Saudi Arabia (Bin Seddeeq et al. 2019; Alshihri et al. 2022), Malaysia (Endut et al. 2009; Memon 2014) and Indonesia (Kaming et al. 1997; Susanti 2020), Turkey suffers from construction delays and poor construction planning. Therefore, in this study, a novel method is developed to estimate the ideal construction duration and prevent delays in public housing projects by considering the influencing factors and evaluation criteria. The calculation method is compatible with the existing bidding system and is user-friendly. Therefore, it can be used for ensuring time saving as well as cost reduction. The method is based on data collected from construction authorities regarding construction duration in Turkey. The findings of this study are expected to enable senior project managers to estimate the ideal construction duration for housing projects during the tender preparation stage.

In Section 1, we review previous studies. In Section 2, we examine factors affecting construction duration by surveying the literature and identify the selected factors for the proposed model. Section 3 presents the research methodology; we explain how the relevant data were collected and describe the statistical methods used for analysis. Section 4 presents the results of the statistical analysis for our calculation method to obtain ideal construction durations for public housing projects. The final sections present a discussion and conclusions.

The completion of housing projects within the specified construction duration is one of the main criteria for evaluating the performance of a construction company (Bromilow 1969; Chan and Chan 2004; Aibinu and Odeyinka 2006; Ting et al. 2021). However, delays in housing projects remain a common problem (Lin et al. 2011). Studies have indicated that only 12.5% of building contracts were completed within the scheduled periods, and the average completion time was approximately 40% longer than the contract duration (Bromilow 1969). Moreover, 40%–70% of these projects deviated from the original schedule, and some incurred delays that lasted for months (Al-Khalil and Al-Ghafly 1999; Blyth et al. 2004; Iyer and Jha 2005).

The duration of a construction project is often overlooked owing to incorrect assumptions or comparisons, and it is easily evaluated during the early stages of the project (Elinwa and Joshua 2001; Walraven and de Vries 2009; Ibironke and Elamah 2011; Oyedele 2013; Oyedele et al. 2015, 2021; Shokri-Ghasabeh and Chileshe 2016). In the tender-issue stage of a construction project, stakeholders seek a reliable estimation of the project duration (Qiao et al. 2019). The project duration must be estimated before beginning a project to complete the project in a timely manner (Thing 2006). Underestimation of the project duration may lead to disputes between the contractor and owner, and overestimation may reduce competitiveness during the tender-issue stage (Jin et al. 2016). Studies have indicated that planning and scheduling during the early stages of construction projects significantly impact the final project outcomes (Wang et al. 2012).

Various factors affect construction duration (Oo et al. 2022). Several factors that affect the construction duration at the pre-construction stage, as reported in the literature, are presented in Table 1.

Although project duration significantly affects project costs, the dependence of the cost on the complexity and size of projects, rate of return, cash flow, type of contract, priority of projects, previous experiences of the contractor, and geographical area are given more importance worldwide (Ahmad and Minkarah 1988; Shash 1993; Fayek et al. 1998, 1999; Hillebrandt 2000; Bageis and Fortune 2009; Jarkas et al. 2013; Alsaedi et al. 2019). Previous studies on estimating construction duration have mainly focused on the financial, economic (Türesoy 1989; Alaghbari et al. 2005; Shane et al. 2009; Musarat et al. 2021), climatic, geographical and topographic factors (Elhag et al. 2005; Shane et al. 2009; Cheng 2014;). Other factors that have been considered are the complexity and size of projects (Chevallier and Russell 2001), project priorities (Yang et al. 2014), supply and logistic conditions of the project region (Asnaashari et al. 2009; Ramli et al. 2018; Tunji-Olayeni et al. 2018; Nayak 2019) and social and cultural factors (Imbert 1990; Assaf and Al-Hejji 2006; Salleh 2009; Al-Sabah et al. 2014). Relative to the several factors considered in various studies, construction duration is mostly ignored.

