Productivity Metrics and Its Implementations in Construction Projects: A Case Study of Singapore

: Although some studies have used or developed different types of metrics to assess construction productivity in the existing literature, few of them investigated those metrics systematically and the differences between assessment results. This study examined the various types of metrics used in the assessment of the productivity of construction projects. First, a literature review was conducted ﬁrst to identify prevailing productivity metrics at four levels, namely trade, project, company, and industry. Then, the questionnaire was developed and disseminated to 53 Singapore-based construction companies for data collection. Subsequently, non-parametric statistical tests were conducted to analyze the data collected by the questionnaire. Results showed that the top ﬁve metrics in terms of usage frequency and relative importance were “constructability score”, “buildable design score”, “square meter of built-up ﬂoor area per man-day”, “square meter per dollar”, and “output per worker.” In addition, results showed that differences existed in the assessment results when productivity metrics at different levels were used to conduct the same measurement. This is the ﬁrst study to explore the most widely used metrics in productivity assessments of construction projects and investigate possible differences in assessment results. This study could help the authorities to review, evaluate, and modify the productivity metrics used in practice. Thus, this study is beneﬁcial to the practice as well.


Introduction
The construction industry is widely reported to have a relatively low productivity level, especially when comparing with other industrial sectors such as mining, manufacturing, and oil and gas [1]. This is mainly because the construction industry is labor-intensive and is deeply affected by various internal and external factors including social culture, environmental and legal constraints, inappropriate management actions, extreme weather, excessive overtime, and transportation conditions [2][3][4][5].
Productivity must be assessed before it can be improved. The assessment of productivity cannot be carried out without appropriate metrics. In the current literature, several researchers have used different metrics to assess productivity. For example, Shehata and El-Gohary [6] used capital productivity, output per worker, and output per work hour as indicators to assess construction labor productivity. Vogl and Abdel-Wahab [7] suggested using total factor productivity, output per worker, and output per work hour to measure the construction productivity performance at the industry level. Ayele and Fayek [8] developed a framework to measure the total productivity of construction projects, which consists of metrics at three levels: activity, project, and industry, and proposed a metric for measuring the total productivity of construction projects, which considers all resources used, including labor, material, capital, energy, construction project indirect cost, and owner cost. Huang et al. [9] stated that productivity can be measured at three levels: task, project, and industry, 1.
What metrics are used to measure construction productivity in Singapore? 2.
Are those metrics frequently used in practice? 3.
Will there be any significant differences in productivity measurement results when different metrics are used to measure the same project?
The research efforts described in the paper were carried out in the context of Singapore. Singapore is an island country featured in tiny size but developed economy. The population of the country is limited making the country has no choice but to seek for a higher productivity. To improve construction productivity in Singapore, the governmental department Building Construction Authority (BCA) has announced several initiatives in recent years. For example, BCA launched the 2nd Construction Productivity Roadmap in 2015 which encourages using advanced technologies such as Building Information Modelling (BIM), Prefabricated Prefinished Volumetric Construction (PPVC), and 3D printing to improve construction productivity [11,12]. In addition to the efforts made by the government, the academia has also carried out a series of research attempting to improve the construction productivity in Singapore. Ofori [13] investigated the barriers to achieving high construction productivity in Singapore. Ofori [14] examined the initiatives that would enable contractors in Singapore to improve the level of construction productivity. Lin et al. [15] proposed a framework for developing a productivity and safety monitoring system using Building Information Modelling. Hwang et al. [16] identified the critical factors affecting the productivity of green building construction projects in Singapore. Hwang et al. [17] identified and prioritized the critical management strategies that can help improve productivity in the construction industry of Singapore. Although great efforts are made both by the practice and academia, the productivity of Singapore's construction industry is still low [18,19]. Hence, it is imperative to review and evaluate the practices conducted for improving construction productivity in Singapore. Moreover, according to the World Population Review [20]. Singapore is the second most productive country in Asia, which merely ranks behind Japan. A lot of countries, particularly for the emerging economies in Asia, are following the practice of Singapore for seeking productivity enhancement. For example, China chosen Singapore as its model and privileged partner [21]. Vietnam is learning from Singapore's experiences in preparing for a productive workforce targeting the Industrial Revolution 4.0 [22]. India is investigating the model of Singapore trying to secure its tremendous growth and economic success [23]. Thus, research findings generated in Singapore would be useful, informative, and generalizable to the emerging economies worldwide. Therefore, Singapore is a good context to carry out the piece of the research.
Although some studies in the existing literature have used or developed different types of metrics to assess construction productivity, few of them investigated the implementation of these metrics. This study revealed the implementation of different productivity metrics in Singapore by checking the usage frequency, relative importance, and possible differences in the assessment results of various productivity metrics. Thus, this study can contribute to the current body of knowledge on productivity. Meanwhile, this study can help achieve further productivity growth in the industry by enhancing the government and the construction companies' understandings of productivity metrics and updating the corresponding rules and policies. Therefore, this study is also beneficial for the industry. Moreover, this study can be replicated in other countries, which will allow government and construction industry managers to understand how local productivity metrics are implemented in real projects. Therefore, it is believed that this study can make contribution to the global construction community.

