April 2024
by Karl Vantine, Chief Customer Officer at Contruent
Can predictive analytics help forecast possible risks with your capital projects?
Cost overruns and project delays plague even the most carefully planned capital projects. Unexpected problems and full-on crises can catch you off guard, forcing you to scramble to react.
How can you be better prepared? It comes down to deciding which risks are worth spending time on for mitigation plans and which aren’t.
All too often, organizations have historically relied on subjective and qualitative risk assessment when evaluating risks. While these can add value and are simpler to do, incorporating more objective and quantitative measures leads to increased precision and better results. Large, complex projects need a data-driven approach that empowers more informed decision-making around risks, including which risks need attention and which do not. That’s where predictive analytics enters.
Predictive analytics leverages your historical project data, enabling you to model potential risks, gauge their likelihood and impact, and create appropriate mitigation plans—so you’re better prepared for the unexpected.
Why Implement Predictive Analytics?
Construction projects have long relied on familiar risk assessment methods like expert judgment, fixed percentages and PERT to help identify and plan for potential risks. However, because they involve a degree of subjective or qualitative input, limitations emerge. Predictive analytics overcomes these limitations, though it requires a lot of data inputs to be most useful and effective. Here’s how:
At issue: Optimism bias
- Limitation: Optimism bias can creep in, downplaying the likely occurrence and/or severity of risks.
- Solution: By evaluating large amounts of data from different sources—historical data, jobsites, IoT-enabled equipment, weather reports, etc.—predictive analytics presents a more comprehensive, objective landscape of possible risks. With assessment rooted in data, a broader range of real-world, indisputable risks can be identified, including those that traditional risk assessment methods might miss.
At issue: Determining risk potential
- Limitation: Quantifying the likelihood of risk occurrence can be difficult using qualitative assessments.
- Solution: You can only predict where things are headed if your forecast reflects a good understanding of what’s happening with your project’s trends and what caused them. Quantitative measures provide this; the more evidence points you can supply, the better. Leveraging all your data allows you to model out the what-if scenarios with quantifiable risk potential and impact for each one, helping you decide which you need to worry about and which you don’t.
At issue: Factoring for the cost impact of risks
- Limitation: Subjective measurements can’t account for the differing cost impact of varied risks.
- Solution: Having data-driven forecasts and insights into project costs — especially from all your historical data — becomes a solid basis on which to set risk-adjusted budget estimates that are grounded in realism. This avoids the danger of underestimating costs and instead establishes budget requirements that account for risks.
At issue: Effective risk mitigation strategies
- Limitation: Subjective assessments lack the accuracy needed to forecast risks reliably and, therefore, the efficacy of mitigation efforts.
- Solution: Once likely risks and their effects are identified and quantified through predictive analytics, it can help to proactively tailor more focused mitigation strategies and contingency funding for each. These could include procuring materials from multiple vendors, negotiating risk assignments in contracts, and identifying opportunities for smarter resource allocation.
These culminate in higher confidence and certainty based on evidence-based forecasting rather than hard-to-measure qualitative approaches to project and risk management. This is especially crucial for mega projects that require a high level of accountability and financial discipline. Predictive analytics can substantially improve budget control and head off the costly surprises that can derail progress.
Now the question becomes: How do you implement it to get similar benefits?
Implementing Predictive Analytics
Collect and Prepare Data. Predictive analytics is only effective if its data is complete and reliable. It first requires identifying accredited data sources for previous estimates, industry benchmarks for similar projects, and external factors like weather pattern data, cost trends for labor/material/equipment resources, and economic reports.
Quality matters. It must undergo a cleansing process to remove duplicate data, fill in anything missing and fix mistyped values. Once cleaned, does that data make sense? Look for extreme deviations or anomalies that could indicate errors or random events. Then, standardize everything by ensuring specific types of data are in their expected format, such as units of measurement, so that they can be manipulated, calculated and analyzed accurately.
Think of it this way: better quality data inputs will deliver better quality predictive outputs.
Select and Train a Modeling Technique. What is the appropriate modeling technique(s) for your predictive analysis? There’s a wide range to choose from; the one(s) you choose depends on your project needs.
For example, if you want to simulate a wide range of possible outcomes to ensure the right contingencies are in place, use Monte Carlo analysis. To determine how much certain variables influence others (how a supply chain glitch impacts project costs), consider regression analysis, which taps into historical data to identify these interdependencies.
These are just a few examples of how different ones can be used for specific needs.
The effectiveness of each one will depend on data availability (and quality, as we mentioned above). The more data each model has to work with, the more comprehensive your analyses will be.
Going forward, improvements in artificial intelligence (AI)/machine learning (ML) open the possibilities of extracting more value and deeper insights from historical datasets. Using AI/ML-enabled models can help improve the predictive accuracy of project outcomes and detect intricate relationships among variables that may be overlooked through traditional methods.
Simulate Scenarios to Identify Risks. Now run the what-if scenarios. Using those trained models, simulate various project scenarios while adjusting the likelihood value of each risk and testing different risk combinations. What key risk drivers emerge from these simulations? This step helps you identify the most likely risks and determine their corresponding impact on project costs, schedules and scope.
Develop Mitigation Strategies and Contingency Funding Plans. With key risks and their predicted impact identified, you can now move to the proactive stage of risk management: developing and implementing risk mitigation strategies. This is also where you can use the model’s predictive output to determine how to allocate contingency funds appropriately. So, rather than reacting to a surfacing risk down the line, you’ll have a chance to proactively strategize workarounds and backup plans before they occur, giving you more ownership over risk management and, therefore, more control over project outcomes.
A caveat to keep in mind: Once you’ve done all this modeling based on objective and quantitative data, resist the urge to apply an optimistic bias. This isn’t the time to look for justification to ignore the data, even if it’s not a rosy prediction. It can be tempting to assume that the estimate or project won’t get approved without an optimistic story — especially if there’s strategic value for the key project stakeholders, whether a major corporation or a municipal or national entity.
Monitor the Project and Compare it to the Model. When construction is underway, monitoring real-time progress and comparing it to the model’s predictions are a must. Are tasks being completed on time? Are costs on track with where they should be? Are any risks materializing?
This monitor-compare process helps detect any deviation from the model. Discovering a potential issue early gives teams a distinct advantage in risk-impact management: Examining what’s behind the disparity allows them to strategize the next steps, make adjustments, or implement the appropriate contingency plan. Updating the model accordingly further improves its accuracy.
It’s an ongoing loop that ensures your predictive analytics remains an effective tool for data-driven decision-making and improved outcomes.
Predictive Analytics Needs an Analytics Solution
How confident are you in being able to forecast future possible outcomes successfully? Do you have the right insights to manage risk impact and contingencies effectively?
Leveraging the power of predictive analytics helps answer these questions. A robust analytics solution supports this, integrating volumes of project data and cycling it back into the model. Only then will you get the insights to make data-driven decisions to anticipate, model and manage risk impact.
As you consider such a solution, learn how Contruent Enterprise can help you improve your risk management processes. Our value comes in our ability to mine project data and historical trends for insights to augment your contingency planning and risk management practices. To learn more, request a demo today.