Several methods have been developed for estimating construction duration according to these factors. Many studies have demonstrated the applicability of regression analysis for estimating the duration in the early phases of a project (Khosrowshahi and Kaka 1996; Lin et al. 2011). Simulation models have also been used to estimate construction duration. Sanni-Anibire et al. (2021) developed a machine-learning model that involved multilinear regression analysis (MLRA), k-nearest neighbours (KNNs), an artificial neural network (ANN), a support vector machine (SVM) and ensemble methods for tall-building projects; the most crucial factor affecting construction duration was the total number of floors. Fan et al. (2021) used an ANN to estimate the construction duration in the preliminary stage and found that accurate durations could be obtained by using two-stage ANNs and feature selection via sensitivity analysis. Yogesh and Rao (2021) developed a system to estimate road construction duration using individual activity output rates and Delphi analysis. Yaseen et al. (2020) developed an accurate method using a hybrid artificial-intelligence model to estimate construction duration and monitor risk levels. Lines et al. (2014) demonstrated that a scheduling model used during the tender preparation stage could realise cost and time reductions of 44.0% and 44.9%, respectively. In another study, Lines et al. (2015) proposed a planning model for the tender preparation stage that reduced the cost and duration by 54% and 70%, respectively. Decision-tree algorithms, e.g. classification and regression tree (CART) and chi-squared automatic interaction detection (CHAID), are also used for estimating duration (Godinho and Costa 2004). However, there have been few studies on the use of the CART and CHAID algorithms in the construction industry. Lin and Fan (2019) used CHAID and CART to identify defects in public construction projects and reduce adverse delays. Pospieszny (2015) used CHAID to estimate the effort and duration required for software projects. Papatheocharous and Andreau (2012) developed a hybrid software cost estimation approach using CHAID and CART.

The time performance can be enhanced by selecting appropriate factors to reduce delays in construction projects. The construction duration should be estimated at the tender preparation and initial planning stages such that stakeholders can prevent potential disputes, as well as time and cost losses.

Delay factors hinder construction activities during construction and consequently affect the project completion time (O’Brien and Plotnick 1999). Research published over the past few decades has identified numerous factors that generate delays in construction projects. The delay factors encountered by many construction projects as reported by previous studies were remarkably similar (Doloi et al. 2012). Table 2 presents 56 factors that affect the construction duration; these factors were collected through a literature review.

Among the 56 factors presented in Table 2, the two FM factors were eliminated, and the remaining 54 factors were categorised into factors associated with the tender stage and those related to the construction stage. Since the proposed calculation method is to be used in the tender stage, only the 21 factors belonging to the tender-stage category were considered. The 54 factors were also categorised as factors attributable to the employer and factors attributable to the contractor, and only 29 factors belonging to the former were selected. In total, there were 18 factors belonging to both the tender stage and attributable to the employer categories, which were then considered for calculating the ideal construction duration. Next, these factors were examined individually with regard to their eligibility and applicability to statistical methods, and three factors were finally selected.

Eight factors used by construction authorities were selected, and their validity was crosschecked through a literature review to confirm that they were consistent with the selection criteria. These eight factors were the number of flats, number of working days with a schematic design or construction documents, number of non-working days, priority of the project, complexity of the project, special request for the project, logistic conditions of the project region and climatic conditions of the project region. The first three factors (number of flats, number of working days with a schematic design or construction documents, and number of non-working days) are combined and considered as one factor: the baseline construction duration (BCD).

The formula used by construction authorities to estimate the total construction duration for public housing projects is given below. 1 BCD(days)=(numberofworkingdaysinFactor#2acorrespondingtoFactor#1)+(50workingdaysincaseofuseofFactor#2b)+(Factor#3) \[\text{BCD}\,\left( \text{days} \right)=\left( \text{number}\,\text{of}\,\text{working}\,\text{days}\,\text{in}\,\text{Factor}\,\#2\text{a}\,\text{corresponding}\,\text{to}\,\text{Factor}\,\text{ }\!\!\#\!\!\text{ 1} \right)+\left( 50\,\text{working}\,\text{days}\,\text{in}\,\text{case}\,\text{of}\,\text{use}\,\text{of}\,\text{Factor}\,\#2\text{b} \right)+\left( \text{Factor}\,\#3 \right)\]

Here, Factors #1, #2a, #2b and #3 represent the number of flats, duration with construction documents, duration with a schematic design, and number of non-working days, respectively.