Productivity
Generally, there are two types of productivity, namely single factor productivity, and multi-factor or total-factor productivity. Single factor productivity, which includes labor productivity and capital productivity, relates to just one input factor, while multi-factor or total-factor productivity takes into account all the inputs [7]. Labor productivity is a singlefactor productivity often expressed as output per worker or output per hour worked [6,7,10]. The Construction Industry Institute (CII) [24] attempted to propose definitions of productivity for consistency, where productivity refers to "work-hours performed per units of work completed", and "the ratio of planned productivity to actual productivity" refers to productivity index. Another single-factor productivity is capital productivity. Capital productivity can be defined as a percentage return on capital invested [25]. Li and Liu [26] stated that it can provide an overall capital utilization level of the construction industry. Multi-factor productivity considers the contribution of both labor and capital as inputs [10]. Total factor productivity (TFP) takes into account a combination of inputs with adjustments for technological progress [27]. Carson and Abbott [28] also stated that Data Envelopment Analysis (DEA) is a linear programming technique that computes the ratio of total inputs employed to total output produced for each unit to estimate organizational efficiency.

Construction Productivity in Singapore
Construction is a mainstay of Singapore's economy. It accounts for 3-6% of Singapore's gross domestic product (GDP) and takes up an estimated 12% of the total workforce in the country. According to Bowen [29], the vitality of one country's economy is best indicated by its productivity performance. Therefore, improving the productivity level of the construction industry is important, critical, and imperative to the national development of Singapore. BCA has launched a series of initiatives in the past years to improve the construction productivity of Singapore. For example, in 2013 BCA initiated the Balcony Bonus Gross Floor Area Scheme which rewards developers who work closely with architects, engineers, and contractors towards higher construction productivity [30]. In addition, BCA operates a Certified Construction Productivity Professional Scheme which grants registered contractors additional bonus points for project bidding [31]. However, in the current construction industry of Singapore, there is a lack of systematic research on productivity metrics, making the industry companies unable to interpret their productivity level accurately, thereby leading to poor management and hindering the further improvement of the industry productivity.

