Unlocking Productivity: Revealing Waste and Hidden Disturbances Impacting MEP Workers

Traditionally, waste in construction is seen as the physical byproduct of production resulting from the excessive use of resources (Bølviken et al. 2014). The Toyota production system added a lean, process-oriented perspective to the concept of waste by focusing on quantifying and eliminating time-related inefficiencies (Ohno 1988). Lean construction has adopted this process perspective of waste, examining the activities of individuals or teams on the shop floor (Koskela 1992; Ohno 1988; Womack and Jones 2003). In lean construction, wasted effort is an inefficient or unnecessary allocation of resources, time, and labor that does not add value to the overall progress or quality of a construction project (Koskela 1992). There are seven types of wasted effort in the lean literature: transportation, waiting, motion, extra processing, overproduction, defects, and inventory (Ohno 1988), to which making do (Koskela 2004) and unutilized talent (Ansah et al. 2016) can be added. These categories are commonly referred to in construction (Koskela 2000; Poreddy et al. 2015; Wandahl et al. 2021).

Waste arises from missing task preconditions (Koskela 2000). Preconditions include building design and specifications, directives, components and materials, workers, equipment and machinery, sufficient space, connecting work, external conditions, and working conditions (Bertelsen et al. 2006; Koskela 2000). These preconditions are subject to wide availability fluctuations in construction projects (Bertelsen et al. 2006), and thus, it is difficult to plan and control production such that preconditions are available to all workers at all times (Henrich et al. 2007). Starting work with missing preconditions causes waste and decreases the time spent on direct work (Dave et al. 2010).

Approaches to studying wasted effort in construction include surveys (Bajjou and Chafi 2022; Issa and Alqurashi 2020), interviews (Formoso et al. 2017; Kalsaas et al. 2014), and activity analyses (Gouett et al. 2011). Work sampling is often used to measure waste (Kalsaas 2011; Josephson and Bjorkman 2013; Wandahl et al. 2021). In work sampling, observers track workers’ activities at random or specified time intervals to measure their shares of direct work and wasted effort. Although categories of wasted effort vary, the share of direct work is consistently used by scholars (Neve et al. 2020a). Studies have reported direct work shares between 30% and 60% (Thomas et al. 1984; Horman and Kenley 2005; Gong et al. 2011; Neve et al. 2020a, b), meaning that 40%–70% of effort is wasted in construction.

Previous studies have not categorized preconditions or waste categories, possibly because detecting waste and preconditions is difficult when observations are not continuous (Seppänen and Görsch 2022). These shortcomings could be addressed by continuous observations, such as in time-motion studies, which combine an efficiency measuring method (time study) by Taylor (1911) with work activity analysis (motion study) by Gilbreth and Gilbreth (1922). Various time-motion studies have shown that unnecessary movement and waiting are the main waste types in construction. Demirkesen et al. (2022) found that about three-quarters of work time is wasteful, mainly due to waiting. However, as the observability of waste varies by type, certain types may be disproportionately represented (Demirkesen et al. 2022; Kalsaas 2011). Given this possibility of bias in evaluation, other approaches, such as observing direct work disturbances, may generate additional insights.

Researchers have investigated interruptions in construction. Hassan and El-Rayes (2020) and Dai et al. (2023) investigated why work comes to a complete stop and causes cascading delays. Neve et al. (2020b) used work sampling to investigate shares of activities before and after direct work stops and revealed that higher performance is achieved when direct work starts earlier and stops later throughout a workday. Tetik et al. (2021) reported that value-added work is less interrupted when kitting logistics are used to enable just-in-time material deliveries to installation areas. In addition, Zhao et al. (2019) concluded that longer uninterrupted presence time at the workplace is associated with higher shares of direct work, which is in line with Wandahl et al.’s (2023) finding of a negative correlation between direct work and movement. Similarly, in applied ergonomics, the role of the organized workplace in uninterrupted workflows has been established (Neyens et al. 2019). Interruptions divert the worker’s attention from the main task, which is stressful and exhausting, decreasing the worker’s situational awareness (Weigl et al. 2020). Görsch et al. (2022) examined time intervals between direct work activities and found that preparatory work, such as working with materials, reviewing plans, or making spatial adjustments, most frequently stopped direct work.

However, none of these studies considered waste types and missing preconditions. This study analyzed how missing preconditions affect wasted effort and lead to prolonged direct work disturbances. Thus, it expands research by Görsch et al. (2022) on stoppages and continuation of direct work by analyzing how much time is spent on activities before and after direct work stops. The study departs from Neve et al.’s (2020b) work by using a time-motion rather than a work-sampling approach. The term “disturbances of direct work” is used to describe the studied phenomenon.

Waste in the construction industry has been extensively studied (see Table 1). Various forms of physical and nonphysical waste in different countries and lifecycle phases of a building (design, construction, maintenance) have been explored, although most studies have focused on the construction phase. Literature reviews, surveys, and interviews are frequently used to investigate the causes of waste. Table 1 compares the methods and scopes of studies of the factors impacting waste and presents the most common causes of waste identified in each. The most frequently cited causes are preconstruction activities, such as improper planning and scheduling, poor coordination and supervision, and poor communication. However, previous research does not connect waste types and causes at the activity level on the construction site.