If the number of flats and construction documents are available in the tender stage, the corresponding duration with construction documents can be determined. The durations corresponding to ranges of the number of flats are as follows: 400 days for 0–250 flats, 500 days for 250–750 flats, 550 days for 750–1,250 flats, and 600 days for 1,250 or more flats (for example, the duration for a public construction project with 850 flats is 550 days). If a project uses a schematic design, 50 days are added to its duration with construction documents. Additionally, the number of non-worked days is added to the construction duration. The list of non-worked days by province was obtained from the Ministry of Environment, Urbanization, and Climate Change.

Other factors used by construction authorities are the priority of the project, complexity of the project, special request for the project, logistic conditions of the project region, and climatic conditions of the project region, which were shown to be effective for calculating the construction duration in various studies. In total, 11 key factors were selected as variables for the calculation method, and they are presented in Table 3.

To quantify the effect of each variable on the construction duration, evaluation criteria values were used. For the first eight factors, the evaluation criteria values were set as ‘0, 1, and 2’, where 0 indicates that the factor had no effect on the duration, and 1 and 2 indicate that the factor tended to reduce and increase the duration, respectively. For F9 and F10, the evaluation criteria values were set as numbers ranging from 1 to 7; a larger number corresponded to a more significant effect of the factor on the duration. For F11, the evaluation criteria values were coded with numbers ranging from 1 to 5; a larger number corresponded to a less significant effect of the factor on the duration.

A quantitative methodology was adopted in this study. The research method and procedure included two stages in line with the research objectives. In the first stage of the study, the statistical calculation methods were developed and validated, and in the second stage, these methods were tested and implemented.

In project management, one of the most difficult tasks is to estimate the ideal project duration. Our results indicated that construction delays can be reduced using the proposed calculation method, as suggested by previous studies concerning related methods. A similar study by Lin and Fan (2019) demonstrated that the CHAID and CART methods exhibit good accuracy for predicting defects in public construction projects, which have an indirect effect on completion time in that they result in a prolongation of the desired completion time. Papatheocharous and Andreau (2012) developed software to estimate project costs using CHAID and CART. In both of these papers, alternatives to traditional techniques such as regression were proposed, along with the use of analogies considering the problematic nature of public project data. However, in these studies, larger datasets were needed to evaluate the accuracies of the CHAID and CART methods. Moreover, no research has been conducted on the use of CHAID and CART to estimate the construction duration of public housing projects. Studies have been performed on alternative methods for project duration evaluation in the construction industry. For example, the linear scheduling method and Delphi process were used to assess highway building projects, and it was concluded that various risks and the production rates of activities are important for estimating project duration (Yogesh and Rao 2021). Yeom et al. (2018) used MLRA to estimate the project duration for office buildings in the planning phase. Their model provided accurate results with regard to duration prediction, as well as ease of use for owners and other stakeholders. It is important that these models present data and results in a simple and accurate manner such that the users can easily understand how and which data should be used. In other studies focussing on the pre-planning phase of construction projects, similar results were obtained using different methodologies. The use of AAN and sensitivity analysis exhibited good accuracy for estimating construction duration considering the simulation of complex behaviours, which is a limitation for regression analysis and other traditional methods such as standard-curve models (Chao and Chien 2010; Fan et al. 2021). Ujong et al. (2022) showed that the duration prediction performance of the ANN model is better than that of MLRA for buildings. However, ANN has limitations for determining various control features considering input–output processes, and the construction duration can be determined by considering critical activities rather than via the summation of all activities (Fan et al. 2021). Additionally, inaccurate training data can affect the predictive accuracy of and distort ANN models during their development, as they are data-driven models (Adul-Hamid 1996). Conversely, an action-based research methodology was used by Lines et al. (2014) to develop a scheduling model for the tender stage, and they showed that their model could reduce cost and time overruns by up to 44.0% and 44.9%, respectively. Similar findings were obtained by Lines et al. (2015) for construction cost and duration.