Metrics Used to Measure Productivity in Construction Projects
The research team carried out a comprehensive literature review to identify the metrics that are used for construction productivity assessment. The identification of productivity metrics refers to the "PRISMA" guidelines [32]. Ten journal articles and two official reports that particularly addressed construction productivity metrics were identified. After examining the results of the literature review, the research team found that the prevailing productivity metrics can be grouped at four levels, namely, trade level, project level, company level, and industry level. Table 1 summarized the specific metrics used at four different levels. Amount of rebar/prefab mesh per manhour P23 [36] Volume of concrete per manhour P24 [36] Area of drywall per manhour P25 [36] Painted area per manhour P26 [36] Number of doors installed per manhour P27 [36] Wall tiled area per manhour P28 [36] Floor tiled area per manhour P29 [36] Suspended ceiling per manhour P30 [36] Length of ducting per manhour P31 [36] Length of electrical conduit per manhour P32 [34] [36] Length of water pipe per manhour P33 [36] At the trade level, productivity metrics involving measuring the output units for activities carried out at construction sites. Shehata and El-Gohary [6] mentioned that activity-oriented models are commonly adopted by contractors to measure productivity at construction sites. CII [24] suggested using work hours expended over quantity installed to measure activities' productivity on site. BCA [36] published a guidebook that sets out the best practices to measure productivity for 12 key trades which are commonly found in most construction projects. The 12 key trades include formwork installation, concrete placement, reinforcement placing and fixing, drywall installation, timber door installation, painting, wall tiling, air-con ducting installation, suspended ceiling installation, floor tiling, electrical conduit installation, and water pipe installation. A study by Singapore Contractors Association Ltd. (SCAL) also stated that construction firms used trade-level labor productivity that can be expressed as either unit of output per dollar, per work-hour, or per man-day to monitor site activity [10]. Therefore, when measuring productivity at construction sites, it can be measured by the work hours expended over quantity installed.
At the project level, metrics such as square meters of built-up floor area per man-day or square meters per dollar are used as a composite measure of labor productivity [10]. In addition, "output per person-hour on key trades", "total revenue per month", "square meters per man-day", "value-added per work" and "constructability scores", are also metrics that can be used to assess project productivity [10]. Shehata and El-Gohary [6] also used square feet per dollar as a form to assess project productivity. However, due to the unique nature of the construction projects, measuring productivity at the project level poses limitations for benchmarking of data as construction items are non-identical [10]. Therefore, when measuring productivity at the project level, it is important not only to consider quantitative metrics but also to adapt qualitative metrics in the light of the actual situation of the project to obtain appropriate results.
At the company level, metrics such as "gross output per worker" and "square meters per man-day" are used to assess productivity [10]. In addition, value-added productivity is also used in some instances in the context of Singapore. For example, value-added per worker or value-added per hour worked are used and are estimated from progress payments or renovation bank loans and data from Housing Development Board and Urban Redevelopment Authority or production volume [10]. Subsequently, using price indices to deflate nominal value to get the real value-added. However, the challenges facing by the metric involves the difficulties in understanding the concept of value-added and in obtaining accurate figure due to project complexity [10]. Therefore, if productivity is to be measured within a certain period, metrics of company level can be used, but it will lack accuracy when considering long-term productivity.
At the industry level, productivity metrics include labor productivity, multifactor productivity, total factor productivity as well as estimations using cost or production function such as the growth accounting approach or data envelopment analysis. Issues with metrics at the industry level are the measurement to use for output or value-added, and what measurements should be used use for labor input. Additionally, problems in finding appropriate deflators for data adjustments happen due to affected output and inputs data by business cycles [10]. Despite productivity metrics at the industry level have many advantages in theory compared with other levels, the value of inputs and outputs cannot be measured only by quantitative metrics in practice. Therefore, errors due to uncertainties should also be considered.

Methods
This study adopted data collection methods, including literature review, pilot interviews, and questionnaire. The data analysis includes frequency and importance analysis, alignment analysis, and difference analysis.

Literature Review
Literature review is an effective way for researchers to learn some specific areas [37]. To identify the productivity metrics used by the current construction industry, this study conducted a literature review. The search scope of the literature expands vastly, covering the relevant books, journal articles, government reports, authoritative documents, and information on websites.

Pilot Interview
Pilot interview with experienced industry expert is widely used in the construction engineering and management research to verify whether the information retrieved from the literature review is reasonable [38,39]. Therefore, this study carried out pilot interviews with three experts from the construction industry of Singapore, to verify the productivity metrics identified in the literature review, checking their applicability to the context of the country. The list of productivity metrics retrieved from the literature review was then used in a focus group meeting with three experts to check the applicability of the metrics to the Singapore industry. All three experts have at least 14 years of experience in the construction industry and at least 8 years of experience in assessing construction productivity. These experts are experienced enough to provide views and opinions that are correct and reliable. The experts did not add any new metrics but modified the descriptions of some metrics and provided suggestions in the discussion section. Table 2 presents the profiles of the post-interviewees.