Studies using manual and automated onsite observations of individual workers’ activities to identify causes of waste have found preconstruction activities to be among the most common causes. Missing materials and equipment are most frequently identified and can be considered direct causes of work stoppages and disturbances. Observation approaches can be used to distinguish waste events by project type, trade specifics, and task specifics. Seppänen and Görsch (2022) examined differences in the proportions of work activities in a time-motion study and proposed explanations and root causes by reviewing the extent of missing preconditions and differences in project types. However, their analysis was not statistical and did not investigate the direct connection between waste types and their causes or the influence of trade factors.

Previous research has focused on wasted efforts and reasons for them through survey and interview approaches from a management perspective and in observational studies from a process and worker perspective. Here, activities were not considered subsequently and associated with waste events and missing preconditions simultaneously in the form of disturbances of direct work. This study shifted the focus to such disturbances and their causes. This approach enables a deeper understanding of why and how value-adding work is disturbed and how the management of preconditions can enable workers to perform tasks smoothly and resource efficiently in construction.

The study aims to improve the understanding of installation work disturbances to improve workers’ task performance and resource efficiency. The approach is to analyze a series of consecutive activities because reasons for disturbances are not apparent when observing individual tasks in isolation. Thus, installation work disturbances, related waste categories, and missing preconditions can be difficult to study with work-sampling approaches (Seppänen and Görsch 2022). Therefore, the study adopts an approach that makes it possible to uncover longer chains of events by conducting a time-motion study with cameras attached to worker safety helmets. Such a study can continuously track workers’ activities without direct onsite observation. The resulting video footage can be used to observe long chains of events (e.g., by stopping, reviewing, or adjusting the playback speed) to identify disturbances in direct work and to determine whether activities include wasted effort due to missing preconditions. Fig. 1 visualizes the research steps taken from initiating the project through data collection and classification to data analysis over more than three years.

Before the data were collected, the university’s ethics committee reviewed the research plan, and consultations were held with employee and employer organizations. Measures were taken to address the concerns of the unions and the ethics committee, including muting all audio tracks in the recordings and using anonymizing software to protect the personal data captured by the cameras, such as faces and license plates. It was agreed before data collection that only project- and trade-specific information of participants would be recorded and analyzed. No personal data (e.g., age, nationality, or company affiliation) were recorded. Neither raw video data nor any results traceable to individuals were given to the employers.

The video footage captures the participants’ work activities, covering an angle of approximately 180° in front of a worker’s face. Fig. 2 shows the setup of a helmet with a camera, power bank, and other safety equipment. Before data collection, five research assistants responsible for data collection and classification were recruited. They were taught the classification process that was then collaboratively tested and improved. A face-to-face introductory session was held with the assistants to explain the classification categories and process so they could reach a common understanding. Everyone had the opportunity to ask questions and discuss unclear distinctions between categories. All five assistants individually tested the classification approach with a dummy data set. They were asked to identify any unclear or confusing classification categories and process steps to ensure a better common understanding and eliminate overlaps between categories that had not been described and distinguished sufficiently.

The researchers intended to conduct time-motion studies on at least four construction projects, each with at least two MEP trades working on the jobsite at the same time. The COVID-19 pandemic limited the number of projects available for the study, and several originally selected projects could not be included due to COVID-19 exposure on the jobsite. Although unions supported the study and encouraged their members to participate, it was difficult to find consenting workers, primarily because of the additional COVID-19 exposure and the need to make up time delays. Many potential participants were skeptical of the research approach, collection of personal data, and potential additional burdens due to helmet cameras. Thus, an onsite presentation with the principal investigator was given to potentially interested parties per project. Participants were able to understand in detail the research approach and use of the data and to ask questions, leading to the recruitment of at least three participants from two trades per project. Initial skepticism turned to support through the onsite presentation and ongoing onsite discussion with the research assistants as the data collection progressed.

Data were collected from March to May 2021. In total, the activities of five workdays of 14 construction workers from four construction projects were filmed. Table 2 shows the main aspects of the studied projects.

Video data were collected via helmet cameras, with research assistants setting up the cameras every morning and checking the setup twice during each workday. Due to technical problems (e.g., the power bank was unconnected, and the recording switched from movie to picture mode) and fluctuations in working hours, fewer data were collected than the theoretical maximum of eight hours of video material per working day per worker. Moreover, participants were told to turn off the camera during breaks to reduce the amount of personal data recorded. Overall, 408:07:03 h of video footage was collected (see Table 3).

The recorded video footage was subjected to manual categorization in spreadsheets (Microsoft Excel). This involved a thorough categorization of worker activities, waste events, and missing preconditions based on predefined categories. Waste categories are Ohno’s (1988) original seven categories, augmented with unutilized talent (Ansah et al. 2016) and making do (Koskela 2004). Preconditions are those of Koskela (2000) and Bertelsen et al. (2006). Activity categories were developed based on prior literature and work-sampling studies utilizing direct observations (Kalsaas 2010; Neve et al. 2020b). In a master thesis project completed before this study, a smaller time-motion study was conducted (Pasila 2019), and the categories were employed, evaluated, and tested.

