BIM Log Mining: Measuring Design Productivity

Abstract

There has been a long debate on how to measure design productivity. Compared to construction productivity, design productivity is much more difficult to measure because design is an iterative and innovative process. Today, with rapid extension of building information modeling (BIM) applications, tremendous volumes of design logs have been generated by design software systems, such as Autodesk Revit. A systematic approach composed of a detailed step-by-step procedure is developed to deeply mine design logs in order to monitor and measure the productivity of the design process. A pattern retrieval algorithm is proposed to identify the most frequent design sequential patterns in building design projects. A novel metric for measuring design productivity based on the discovered sequential patterns is put forward. A large data set of design logs, provided by a large international design firm, is used as a case study to demonstrate the feasibility and applicability of the developed approach. Results indicate that: (1) typically, each designer executes specific commands more than any other commands; for instance, it is shown for a designer that the accumulative frequency of three commands can reach up to 56.15% of the entire number of commands executed by the designer; (2) a particular sequential pattern of design commands (\”pick lines →  \”trim/extend two lines or walls to make a corner→ \”finish sketch”) has been executed 2,219 times, accounting for 46.75% of instances associated with the top five discovered sequential patterns of design commands; (3) the identified sequential patterns can be used as a project control mean to detect outlier performers that may require additional attention from project leaders; and (4) productivity performance within the discovered sequential patterns varies significantly among different designers; for instance, one of the designers (designer #6 in the case study) is identified as the most productive designer in executing both Patterns I and II, whereas another designer (Designer #1) is found to be the most productive designer in executing both Patterns III and IV. It is also uncovered that designers, on average, spend less time running the most observed sequential patterns of design commands as they gain more experience. This research contributes: (1) to the body of knowledge by providing a novel approach to monitoring, measuring, and analyzing design productivity; and (2) to the state of practice by providing new insights into what additional design process information can be retrieved from Revit journal files.