Development of Questionnaire
Questionnaire is a widely used method for obtaining the opinions of industry professionals in construction engineering and management research [40]. This study used questionnaires to gather industry professionals' perceptions of productivity metrics. The questionnaire was developed based on the results of literature review and was composed of three sections. The first section recorded the respondent's profile including the designation, years of experience in construction industry, and years of experience in assessing construction productivity. The second section asked the respondents to assess the identified metrics in terms of their usage in frequency and relative importance, using two different five-point rating scales, namely, 1 = never, 2 = rarely, 3 = sometimes, 4 = often, 5 = always, and 1 = not important at all, 2 = low important, 3 = moderate, 4 = important, 5 = extremely important. The third section of the questionnaire requested the respondents to assess the productivity of one project they have participated in, based on the identified metrics. During the assessment, another five-point rating scale, namely, 1 = very poor, 2 = poor, 3 = moderate, 4 = good, and 5 = very good, were used. A sample of the questionnaire has been attached as Appendix A.
The sampling frame of the questionnaire covered 4168 contractors and licensed builders registered under the BCA Directory [41]. Referring to a frequently used sample rate of 10% [42], 416 contractors were randomly selected from the BCA Directory as potential respondents of the questionnaire, which was distributed via email, phone calls, and in person. A total of 53 responses were received, yielding a low response rate of 12.7%, but for the survey research of Singapore's construction industry, this response rate is acceptable [43,44]. Such a low response rate is understandable, because construction industry is always full of stress and challenges, and that construction professionals within a pressured and dynamic work environment are unlikely to respond in the same way as members of the public receiving questionnaires in their own homes or others who may work in more stable work environments (e.g., office workers in large organizations) [45]. The data set was checked for outliers, and the result showed that the data of the 53 responses were all valid. The respondents had different types of designations, such as contractors, consultants, owners, and government agency personnel. Among the respondents, approximately 90% had at least five years of working experience in the construction industry and approximately 50% had at least five years of experience in assessing construction productivity. These proportions indicated that the questionnaire respondents were experienced in the field and that data collected were reliable. Table 3 presents the profiles of the survey respondents.

Data Analysis Methods
Statistical tests were conducted to analyze the data collected by the questionnaire. Although the sample size of this study (i.e., 53) was not large, statistical analysis could still be performed because the central limit theorem holds true as long as the sample size is no less than 30 [46,47]. Two types of tests are widely used for statistical analysis, parametric statistical tests and non-parametric statistical tests [48]. For parametric statistical tests, they have a specific requirement which is data to be analyzed must fall in normal distribution, while non-parametric statistical tests have no such constraint [49]. Based on the experience of general questionnaire survey research in the construction industry, this study uses non-parametric statistical tests to analyze the data. All tests were conducted with the aid of SPSS Statistics 26.0.

Identification of the Most Frequently Used and Most Important Metrics
One sample Wilcoxon signed-rank test was used for the identification of the metrics that are most frequently used and most important. One sample Wilcoxon signed-rank test is a non-parametric statistical method, which checks whether the median of the sample data is different from a test value [50,51]. According to [52], the null hypothesis of the test is that the median of the sample data is statistically same as the test value. If the p-value generated by the test is greater than 0.05, the null hypothesis is supported. By contrast, a p-value less than 0.05 indicates the median of the sample data is statistically different from the test value. For this study, "4" was selected as the test value, which means the metric is frequently used and important according to the five-point Likert scale. Thus, by conducting the One sample Wilcoxon signed-rank test using the test value of 4, the metrics that are most frequently used and most important could be revealed. One sample Wilcoxon signed-rank test used the data from the second part of the questionnaire to analyze the frequency and importance. In addition, the mean generated for each productivity metric is evaluated as well, with top five productivity metrics for both level of frequency and importance identified for discussions.