The ability to pause, rewind, and review video footage facilitated a more precise categorization of work than in previous studies, leading to category expansion (see Fig. 3). The extension of the categories included, for example, hauling activities being distinguished depending on the distance from the installation area (work $\mathrm{preparation}<5\mathrm{m}$; short-distance hauling between 5 and 30 m; and long-distance $\mathrm{hauling}>30\mathrm{m}$). This extended set underwent validation in the research project’s steering group meetings including Finnish MEP employer associations and companies, ensuring alignment with the study’s scope and interests.

As voices were not recorded, any verbal interaction was categorized as “Discussion.” Furthermore, as workers often performed more than one activity at a time (e.g., discussion and installation), the duration of these activities was divided into equal parts. Watching the footage allowed the researchers not only to classify activities but also to determine if any were impacted by a missing precondition and contained waste types (see Fig. 3). All types of activities, waste, and preconditions are described in detail in the Appendix.

Despite prior teaching and training, the research assistants classified some data differently. Rather than going through all the videos again, a spot-check analysis (consistency check) was conducted to check samples of activity categories, their description, and associated waste and precondition categories across the different classifiers, projects, and trades. Subsets of data were randomly selected, and categories were manually examined for inconsistencies, errors, or discrepancies between the data points. To ensure the reliability of the findings, the process was repeated with different random samples and then aligned with the description of each category (see Appendix) to ensure a consistent data sample. After the research assistants’ budgeted project hours for classification were used up and the consistency check was performed, the main researcher classified the remaining video material until data saturation was reached (see Fig. 4).

The process of data classification required more resources than originally planned. Therefore, due to budget constraints, not all collected data could be classified. After discussions within the project consortium and based on insights from previous research studies, two criteria were established to decide when data classification would be considered sufficient without compromising findings. First, it was decided that the study had to reach a 95% confidence interval, as recommended by Thompson (1987) for sampling studies. The classified data contain at least 1792 activities (data points) per project and at least 372 activities per participant. This is consistent with the number of data points necessary to achieve 95% confidence in a sampling study. Second, following Neve et al. (2020b), the validity of the data samples was tested using a stabilization curve that represents the standard deviation of the proportion of classified activities between accumulated data sets of 5 h. The more data collected, the smaller the deviation (the more stable the shares of categories become). Fig. 4 shows the standard deviation of activity shares between the samples.

With 150 h of video classified, a reduction in the standard deviation of more than 95% was observed compared to 5-h and 10-h samples. These results were considered to indicate sufficiently analyzed data such that additional sets of 5-h samples would not have a significant effect on the standard deviation and shares of activities, explaining the difference between the collected and classified data (see Table 3).

After the classification and before the analysis, the data were cleaned to exclude breaks and research-project-related activities (checking helmets, discussions with researchers) and scenes in which no clear activity was evident (e.g., overhead work when the footage shows only the ceiling). In the analysis phase, the activities were divided into electrical and plumbing/heating ventilation air conditioning (HVAC) work. Plumbers and HVAC installers have similar duties in terms of work patterns, size of components, and preparation needs, as both trades focus on preparing and installing conduit systems. Moreover, they require similar mechanical skills to understand how pipes, ducts, valves, and other components work. Because of the similarities between these two trades and the fact that the study collected data from one HVAC installer and six plumbers, it was decided to distinguish the two professions together from electrical work.

The data analysis included the calculation of direct work disturbances, the share of resource inefficiencies (waste types), and missing preconditions per disturbance. Fig. 5 shows a sample of the classified data: each row represents an activity with its duration and whether waste and missing preconditions occurred within it. The duration of a direct work disturbance (i.e., a set of activities between two direct work activities) is calculated by the sum of durations of individual activities within the disturbance (e.g., the sum of activity durations in rows 4, 5, 6, 7, and 8 to calculate the duration of the second disturbance, which is 4 min 56 s).

As well as tracking activities and their duration, it was determined whether nonvalue-adding work occurred together with waste or missing preconditions (“Waste” and “Missing Preconditions” columns in Fig. 5).

The duration of waste and any missing preconditions are difficult to measure accurately because only parts of an activity can be wasteful, and it is often impossible to determine through video observation alone which parts of the activity duration were affected. For example, in line 4 (Fig. 5), components and materials were missing while work preparation was in progress. However, only parts of the work preparation may have involved materials. In fact, the observer can only detect the wasteful impact of missing materials by looking at activities 5 (searching) and 6 (nonwork-related activities). To analyze waste and missing preconditions, it was decided to count activities affected by forms of waste and missing preconditions instead of attempting to calculate durations of waste. Disturbance durations and numbers of activities affected by wasted and missing preconditions were recorded numerically and as a percentage per disturbance (see Fig. 6).

To show the frequency and duration of direct disturbances of MEP work and reveal associations between resource inefficiencies and prolonged durations of direct work disturbances and the factors contributing to them, the data were analyzed statistically in three steps. Following a descriptive analysis of activities, wasteful instances, instances with missing preconditions, and disturbances, a two-step cluster analysis grouped disturbances based on their duration and the proportion of waste types within them. Cluster analysis is frequently employed to extract insights into associations in a sample (Jang et al. 2020), and various clustering methods were considered before two-stage cluster analysis was chosen for advantages over hierarchical (Balderjahn et al. 2018) and K-means (Sarti et al. 2018) clustering methods thanks to its use of a statistical goodness-of-fit measure. This measure determines the number of clusters based on algorithms, eliminating the need for subjective decisions (Jang et al. 2020). The two-step cluster analysis comprised 10 factors (duration and proportion of nine waste types per disturbance) to examine structural similarities within durations of disturbances. In the final step, the impact of missing preconditions on waste types was assessed using general linear model multivariate analysis. The nine waste categories were set as dependent variables, while the nine categories of missing preconditions acted as independent variables to determine how often waste occurred when the preconditions were missing. All tests were performed using IBM SPSS Statistics version 28.

