Productivity Reimagined, FTW

Swinerton deploys innovative technology stack to measure construction progress and productivity

The construction industry, one of the world economy’s largest sectors, has been notoriously slow to embrace technology and innovation. According to McKinsey, the industry’s productivity growth has been a mere 1% annually for the past 20 years, costing the global economy $1.6 trillion a year. Fortunately, the industry is now adopting new technologies in earnest to improve productivity gains. One such technology is BIM (Building Information Modeling), whose power is to act as the living digital twin of a building. BIM is designed to make building data accessible and connected throughout an asset’s lifecycle, starting with the original design model (or an as-built model in the case of renovation) and maintaining it throughout construction further to operation.

BIM is now a part of the majority of construction projects, though it is largely employed in the design and planning phases limiting its usefulness in enhancing project productivity. According to Ernst & Young Construction and Real Estate Advisor Eric Ottinger, this phase only accounts for 7% - 12% of a total cost of a construction project. BIM does not do the project much good if it lives with the architect, is simply used to create plans, and then issued or “thrown over the wall” for construction. This is where data silos begin to develop. Much is lost in translation as it moves from the office to the field. As soon as construction begins, project variables increase exponentially and with that, the amount of risk. It’s unrealistic to rely on the broad strokes early on in the schedule process and then simply press “GO” hoping everything operates exactly as planned.

The big opportunity to get real productivity gains is to apply technology and automation to the majority of the project that happens on site, enabling construction teams to closely monitor progress and catch mistakes before major rework is needed, ideally on a daily basis. It’s the missing link to close the feedback loop and keep building models continuously up-to-date. 


The complexity of commercial projects make it virtually impossible for the human eye and human intuition alone to track site progress and quality. For years, site inspection involved teams of sub contractors, general contractors and owners doing walk throughs with clipboards, measuring tape and cameras to document progress. Some forward-thinking construction firms began collecting data through reality capture technologies, utilizing terrestrial scanners, drones, photogrammetry and 3D cameras. Yet, these processes can be laborious, expensive and difficult to scale. The data is essentially a stamp in time, whereas a construction project is an ever-changing and evolving system. To implement such technologies at a regular frequency can limit the rate of adoption and potentially wipe out any productivity gains achieved.

With a mission to move construction productivity forward in the built environment, U.S. construction giant Swinerton is eagerly taking on the challenge of automating construction monitoring, progress tracking and deviation reporting. The individuals of Swinerton’s Innovation Team, headed by Eric Law, are champions of technology, not simply in and of itself, but for analyzing and solving the real world business problems they face in the construction industry. 

Swinerton, which operates between 150 and 200 active job sites per year, needed to put a finer point on their productivity metrics. Industry standard SWAG estimates based on inexact metrics were causing cascading cash flow issues for their vendors, subcontractors, and the trades. On average, the invoicing cycle was running up to 90 days for approval and final payment. “Thumb in the air” percent completion estimates were submitted and run through a lengthy verification process. Payment lag often forced subcontractors into float or bridge loans whose carrying costs are then built-in to future estimates and ultimately passed on to the customer, ballooning project costs and causing overruns and inefficiencies. In an effort to shorten the payment window and the overall project schedule, and to mitigate the risk of variances against design intent, the search was on for a comprehensive process to aggressively track the amount and quality of work completed and reduce or eliminate human error in the process.

Proof of Concept

To automate the process of construction monitoring, progress tracking and variance detection, the Swinerton team asked themselves two fundamental questions. Is the technology available today to make this happen? If so, how do we integrate it into our workflow? To address the first, Law and his team did their homework – and their fieldwork – identifying an ingenious solution that demonstrates that construction technology is playing more than a game of catch up, it’s going for the win! 

To start, Swinerton partnered with Avvir, a NYC-based construction software company automating quality assurance for the construction industry through scanned-based AI and machine learning. Avvir’s platform proved to be highly effective at assessing reality capture data, catching errors and delays, and providing comprehensive information to stakeholders about the project completion. 