Investigation of Stakeholders' Alignment on the Various Metrics
As the respondents of the questionnaire are from different institutions such as contractors, consultants, owners, and government agencies, it is necessary to carry out an inter-group comparison checking whether the respondents opine differently in sense of their institutions. Kruskal-Wallis test is a rank-based non-parametric statistical test method checking the potential differences among two or more different groups [53][54][55]. This study used the Kruskal-Wallis test rather than the one-way ANOVA test because the studied groups in Table 3 were of non-equal sample size and almost all metrics have their Levene's values ranging from 0.013 to 4.967 for frequency, from 0.057 to 6.074 for importance, and from 0.019 to 3.524 for performance (significant at p < 0.05) [56,57]. The null hypothesis is that there are no differences in the median between groups. A significance level (α) of 0.05 is used to assess the hypothesis. A p-value lesser than 0.05 rejects the null hypothesis and concludes that there is a difference in the median between the groups. On the other hand, if the p-value is more than 0.05, the null hypothesis is accepted indicating a consensus in the median between the groups. Kruskal-Wallis test used the data from the second part and third part of the questionnaire to analyze the scoring of frequency, importance of metrics and productivity performance by respondents from different groups.
To further evaluate the results of differences between two groups (e.g., between contractors and consultants, contractors and owners), Mann-Whitney U test was conducted, it is also a non-parametric test [58]. The Kruskal-Wallis test conducted in the previous step only identified a difference between subgroups for the productivity metrics and does not tell which specific groups are significantly different. Hence, a more extensive analysis to understand which specific groups were statistically significant different from each other. Use the Mann-Whitney U test for post hoc analysis, Bonferroni adjustment should be considered to adjust the significance level to control for Type 1 errors [59]. The null hypothesis is that there are no differences between two groups. A p-value lesser than adjusted significance level rejects the null hypothesis and concludes that there is a difference between two groups. Mann-Whitney U test used the same data as Kruskal-Wallis test for analysis.

Investigation of Differences on Assessments at Different Levels
To check whether there are significant differences in results when different metrics are used to assess the same project, the Friedman test was conducted. The Friedman test is suitable for two-way rank analysis of variance in the randomized block design [60]. It is similar to the Kruskal-Wallis test, but takes into account the influence of different groups [61]. In this step, according to the results of the Kruskal-Wallis test, the data set was divided into different blocks to minimize the difference between the groups. This step aims to reduce the error caused by the influence of the blocks to effectively distinguish whether there is a significant difference in results when different metrics are used to assess the same project. The null hypothesis is that there is no difference between the four levels of metrics when assessing the same project. A p-value lesser than 0.05 rejects the null hypothesis and concludes that there is a significant difference between the four levels of metrics. The Friedman test uses the data in the third part of the questionnaire. The test item is the rank of productivity performance. To determine which level has the highest and lowest productivity performance, the average performance score of each level (i.e., industry, company, project, and trade) was calculated for every group of respondents (i.e., designation or years of experience in industry or years of experience in assessing productivity), and the ranking of the scores of the four levels for different groups were presented, then SPSS was used for significant difference analysis.

Statistical Test Results of Productivity Metrics
This study mainly uses four non-parametric statistical test methods to analyze the data, one sample Wilcoxon signed-rank test was adopted to check whether the identified productivity metrics are frequently used in practice, as well as whether they are of importance to the practice; Kruskal-Wallis test was used to carry out inter-group comparison; Mann-Whitney U test was used for post hoc analysis between two groups; and Friedman test was suitable for checking the significant difference in results achieved by different metrics upon one project. Table 4 presents respondents' assessments of productivity metrics (usage in frequency and relative importance) as well as the relevant results of statistical analyses. Table 4. Assessments of productivity metrics in sense of usage in frequency and relative importance. According to the one sample Wilcoxon-signed rank test results in Table 4, assessments of most productivity metrics were statistically lower than the test value 4, except for P10, P15, P16, P19, and P20. Such results indicate that, although many metrics are proposed by the authorities to assess productivity, only several of them are frequently used and considered important by the practice.