The classification of 28 workdays of 14 workers based on the 16 categories (see Table 3; for details, see Appendix) resulted in 9,575 analyzed activities. Table 4 shows the total time and the share of each activity type in the total time. The results are presented for the overall sample and separately for different trades. Counts and average durations of classified activities were also calculated. The most frequently occurring activities were direct work and activities that facilitate direct work, namely, work preparation and working with materials. Discussion and nonwork-related activities (such as waiting) had the highest frequency of all nonvalue-adding categories. These mentioned categories accounted for 70.4% of all time in the sample.

On average, direct work accounted for 18.9% of work time. Electricians had a remarkably higher share of direct work (24.0%) than plumbing/HVAC installers (13.6%). The same pattern in proportions of direct work between electricians and plumbing/HVAC installers was observed in all four projects (the project breakdown is not shown in the table due to space limitations). Electricians showed higher shares of direct work-facilitating activities: work preparation, working with material, and short-distance hauling. Plumbing/HVAC trades engaged more often in searching, discussing, hauling (long-distance), measurement, and movement activities than electrical trades. Qualitative observations of the video footage showed that electricians were more likely to be involved in arranging and preparing cables in the immediate vicinity of the installation area, while plumbers/HVAC workers were more likely to organize work within a wider radius, including hauling materials and equipment from further away and reviewing and discussing plans while on the move. The mean activity duration was 63 s, demonstrating a highly fragmented workflow. Electricians’ average duration per activity was lower than that of plumbing/HVAC installers, indicating a higher change rate of activities in electrical work.

Table 5 shows the total count and proportion of waste events and missing preconditions. The shares are calculated separately for waste events and missing preconditions. Waste events and preconditions were included in the analysis when an activity contained wasteful elements or was carried out despite or because of missing preconditions. The most common waste events were motion, waiting, defects, and transportation. Differences in motion differentiate the trades, with plumbing/HVAC workers suffering more than electricians. Waste due to waiting was also higher in plumbing/HVAC work. In contrast, electrical work showed higher shares of making do, transportation, and defect waste events. Other types (e.g., extra processing, unutilized talent, and, in particular, overproduction) were detected much less frequently, as they were difficult to observe in the footage. Interestingly, some direct work observed in the video was wasteful because an electrician performed an installation task that was not part of his duties.

Components and materials, equipment and machinery, and connecting work were the preconditions most often missing. Plumbing/HVAC work suffered more from missing components and materials, equipment, and machinery. Electrical work more often lacked sufficient space and previous or connecting activities meeting requirements (connecting work). Building design and specifications, directives, and external conditions were identified much less frequently than other missing preconditions.

In total, 1,771 disturbances were classified and analyzed. Table 6 shows the descriptive results. Of these events, 1,196 were related to electrical work, and 575 were related to plumbing and HVAC work. Electricians had, on average, 85 disturbances per day, roughly double that of the plumbers/HVAC (an average of 41 per day). The total time of all disturbances per trade was about the same, though disturbances of plumbing/HVAC work were, on average, twice as long as in electrical work.

To analyze associations between resource inefficiencies and durations of direct work disturbances, a two-stage cluster analysis with 10 factors (proportions of nine waste types and duration of disturbances) was performed. The two-step cluster analysis resulted in a three-cluster solution, where the factors of motion and transportation and the duration of disturbances most impacted the cluster segmentation. Medium-impact factors were making do, waiting, defects, and inventory, while extra processing, unutilized talent, and overproduction had low impacts. The overall quality of a cluster solution was evaluated using the silhouette measure, assessing how well individual data points can be distinguished between clusters. It ranges from $-1$ to 1: $-1$ means that individual data points are likely to be clustered incorrectly, and 1 means that the data points are well arranged within their cluster. A silhouette value greater than zero is considered an indicator that a cluster solution is valid; thus, the value of 0.8 for the two-step cluster analysis in SPSS is reasonable to a strong indication of validity (Van Den Berge et al. 2017; Norusis 2011).

Table 7 shows the counts and total time of the clusters of disturbances. In cluster 1, 72.8% of all disturbances accounted for 25.9% of the time, representing many short-term disturbances with a mean duration of 1:31 min. In cluster 2, 20.3% of the disturbances accounted for 41.6% of the time, resulting in a medium-length mean duration of 8:46 min. Cluster 3 included just 6.9% of disturbances but accounted for 32.5% of the time, resulting in the longest mean duration of 20:15 min.

As the average duration for the distinction of the clusters is crucial, clusters are labeled short-term (1), medium-term (2), and long-term (3) disturbances. Short-term disturbances do not include any wasteful activities. Medium-term disturbances contain the highest proportion of motion among the clusters and an increased proportion of waiting and defect incidents. Long-term disturbances include the highest proportion of transportation waste and show an increased proportion of motion, making do, and inventory waste.