Regression analysis evaluates the significance of the independent variables numerically rather than categorically. In contrast, the CHAID and CART methods introduce variables as decision trees instead of equations. Regression analysis is used to predict continuous outcomes, while CHAID and CART are used for classification and segmentation tasks. The choice of technique depends on the research question and the type of data being analysed. The type of data for housing projects in this study involve both continuous and categorical data. For example, a categorical ranking was made by considering the values in the ranking of the geographical regions in Turkey according to the number of rainy days (precipitation). As an evaluation criterion for Climatic Conditions of Project Region (F10), ‘1’ represents the geographical region that receives the least amount of precipitation and ‘7’ represents the highest precipitation. Likewise, F9 and F11 are categorical variables that should be evaluated by classification methods such as decision trees. Overall, to improve the evaluation method used by the construction authority, involving calculation of the ideal construction duration by means of an accurate equation, regression would be needed. On the other hand, CHAID and CART are preferred when the need is felt for an accurate computation of the ideal construction duration, owing to categorical data in the factors affecting housing projects, such as F9, F10 and F11. These three methods were used to estimate the ideal duration to determine the optimal solution. Therefore, the cons of regression analysis for categorical variables is fixed by decision trees while the need for an equation is resolved by regression.

The goodness of fit of a regression model is measured by varying R² values between 0 and 1. It is difficult to suggest a rule for how appropriate R² is due to its varying value by research area. For example, R² values of 0.90 and higher are common in longitudinal studies while values around 0.30 are common in cross-sectional designs, and R² values around 0.10 are acceptable for exploratory research using cross-sectional data (Mooi and Sarstedt 2011). Since the study was exploratory and cross-sectional data were used, the R² value of 0.356 can be considered acceptable. In addition, since a large number of independent variables, such as the six variables employed in the present study, were used in the regression analysis, it is an expected result that the deviations increase and the R² value decreases due to the numerator being ‘the regression sum of the variance of squares’, while the denominator the ‘total sum of the variance of squares’ in the R² equation (Lewis-Beck 1980; Hagquist and Stenbeck 1998). The CHAID and CART results were validated using 10-fold cross-validation, with 70 training sets and 30 test sets, and standard error values were derived for the estimations. The standard deviation increases and an overtraining problem can occur when 90%/10% is used for the training and data set (Geng et al. 2015). Besides avoiding this problem, studies also suggested that 70%/30% was found to have the highest classification success compared to 60%/40% and 75%/25% (Koc and Ulucan 2016; Aksoy and Boztosun 2021).

The objective of this study was to develop a novel calculation method for the ideal construction duration. The factors used by construction authorities to determine the BCD were proven to be insufficient. Additional factors from the literature were needed to achieve more accurate duration predictions. Using the proposed calculation method reduced the number of delayed public housing projects and associated delay times. Therefore, the results indicated that the proposed calculation method was useful and valid, and the objective of the study was achieved. This method can also be used to determine the ideal construction duration, which can ensure the on-time completion of public housing projects and prevent delays. This will improve cost management and result in fewer disputes among stakeholders.

The results of a statistical analysis indicated that all three investigated statistical methods were valid. The regression method yielded results that were more accurate than those of the CHAID and CART methods. Therefore, in future studies, the ideal construction duration based on six identified factors can be predicted using the proposed regression method (Table 9). Although the CHAID and CART methods exhibited lower performance than the regression method, they required fewer factors to estimate the ideal construction duration. Therefore, it was proven that the ideal construction duration could be calculated using all three methods. All factors that exhibit significance in any of the investigated statistical methods should be considered in the estimation of the ideal construction duration.

This study had three limitations. First, the proposed calculation method involves only public housing projects because the majority of Turkish public construction projects are public housing projects, which incur major delays. Second, the proposed method is limited by the accuracy of public construction project data, which may influence its predictive accuracy. Third, key factors identified as having significant impacts on the construction duration in previous studies were only partially included because many factors are not related to the pre-construction (procurement) stage. Therefore, only the factors affecting the construction duration at the procurement stage were considered in this study.

We developed a practical and consistent project management tool using the proposed calculation method. It can be used to prevent problems during the tender stage and lead to fewer risks during the construction stage. In future research, the proposed calculation method can be applied to a digital environment and converted into computer software to ensure ready availability and that a greater number of users have access to it, as well as accommodate international users and authorities such that it can be utilised worldwide.

Raw data were generated at TOKI (Housing Development Administration of the Republic of Turkey). Derived data supporting the findings of this study are available from the corresponding author on request.