Stakeholders' Alignment on the Various Metrics
As for inter-group comparison based on the second part of the questionnaire, the Kruskal-Wallis test results show that all respondents shared unanimous opinions on frequency and importance regardless of their years of experience in the construction industry and in assessing construction productivity, but there are differences between designation groups. To determine which two designation groups have a statistically significant difference, the Mann-Whitney U test was conducted for metrics P7, P19, and P20 in the frequency section, and P7, P10, P19, and P20 in the importance section. According to Bonferroni adjustment, adjusted significance level α = α/m = 0.05/6 = 0.00833 [62]. Results indicated that there is a difference between the designations with government agencies, as presented in Table 5. The values in bold show a statistically significant difference.  As for inter-group comparison based on the third part of the questionnaire, Table 6 lists the differences in productivity performance among different groups of respondents. The Kruskal-Wallis test results show that although respondents of different designations and years of experience in the construction industry shared unanimous opinions on productivity performance, according to the different years of experience in assessing construction productivity, the respondents' perspective on metrics are inconsistent, especially the trade level (P21-P33). To determine which two specific groups are statistically different, the Mann-Whitney U test was conducted as the results presented in Table 7. Adjusted significance level α = α/m = 0.05/3 = 0.0167. Results observed a difference in the group of more than 10 years of experience in assessing construction productivity. Respondents in subgroup ">10 years" indicated a perception of trade level productivity metrics having a better productivity performance compared to the two other groups.

Differences on Assessments at Different Levels
The results were affected by the subjective judgments of the respondents, thus it may fail to accurately reflect the actual situation. It is suitable to use the Friedman test method for comparison. Table 8 shows the average scores and corresponding ranks of productivity metrics performance, based on different levels as well as the different designations of respondents. The result of Friedman test reported p = 0.017, which was less than 0.05. The null hypothesis was rejected and concludes that there are significant differences when using different levels of productivity metrics to assess the same projects.

Discussion
Recalling the three research questions raised in the introduction section, first, 33 prevailing metrics for construction productivity measurement in Singapore were obtained through literature review and pilot interview. Then based on the second part of the questionnaire and statistical analysis, explored the usage in frequency and relative importance of different productivity metrics, and determined the top five productivity metrics in usage frequency and importance. Stakeholders' alignment on the various metrics reflects the main reasons for the differences. Finally, based on the third part of the questionnaire and statistical analysis, reported that there are significant differences when using different levels of productivity metrics to assess the same projects.