Qualitative observations from video footage indicate that plumbing/HVAC workers occupied different areas and rooms to carry out their tasks. Due to larger sizes and lengths of components in plumbing/HVAC work, installers were often constrained by limited space. They prepared components and carried out direct work in other locations and then hauled them to the installation area when needed. Hauling over long distances and from further away seems to be related to higher shares of searching, moving, nonwork-related activities, and discussions in plumbing/HVAC work (see Table 4). All clusters showed a low or nonexistent share of extra processing, unutilized talent, and overproduction, as very few instances of these types were observed in the data set (see Table 7).

Cluster differences per trade were compared using Pearson’s chi-squared tests by the time of their occurrence (morning, noon, afternoon), project, and the proportion of activities they included (see Table 8). Short-term disturbances occurred most throughout the working day in all projects. Their proportions were always higher in electrical work except in multifamily building 2, where more long-term disturbances occurred for electricians. In both trades, the proportions of short-term disturbances were highest in the noon hours and higher in supervisors’ work than in that of installers.

Plumbing/HVAC work saw more medium-term disturbances, indicating that more setup time was needed to perform direct work. Both trades showed a peak of medium disturbances in the morning, representing long starting times before performance increased. Plumbing/HVAC work in the hotel and office project saw the highest shares of medium-term disturbances and lowest of short-term disturbances, reflecting its lowest share of direct work across all projects. The trend for plumbing/HVAC work to have higher shares of medium-term disturbances than electrical work is also present when differentiating between roles.

Long-term disturbances reached their lowest value during noon hours while short-term disturbances without waste shares reached their highest level, which could indicate performance peaks. During afternoon hours, long-term disturbances peaked in electrical work, which could mean more wrap-up times (hauling, long-distance) at the end of the workday. The proportions of long-term disturbances of plumbing/HVAC work remained similar in the noon and afternoon hours.

Supervisors tended to experience more short-term disturbances, while installers encountered a greater number of long-term disturbances. The pattern is different for medium-term disturbances, where electrical supervisors displayed more instances compared to installers. However, plumber/HVAC supervisors exhibited slightly fewer medium-term disturbances than their installer counterparts. These differences may stem from variations in experience and task assignments. For instance, workers often handle tasks like searching and hauling, which are associated with high levels of transportation, motion, and inventory waste. These waste types are particularly high in long-term disturbances.

Table 8 also shows statistically significant differences between trades in morning and afternoon hours in the multifamily building 2, hotel and office, and shopping mall projects. Projects had special characteristics that can explain some of the differences. Multifamily building 2 had design issues related to modular bathrooms that can be seen as causing a higher share of long-term disturbances for electricians. The hotel and office project had schedule delays caused by COVID-19, increasing medium-term disturbances. The shopping mall was the largest project with the widest site layout causing considerable long-distance transportation for the plumber/HVAC trade. One trend seen across all factors was the distribution of short-, medium-, and long-term disturbances, following the pattern of trades. Electricians showed higher shares of short-term disturbances and lower shares of medium-term ones than plumbing/HVAC workers in almost all incidents. There were smaller differences among the trades in long-term disturbances.

The number of missing preconditions was calculated within the three clusters detected (see Table 9). The largest differences between short-term and longer-term clusters were related to components and materials and connecting work. This shows the importance of logistics and production control for minimizing longer, wasteful disturbances. The same categories also sometimes occur in short-term disturbances, but most missing preconditions in short-term disturbances were related to working conditions.

To determine whether the number of missing preconditions (independent variables) and covariates affected the occurrences of waste types (dependent variables) within disturbances, a general linear model (GLM) multivariate analysis was conducted (Minooei et al. 2020; Naji et al. 2022). The categorical variables project type, role, profession, and daytime were set as covariates. Table 10 shows the significance of the impact of the independent variables and covariates on the types of waste. Cells marked with “x” indicate a significant correlation.

The GLM analysis, based on Wilk’s $\mathrm{\Lambda }$, reported a significant multivariate effect of eight out of nine missing preconditions and three out of four covariates on the nine types of waste. Thus, all variables are significant in explaining the differences between the independent variables, except for the covariate variable project [$\mathrm{F}(9,\mathrm{1,679})=0.747$, $\mathrm{p}<0.666$; Wilk’s $\mathrm{\Lambda }=0.996$] and the independent variable working conditions [$\mathrm{F}(36,\mathrm{6,293.728})=1.112$, $\mathrm{p}<0.297$; Wilk’s $\mathrm{\Lambda }=0.977$], although working conditions showed a statistically significant impact on waiting waste. Table 10 indicates that different waste types were significantly affected by various preconditions and other factors (covariates). The availability of components and materials impacted the most waste types: transportation, motion, defects, inventory, unutilized talent, and extra processing. The readiness of previous tasks (connecting work) and worker qualification and presence (workers) influenced the second-most types of waste. Building design and specifications, equipment and machinery, and sufficient space significantly influenced various waste categories. In addition to the readiness of preconditions, other characteristics (covariates) had a significant effect on waste occurrences. Workers’ roles (installers or supervisors) and their trade affected waste types significantly, especially transportation, waiting, and defects. Differences between electricians and plumbing/HVAC (covariate profession) are significant for waiting, defects, transportation, and extra processing. Interestingly, project type did not significantly impact occurrences of waste, indicating that the findings are generalizable to different contexts. The data and factors (independent and covariates) analyzed explain an average of 23.2% (average R squared, see Table 10) of waste occurrences. Motion was explained best (48.2%) by the analyzed factors, and overproduction was the least explained (0.8%). Low explanation results related to overproduction, unutilized talent, making do, and extra processing were due to the small number of instances in these categories in the sample.

