Karthik Venkatasubramanian, VP, Data Science & Analytics | June 30, 2021
Imagine you are planning a family event, a camping trip, or figuring out what to do for the weekend. For most people, checking the weather forecast is a key step before making plans.
While it’s something we do without giving it much thought, how many of us can answer two simple questions?
Surprisingly, not many know the answers to these questions. Forecast accuracy depends on how far into the future the forecast is being made—a one-day forecast is far more accurate than a 10-day forecast. Some meteorological departments have more complex models than others. It’s easy to see why forecasts are difficult. There are several sources of data, most of which are dynamic.
Historical data is the safest bet in terms of providing data to learn from. Modelling and forecasting determine what the weather is likely to be today or seven days from now.
Ultimately, it’s fair to say that most people would prefer to base their decisions on a weather forecast versus no information at all. While the weather predictions may not always be right, there would be no basis for decision-making or planning without these forecasts.
Project teams actively monitor and use data for decision-making. Examples include: Ball-in-court reports, tracking what is outstanding and overdue for approvals, and updating schedules and budgets with the latest information. Often termed “descriptive analytics,” this sort of reporting and dashboarding is considered table stakes these days.
But most of these metrics are “lag indicators;” they focus on what has happened, rather than what can happen. These metrics allow decision-makers to course correct and respond reactively but often with a limited set of options.
Decision-makers can chase overdue RFIs, follow-up on delayed design reviews, and can take the things that need to be actioned ASAP to site meetings. But, what if they could do more?
Akin to weather predictions, construction organizations are looking for lead indicators to help them look ahead and base decisions based on what might happen. Imagine using machine learning (ML) to not only predict project delays, but to also identify the activities that might be delayed.
What if you could look at why these predictions are made and what actions you might take now to mitigate the predicted delay? We are striving to answer these sorts of questions with our newly launched AI-powered solution, Oracle Construction Intelligence Cloud Advisor.
From the time the first schedule is created, the Oracle solution actively hunts for trouble based on past and present conditions to predict the probability of delays. The Oracle tool looks for patterns in the current project data and tries to identify and correlate the patterns to what has happened in the past.
The predictive tool continuously monitors the data that is created on the project to refine these predictions as the project progresses.
Like weather forecasts, the accuracy of these predictions will depend on the availability of historical data and the quality of available information. In general, there is continuous focus on improvement of data quality, like the use of DCMA checklists in most scheduling platforms, to ensure that scheduling errors are picked up early on.
Progressively, ML will also be used in identifying logical and sequencing errors—as well as making recommendations for labelling activities aligned with best practices for easy identification and analysis.
So, what does this mean? There’s almost an infinite number of reasons projects go sideways:
Whatever the reasons for the delay, with machine learning and good models, you are actively monitoring and managing the likelihood of delay. You can take proactive steps to mitigate or minimize the impact of unforeseen issues and ensure there are fewer surprises in general when delivering projects to a schedule.
Schedule predictions will also benefit from the recent wave of standardization. Predictions improve the more you standardize your processes and activities. Many organizations are integrating disparate systems. Because of this, there are more integrated datasets from several siloed systems that provide a holistic context for making predictions.
New and predictive technologies may not always be accurate, but because AI and ML learn and improve over time, having access to this technology is better than the alternative. Both data quality and prediction accuracy will improve with time.
Movies, music, self-driving cars, predictive text in mobile, image recognition, and shopping recommendations have become better over time. Five-day weather forecasts are also now accurate about 90% of the time.
Digital assistants also understand us better now. While the application of these technologies is novel and new to the industry, these solutions have existed for many years and have been successfully deployed across several industries and in consumer products.
The increasing digitization and datafication of the industry, coupled with ever increasing cloud computing power, means that the possibilities are limited only by our imagination as AI and ML become increasingly accessible. Not tapping into the full potential of these tools could be a missed opportunity.
Learn how to make the most of your growing volume of data to improve on-time, on-budget project delivery in our business brief, "Predictive Artificial Intelligence in Construction."
Oracle Construction and Engineering, the global leader in construction management software and project portfolio management solutions, helps you connect your teams, processes, and data across the project and asset lifecycle. Drive efficiency and control in project delivery with proven solutions for project controls, construction scheduling, portfolio management, BIM/CDE, construction payment management, and more.