Top Five Productivity Metrics in Usage Frequency and Importance
In terms of usage frequency, the one sample Wilcoxon signed-rank test in Table 4 showed that the p-values of P10, P15, P16, P19, and P20 are greater than 0.05, suggesting respondents' assessments of these metrics are statistically same as the test value of 4. In addition, as presented in Table 4, none of the identified metrics received a mean value greater than 4; according to the rating scale adopted by the questionnaire, it means that P10, P15, P16, P19, and P20 are the only productivity metrics that are frequently used in productivity assessment. In terms of relative importance, as shown in Table 4, the p-values of P15, P16, P19, and P20 generated by one sample Wilcoxon signed-rank test are greater than 0.05, indicating respondents' assessments of these four metrics are statistically equal to the test value of 4. Furthermore, none of the identified metrics received a mean value higher than 4 except P19 and P 20; according to the rating scale adopted by the questionnaire, it means that P15, P16, P19, and P20 are the only productivity metrics that are considered important by the practice. Due to the word limit, only P10, P15, P16, P19, and P20 are discussed in this paper.
As shown in Table 4, P19 "Constructability Score" and P20 "Buildable Design Score" received the top two highest assessments both in usage frequency and relative importance. As the results of the Kruskal-Wallis test and the Mann-Whitney U test, the government pays more attention to the Constructability Score and the Buildable Design Scores. The Constructability Score in Singapore is an index that measures contractors regarding their adoption level of labor-efficient construction methods and construction processes, such as system formwork and climbable scaffolding [63]. The score can reflect the productivity levels of contractors and it has become a mandatory requirement of BCA in tender evaluation to improve construction productivity. The Buildable Design Score is an index measuring the potential impact of a building design on labor usage. It facilitates the adoption of less labor-intensive construction methods and promotes greater use of prefabricated, modular, and standardized building components to improve site productivity [63]. It has been required by the government that any projects with a gross floor area of 2000 m 2 and above must comply with a minimum Buildable Design Score [63]. Additionally, the expert opined that these two scores are important to benchmark contractors' involvement on values for the project, as well as the client's involvement and consultant's value-added performance. Therefore, ignoring the productivity increase brought about by new construction technologies may lead to an underestimation of actual productivity, at the same time, appropriate productivity metrics should be identified according to the state of the art of construction in a region.
P16 "square meter of built-up floor area per man-day" ranked third both in usage frequency and relative importance. This metric reflects how productive the work conducted by one worker is. This metric is important because it is crucial to know how much work is completed on site so as to decide manpower requirements or if any adjustments are required to be made. A study conducted by SCAL [10] also revealed that square meter per man-day is one of the leading methods of assessing productivity by construction companies in Singapore. The survey respondents are also of the perception that "square meter of built-up floor area per man-day" is frequently adopted in the local construction industry. The metric provides an easy yardstick for the measurement of physical site productivity.
P15 "square meter per dollar" received the fourth highest assessment both in usage frequency and relative importance. The metric checks how much work is done on account of the amount of money spent. It investigates profitability which is governed by how productive per area of works. The expert commented that this metric can be used as a benchmark on the cost-effectiveness of a project, and it has been widely used by the construction industry of Singapore. The importance of financial consideration is underline in productivity assessment in the study by SCAL [10], the cost implications in this metric measurement are imperative to the productivity assessment. Especially in international projects, "square meter per dollar" is used as an important metric to measure project cost and productivity assessment [64].
P10 "output per worker" at a company level (P10) ranked fifth in usage frequency and relative importance. The metric was identified as a common measure of productivity by numerous studies, associating labor productivity as outputs to labor inputs [7,25]. It aids in measuring the competency of workers, which is a good yardstick to measure the productivity of the workers. Additionally, the metric allows review of workers' involvement and cost-effectiveness of workers' output so that changes can be made for future implementations. On the other hand, workers are directly engaged by contractors, which is good for contractors to keep track of productivity data, as opined by experts. Thereby "output per worker" will be more realistic for implementation as it is based on practicability, especially for projects that require many workers.

Differences on Assessments at Different Levels
As shown in Table 8, there are differences in productivity measurement results when different metrics are used to measure the same project. Therefore, it cannot consider only one certain level when assessing productivity. According to the results of the Kruskal-Wallis test and the Mann-Whitney U test, the respondents with more than 10 years of experience in assessing construction productivity indicated a perception of trade level having a higher productivity performance. The reason may be that metrics of trade level are the basis for productivity measurement, they are relatively easy and straightforward for use [10]. Then, the Singapore government concluded almost all trade level productivity metrics in the builders' guide of BCA a long time ago, thus, when the project is planned or implemented, metrics may have been incorporated into the schedule to respond to the need for high productivity. Finally, trade level productivity metrics are related to the individual performance of each trade, and workers will meet higher index requirements to obtain more rewards. Therefore, the trade level is not only easy to measure, but more importantly, rules based on metrics are also easy to implement. So, it is suitable for measuring the productivity of on site, and it also facilitates the government and managers to develop effective incentive policies.
Although the productivity performance evaluated by trade level productivity metrics is higher, stakeholders prefer to consider project level and company level productivity metrics in terms of frequency and importance. By contrast, productivity metrics at the project, company, and industry levels involves are more complex as they involve all the trades and require more details to be taken into consideration. Industry level productivity metrics are persuasive in theory but have initial investment and derived values that are difficult to measure in practice. It is also further supported by experts that trade level is the base of the productivity measurement, as project level is built upon the multiple trades, and company level is built upon projects and the industry level is up to the results of different companies. Consequently, it is important to measure productivity from a holistic view and select appropriate metrics or develop new metrics to measure productivity according to the actual situation of project, such as the character of project, site conditions, human resources, and local regime.