The findings of this study show that task performance at the operational level was fragmented and involved many activity changes in short periods. The results of previous exploration of this phenomenon (Görsch et al. 2022) using a small, 5-h time-motion data sample were confirmed here in a much larger sample.

According to Weigl et al. (2020), the fragmentation of direct work contributes to lower situation awareness and increased levels of stress and exhaustion, which can degrade performance. In contrast to the explanations of Weigl et al. (2020), the study’s results indicate that electricians were disturbed twice as often as plumbing/HVAC installers but still had a higher level of performance in terms of direct work. Based on the findings, it is not possible to establish an association between the number of disturbances and the share of direct work. However, medium-term disturbances impacted performance most in this study and were most occurring in the hotel and office projects, where both trades had the lowest performance values. The finding that medium-term disturbances have a high impact on the share of direct work in MEP is backed up by the fact that such disturbances accounted for most of the time spent not engaged in direct work activities.

Disturbances can be categorized by their duration and the proportion of waste types they contain. Three clusters were found: short-, medium-, and long-term disturbances. The start and continuation of installation work require the organization and preparation of materials and equipment in the immediate vicinity of the installation area and lead to short-term disturbances. Lack of working conditions, materials, sufficient space, and equipment often forces workers to briefly pause direct work and adjust for the next steps before installation work continues. For example, activities such as material handling, work preparation, and measurements account for a high proportion of the short-term disturbances of plumbing and HVAC work required between direct work activities. These short-term disturbances do not include wasted effort; they support the continuation of direct work. This understanding is novel and explains why the frequency of disturbances alone does not explain low shares of direct work.

Most medium-term disturbances were associated with a lack of connecting tasks and components and materials. The importance of production control can be seen in the hotel and office projects, which suffered from schedule delays caused by COVID-19. In that project, medium-term disturbances were the most frequent. In contrast, the performance of electricians was highest in the shopping mall project where lean methods were used and production was controlled using takt principles. Schedule delays impacted performance by increasing the share of medium-term disturbances. This hypothesis should be further tested in future research. Medium-term disturbances occurred in the wider working area, including movement and transportation to/from nearby areas for storing, preparing, cleaning, and so forth. These disturbances were caused by changing from one installation point to another (readjusting materials and tools) or the current workplace being organized to continue direct work. Overall, they resulted in movement close to the installation area. The importance of workplace infrastructure has been highlighted by Koskela (2000, 2013), and Formoso et al. (2017) as a cause of waste. This study shows the impact on medium-term disturbances, which cover most of the time spent on nonvalue-adding work. It also confirms the qualitative observations of Seppänen and Görsch (2022), who identified movement and transportation as the most common types of waste in MEP work. They are caused by uncoordinated and unready materials and equipment and low maneuverability within the wider area surrounding the installation point. Thus, providing tools and materials where and when needed has the potential to reduce waste in construction (see Ballard 2000; Bertelsen et al. 2006; Koskela 2000).

Within long-term disturbances, installers are pulled further away from the installation point and working area as indicated by high shares of transportation, motion, and inventory, which are associated with areas outside the installation area. Thus, it is logical that their shares were highest in the largest project of the study, the shopping mall, where material storage was often distant. By contrast, the project with design issues had many long-term disturbances for electricians, showing the impact of design on performance. In long-term disturbances, installers face issues of not having the necessary materials, tools, space, or connecting work, resulting in significantly higher shares of searching and hauling materials and equipment or the need to discuss issues faced in the working area with coworkers and supervisors. Reaching for materials, equipment, and coworkers causes longer disturbances, aligning with the findings from Tetik et al. (2021) and Neyens et al. (2019). The impact of unmanaged onsite logistics as a cause of wasted effort has been reported (Koskela 2000; Zhao and Chua 2003; Khanh and Kim 2014; Kalsaas et al. 2014; Formoso et al. 2017; Bajjou and Chafi 2022). This study identifies these causes as the missing preconditions with the largest impact on wasted effort, showing that the organization of logistics could have a large impact on productivity, which is in line with earlier studies on kitting (Tetik et al. 2021; Zhao et al. 2021).

Factors that affect wasted effort in MEP work include the level and accuracy of task-level planning, the role of workers, spatial maneuverability, and trade specifics. The amount of discussion was high in both trades, indicating situations in which further work coordination and clarification about task level is required, as these are often not sufficiently addressed in production planning and control. Thus, workers carry out task planning and control activities (Görsch et al. 2024), which are among the most commonly identified factors driving waste, confirming the findings of Alarcón and Serpell (1996), Koskela (2000), Horman and Kenley (2005), Kalsaas (2013), Khanh and Kim (2014), and Bajjou and Chafi (2022).

Supervisors experience fewer and shorter disturbances (most short-term), supporting continuous direct work. In contrast, installers face more frequent and prolonged disturbances in their direct work. They spend more time clearing due to insufficient space and working conditions and clarifying (discussing) unclear preconditions as well as searching for and hauling materials and equipment elsewhere than in the installation area. The finding raises questions about whether supervisors choose easier work for themselves and prioritize the requirements of their own production tasks or whether their role-specific tasks and experience contribute to these differences.

