Exploring the feasibility of measuring individual labor productivity using a wearable activity tracker
MetadataShow full item record
Productivity measurement in the construction industry has received enormous attention from industry and academia. One of the most crucial factors for achieving high performance in a construction project is labor management. Despite technological advances, construction projects are still driven by labor intensive work. Onsite work productivity determines, to a great extent, the performance of construction projects. However, determining accurate productivity metrics remains a challenge for project managers. Productivity measurements are still collected or recorded manually from construction projects. Productivity assessments require that well-trained staff perform the measurements, and that considerable time is spent accurately collecting and analyzing the data. Moreover, whether the common data used for productivity calculations are accurate is questionable due to undocumented/unrecorded data. Productivity is measured at various levels such as by industry, by project, and by activity. Currently, there is no reliable way to measure productivity at the individual laborer level. Such measurement could provide detailed and accurate information about project productivity, if data from an appropriate number of workers is collected. Identification of the poor productivity performance would be improved using productivity recorded per second with personal and location data instead of hourly, daily or weekly productivity summaries. The principal objective of this research is to explore the potential of a new productivity measurement methodology that automatically collects laborer data used for calculating productivity. The basic concept is to utilize a wrist activity tracker for data collection process. It mainly consists of automated data collection, and automated/semi-automated data analysis. The basic work plan for extracting direct work hours from total work hours is to identify, understand, and input specific patterns that appear in each of the different activities in the data analysis process. For proof of concept, field experiments were conducted at three different sites. Results from actual observations (ground truth), and automated/semi-automated data analysis per experiment were compared to evaluate the proposed method.