The Implications of Study
This study has practical implications because it revealed the implementation status of the different productivity metrics in the real world, provides evidence for finding effective management strategies which could help the authorities and the industry update their rules and policies accordingly. The research results also suggest that the authorities and industry practitioners adjust the productivity assessment system in the light of the actual situation, instead of selecting the only commonly used productivity metrics. A holistic assessment of productivity covering different aspects needs to be carried out if they want to gain objective and reliable information regarding the status quo of the productivity in their construction industry.
Apart from being beneficial to the practice, this study contributes to the current body of knowledge of productivity research. Although various types of metrics are used to assess productivity currently, limited research was carried out to investigate their usage in practice. This is the first study that presents a methodological approach to investigates the various productivity metrics used in the construction industry, by checking its usage frequency, relative importance, and possible differences in assessment results. The results indicate that the benchmarking of the productivity metrics system needs to be improved and exploring productivity metrics more friendly for assessment will be a research direction that can be considered for future research. Although this research is based on Singapore, other countries, particularly those emerging economies in Asia could use the methodological approach presented in this study to examine the productivity metrics used in their industry and carry out modification if needed.

Conclusions and Recommendations
This study investigated the different metrics that were used to measure productivity in the construction industry of Singapore. It first conducted the literature review of the productivity metrics and categorized the metrics at four different levels, namely trade, project, company, and industry. Subsequently, the frequency and importance levels of various productivity metrics were assessed. Results showed that the top five productivity metrics used in the construction of Singapore were "constructability score", "buildable design score", "square meter of built-up floor area per man-day", "square meter per dollar", and "output per worker." Results also showed that differences existed when metrics at different levels were used to conduct the same measurement, and that results generated by the metrics at the trade level are relatively higher than those from the project, company, and industry levels. The findings from this study can help government authorities review, evaluate, and modify the productivity metrics used by the practice, which would further help achieve the sustainable development of the industry.
Although the objectives of the study were achieved, there are limitations. First, this study collected perception-based data from experts in the industry, which may cause the issue of subjectivity. Second, the sample size of the questionnaire is relatively small and thus, caution should be warranted when the results are interpreted. Lastly, this study was conducted in the context of Singapore and the results may have applicability issue when applying to other countries.
Future studies could develop productivity assessment benchmarking systems and integrates the overview of project to implement productivity measurements. In addition, how to measure productivity accurately and efficiently in an era of information technology development will be an important task for industry research and policy development. Data Availability Statement: Some or all of the data that support the findings of this study are available from the corresponding author on reasonable request. Additional data that support the findings of this study, including the survey template and raw survey results, are available from the corresponding author on reasonable request.
Single factor productivity includes output per one input, or in some instances valueadded form.
Multi-factor productivity is defined as output per the contributions of both labor and capital as inputs.
Total-factor productivity takes into account combination of inputs (capital, labor and materials) with adjustment for technological progress (shift factor).
Growth Accounting Approach is estimations of cost functions, by estimating productivity over time by estimating production function.
Data Envelopment Analysis is a nonparametric method to determine efficiency of decision making unit (DMU) by the projection of input and output variables in geometric figures.  This section assesses the productivity performance of projects by metrics. You may encounter different outcomes when using different productivity metrics to measure the same project. Based on your experience in the past three to five years projects, please indicate your perception of the productivity performance under different productivity metrics by marking an X in the boxes below, using a scale of 1 to 5 for ALL the metrics (if the measured productivity outcome is good or poor):

Levels Metrics
Performance by Metrics