Another factor affecting wasted effort and the duration of disturbances of MEP work is spatial maneuverability, which describes the effort needed to cope with spatial, material, and equipment limitations. When the materials and equipment used for installation are large compared to the installation area, the work is more challenging, and more effort is needed to plan and coordinate task performance. The understanding of spatial maneuverability in terms of efforts and details needed for operational task planning was qualitatively described by Gil et al. (2000) and Görsch et al. (2024). They defined it as a part of task planning and control in the operational construction process carried out by construction workers themselves in a decentralized manner. In contrast to previous reports, this understanding was identified for the first time as a factor impacting the amount of wasted effort and the duration of disturbances of MEP work. This study showed statistically significant differences in disturbances between trades in relation to specific times of day, roles, and projects, which can explain differences in the amount of waste. Interestingly, project differences affect the proportion of activities and the amount of waste (e.g., hotel and office: scheduling issues; multifamily: design issues; shopping mall: lean implementations) but not how preconditions affect waste types. The fact that the type of project had no significant impact is taken as an indication of the generalizability of these results.

This study is the first to determine that job-specific task requirements and processes significantly influence the frequency and duration of disturbances as well as the extent and type of waste. This pattern was present despite differences in projects, roles, and times of day. These findings parallel the trade-specific differences in decentralized planning behavior documented among electrical and plumbing/HVAC workers (Görsch et al. 2024). The findings of Demirkesen et al. (2022) show that waiting is the main source of wasted effort in concrete work, and the results of this study show that movement and transportation are predominant in MEP work, highlighting the influence of trade and work type on wasted effort and thus on direct work levels. This contrasts with earlier studies (Neve et al. 2020b; Strandberg and Josephson 2005) reporting that trade had no impact on the proportion of wasted effort or direct work. The study’s findings, derived from time-motion studies categorizing activities, wasted effort, and missing preconditions, question whether work sampling can identify both symptoms and root causes of problems empirically. This may explain why differences between trades were unexplained in earlier work-sampling studies.

To reduce wasted effort and shorten the duration of direct work disturbances, providing better preconditions and work processes is key. Short-term disturbances, such as incidents of missing preconditions, occur most frequently and thus hold potential for improvement. They mainly occur in the direct vicinity of the installation area and support the continuation of direct work. It can be argued that their duration depends on arranging and maneuvering individual skills within the installation area. Thus, an orderly and structured installation area and process can decrease the need to stop and rearrange, thus reducing short-term disturbances and supporting activities. Introducing workers to lean principles, such as the concept of waste or 5S (sort, set in order, shine, standardize, sustain), can increase workers’ process awareness in terms of organizing spaces and materials more efficiently by identifying and reducing waste. The concept of maneuverability during the installation process should also be considered when products and processes are prefabricated, with materials and equipment kitted to installation areas.

Medium-term disturbances were characterized by insufficient space and incomplete connecting work, resulting in additional planning and coordination activities. Accordingly, workers should be provided with space in the right condition for installation work so they can work without disturbances. Digital systems and visualization approaches can display available spaces with set preconditions to improve individuals’ situational awareness and reduce the number of site visits to gain an understanding of what is happening and find possible work areas to reduce and avoid disturbances. Future research and practice should explore how digital advancements such as digital twins (Sacks et al. 2020), digital visual management (Reinbold et al. 2022), and shorter cycle times (Riekki et al. 2023) can enhance data granularity. Such approaches will facilitate more efficient data collection, analysis, and distribution, continuously supporting workers’ information needs with task-relevant information based on faster-updated planning and control cycles.

Long-term disturbances to both direct work share and waste can be mitigated by implementing digital systems accessible to workers. These systems provide real-time information on the availability and location of materials and equipment, thereby minimizing search efforts. However, pointing out the current locations of materials and components does not directly reduce moving and hauling. Thus, this paper proposes the development of mobile workstations, which can be equipped with task-specific materials, components, equipment, and digital visual management devices in a structured manner (Tezel and Aziz 2017). When spatial and process limitations hinder the deployment of mobile workstations carrying materials and equipment, on-demand onsite logistics as a service can be an alternative solution.

Finally, a more detailed tracking of workers’ onsite movements seems logical, given its strong connection to the primary causes of waste—motion and transportation—and their high occurrence in medium-term disturbances. Earlier research (Zhao et al. 2019) suggested using uninterrupted presence time as an indicator for increased direct work. Wandahl et al. (2023) demonstrated a significant negative correlation between movement and share of direct work. While the study’s results align with these, there are potential improvements to consider. Depending on the precision of location tracking, short- and medium-term disturbances may go unnoticed, hindering direct work. Because several disturbance types take place in the immediate vicinity, the assumption that uninterrupted presence in assigned work locations automatically contributes to increased value-adding times is challenged. Trade-specific needs lead to various work areas spread across construction sites, which is not accounted for in current scheduling methods. Future research should detail how workers utilize spaces on construction sites and develop accurate, real-time tracking methods. Since movement is a symptom of missing preconditions, more focus should be placed on network and complex systems behavior in onsite movement. Doing so could identify both problems with preconditions and recognizing medium- and long-term disturbances.