With the goal of producing high-fidelity 3D scans on a weekly – possibly even a daily – basis, Swinerton set a high bar for selecting a reality capture system. Photogrammetry-based systems produce high visual fidelity scans, but lack dimensional fidelity and require multiple set ups and registration of those set ups. Stationary laser scanners produce the visual and dimensional quality needed, but require the same cumbersome and lengthy set ups making it impractical for this application. A mobile continuous scanning solution would fit the bill just right. After extensive field-based comparison and vetting, Swinerton partnered with Kaarta, and chose its Stencil scanner as its mobile mapping system of choice. The Kaarta system allows them to move continuously and quickly, even on an active job site, and produces a single, unified scan of an entire floor with the coverage, visual clarity, accuracy, and density needed. 

To take efficiency and repeatability to even greater heights, Swinerton looked to automate the walk-through capture process by deploying a mobile robot to carry the laser scanner. Swinerton initially evaluated several ultra-rugged wheel and track robot platforms but ultimately chose Boston Dynamics four-legged robot, Spot. Named as it resembles and moves like a dog, Spot substantially outperformed other solutions in obstacle avoidance and it can climb stairs. With the addition of an arm that's currently in development, it can also open doors and fetch (for real). For fast and continuous scanning, the Spot mobile robot and the Kaarta mobile scanner were perfectly paired. 

[Quote from Eric about choosing the Avvir/Kaarta/BD stack]

The Application

Proof of concept and real-world application are two different animals, however, especially in the dog-eat-dog (or Spot-eat-Spot) construction industry. Can the hypothetical, when pushed into real world workflows, produce viable results? And more importantly, can it significantly improve productivity? 

Swinerton put the Spot + Kaarta + Avvir technology trio to the test at one of their construction management sites in the Bay Area, a $xxmillion, 4 floor, 18,400 total square meter [197,800 square feet] LEED Gold Certified medical center with 143 doctors’ offices and 116 exam rooms. Project progress would be tracked with the technology solution in parallel with the typical manual progress effort and reporting.

The workflow is as follows: Kaarta Stencil mounted to the Spot robot captures a scene. In these early tests the robot is guided, at a high level by a person, but once a route for scanning is finalized the robot can be taught to document a path at scheduled times in the day or week. The resultant point cloud is optimized onboard Stencil. The point cloud is transferred to the Avvir platform where it is compared to the design model (BIM) and to previous project scans. It is then followed by geometric analytics, such as construction progress and error identification, and a series of reports are generated. Swinerton is notified immediately through the online Avvir platform of construction errors, project progress and up-to-date installation quantities of materials on the project.

Prior to scanning, preliminary planning ensured a smooth process. The medical center site is scanned by each floor (approximately 4,600 square meters [50,000 square feet] per floor); planning out the scan path and preparing the space (i.e., open doors, turn on lights, remove hazards) to ensure each scan pass is efficient and effective. Stencil’s real-time position estimation and rapid capture of data within a range of 100 meters (328 feet) makes for quick and easy scanning of the massive floor space. 

Being able to replay live scan data and localize to previous scans, has been a game changer for Swinerton. This is especially true as the capture becomes more difficult, scans become larger, and capture time increases as the framing is closed-in and the maze of rooms becomes more complex

Tristen Magallanes, Innovation Analyst at Swinerton, is the project manager for the pilot program. What began as twenty-minute scans of open framing has snowballed into over forty-minute scans of dry-walled rooms with intricate MEP systems and segregated spaces,” explained Magallanes. 

Progressive scan data is fed into Avvir’s online portal and its proprietary deep learning algorithms. The algorithms measure the work put in place against the fabrication model/BIM, comparing quality of installed work in real time. Visualizations for each floor are broken up by content area (such as HVAC, Fire Suppression, Plumbing, etc.) for better management of data and specificity of reporting. Stakeholders can react to deviations almost immediately. The progression of scans is monitored, chronicling the spatio-temporal aspects of the construction from start to finish. The result is almost immediate site awareness, detection of risky “clashing” elements, and prompt feedback on scheduling and costs.