Certain types of waste, including extra processing, unutilized talent, and overproduction, were infrequently detected and had low explanatory power in the statistical analysis. The challenges in identifying and tracking these waste types can be attributed to the visual nature of time-motion and work-sampling studies, often conducted by nonexperts. A deeper understanding of trade-specific installation processes would enable better detection of these waste types, similar to detecting making do. In the study’s methodology, these waste types tended to remain hidden, aligning with the findings of Demirkesen et al. (2022) and the observations of Kalsaas (2011). Their hidden nature can have a significant impact on the outcomes of cluster and regression analyses. While these waste types are likely prevalent in reality, the current camera observation method was not effective in capturing them.

Another factor that affected activity classification was the muted audio of the videos, required by the ethics committee and workers’ associations, which limited the analysis of the causes of wasted effort. The content of conversations and ambient sounds could have provided further information about activities, waste, and missing preconditions if these were unclear in the footage. The content of conversations could also have indicated whether they were work-related and exactly what problems required discussions, thus preventing misinterpretations in the classification.

The filming and classification of construction work is not a new research method. However, the current manual classification is not an efficient option for commercial use. The observation and classification of activities by nonexperts are fundamentally prone to misinterpretation, as previous work-sampling studies have shown (Josephson and Bjorkman 2013; Neve et al. 2020b), although these and earlier time-motion studies yielded valuable insights. Consequently, additional efforts must be made to supervise data collection and classification to obtain valid data. Moreover, the current observation and classification approach was resource-intensive and did not allow for real-time analysis and intervention in the projects. Future research activities should focus on the development of monitoring and sensing concepts that support or reduce subjectivity and manual effort through automation. However, automation is likely to be complex because events cannot be understood as snapshots; the interpretation of waste requires lengthy sequences of data to be analyzed.

The geographical limitations and limited number of professionals in the study limit the generalizability of the results. Investigating 9,575 activities (including 1,771 disturbance, 1,511 waste, and 1,320 missing precondition incidents) over 28 workdays generated by 14 participants is in line with earlier work sampling and time-motion studies in terms of sample size and the conclusions derived (Josephson and Bjorkman 2013; Neve et al. 2020b; Strandberg and Josephson 2005; Wandahl et al. 2023). Furthermore, the study examined similar patterns of waste proportions and missing preconditions in projects with different layouts and approaches. Some factors were project-specific and depended on schedule delays, design issues, and site layouts. However, the relationship between preconditions and waste categories was generally applicable in the projects, and earlier studies have found similar patterns and causes of waste with different study methods and in different countries. Thus, the overall results can be considered generalizable but should be considered with caution when applied to individual projects. Further time-motion studies with the same trades in different projects and contexts are needed to deepen the paper’s results and test their reproducibility and generalizability.

The study has provided empirical data and evidence on trade-specific practices, an under-researched area, although around 70% of construction projects are carried out by contracting companies (Josephson and Bjorkman 2013). The study’s data set is unique, being an extensive and detailed data set of operational task processes in construction, including work activities, waste categories, missing preconditions, and a statistical analysis of their interdependencies. Furthermore, the depth of analysis (16 activity, nine waste, and nine precondition categories) contributes to the unique character of the study and was not shown in previous time-motion (Demirkesen et al. 2022) or work-sampling studies.

Due to its second-to-second classification approach, the study’s method can examine work processes in greater detail than work sampling. This raises the question of how well work sampling can represent reality since it is based on isolated and irregular observations that can be independent of context. The study shows lower shares of direct work (13.6% for plumbing/HVAC and 24.0% for electrical work) than earlier studies (30%–60%) (Gong et al. 2011; Neve et al. 2020a, b; Thomas et al. 1984). This finding can be related to two factors. First, using more activity categories (16 in total) than earlier studies could have affected direct work share. The category measurements (working with material, work preparation, maintenance and cleaning, inspection, and direct work) were not similarly distinguished in earlier studies; thus, it can be assumed that differences in direct work proportions are related to the classification scheme used and its level of detail. This supports Josephson and Bjorkman’s (2013) claim that these methods are limited in measuring productivity and comparing performance over time, which calls into question comparisons of direct work proportions across decades, projects, and disciplines, such as in the work of Neve et al. (2020a) and Wandahl et al. (2021). Second, using continuous observation rather than work sampling with regular intervals (e.g., every 5 min) could lead to lower shares of direct work being detected. Continuous observation of activities can create a more comprehensive understanding of the effects and causes of activities. If observation is carried out every 5 min or at irregular intervals, many activity switches are missed. The transition from nonvalue-adding to value-adding work happens quickly, which may lead the observer to see direct work, this being easier to recognize. Short segments of nonvalue-adding work add up to a significant share of time, likely explaining the differences between work-sampling studies and this study. The differences between the two methods could be further evaluated by using both methods on the same data set.

The study revealed that some direct work is wasteful, although previous research tends to assume that direct work is inherently nonwasteful, given its direct contribution to project completion (Koskela 2000; Neve et al. 2020b). Previous research, primarily based on surveys and interviews, has often pinpointed the causes of waste in preconstruction phases. In contrast, observational methods have demonstrated that waste causes are predominantly observable in the construction process. In summary, the aforementioned points underscore a research gap in our understanding of the concept of waste and its underlying causes in the construction industry. Addressing this gap should be a priority for future research theoretically, methodologically, and empirically.