Q: Can you tell us about Performance Prediction and why it was developed?
Russ: “Businesses around the world are challenged to create effective investment plans—and deliver on them as well. They want to ensure that every dollar they spend is maximizing the benefit to their organization—and that they only spend what they need to.
However, things aren’t likely to go exactly as planned. Labor shortages, material delays, and extreme weather events can materially delay projects, making it very challenging for organizations to deliver their investment plans on time and within expectations. The current pandemic, for example, underscores how unpredictable events can have disastrous consequences on the plan and require us to be agile and react quickly in the face of such uncertainty.”
Danilo: “Performance Prediction enables organizations to gain accurate insights into the expected costs, benefits, and risks mitigated by their capital plans. It uses machine learning (ML) and statistical analysis techniques to predict the performance of investment portfolios based on historical data.
This unique capability can help organizations account for cost and schedule uncertainties and use this information to understand the likelihood of overspending or underspending—and make contingency plans accordingly. Performance Prediction ensures you are creating a plan that will deliver the expected outcomes, and it can also help you adapt to unpredicted changes once the plan is being executed.”
Performance Prediction enables organizations to gain accurate insights into the expected costs, benefits, and risks mitigated by their capital plans.
Q: How do organizations address this problem today?
Russ: “During the discovery phase of developing this product, we explored the different methods our clients currently use to address this problem. Some rely on extensive Excel spreadsheets to try to predict uncertainty based on historical data. Others overprogram, meaning they intentionally plan more than what their budgets allow, betting that some projects will be delayed. Both approaches vary greatly in their effectiveness from year to year. We knew there had to be a better, more automated way to get a consistent, accurate, and unbiased understanding of project execution risks.”
Q: How was Performance Prediction developed?
Danilo: “We worked closely with our clients on the development of Performance Prediction. Using their historical data, we applied machine learning techniques to develop prediction categories. These categories, as well as the overall prediction methodology, were validated using our clients’ actual investment data. Our approach is explained in a lot more detail in the white paper.”
Q: What are the benefits of Performance Prediction?
Russ: “Performance Prediction is part of the Copperleaf Decision Analytics suite and provides the following benefits:
- Increased forecast accuracy: provides a programmatic approach to predict portfolio performance, allowing you to use uncertainties to your advantage
- Quick scenario analysis: offers a simple, intuitive interface that allows users to quickly and easily change prediction parameters to better understand the impact of uncertainty on expected portfolio performance
- Proactive reallocation of funding and resources: provides the ability to update funding and allocation of executing portfolios based on new uncertainties and changing business conditions
- Natively integrated in the Copperleaf system: users can quickly and easily leverage all existing investment data in the system”
Q: Do I have sufficient historical data to use Performance Prediction?
Danilo: “This is a great question. There’s no set amount of data needed to use Performance Prediction. In fact, we don’t need a vast quantity of data to produce a highly accurate model. The important thing is to get started with the data you have. Once you begin using Performance Prediction, and as more projects are processed by the ML algorithms, the parameters will continuously improve, resulting in more consistent and reproducible predictions over time.”
Q: How do I calculate the prediction parameters? Do I need data scientists in my organization?
Danilo: “Organizations with data scientists can calculate the required prediction parameters themselves. Copperleaf also offers consulting services to assist those that may not have this expertise internally. Our team can help you get up and running quickly—by collecting the historical data, using machine learning techniques to define the prediction categories, and assigning investments in the plan to the appropriate category.”
When we tested Performance Prediction on our clients’ investment portfolios, the predicted spend accuracy was 99.8% compared to their actual 2019 portfolio spend.
Q: Can you tell us about the Performance Prediction team?
Russ: “We are a small cross-functional team of about 8 people that spans development, quality assurance, design, and product management. Two of our clients, Duke Energy and Bonneville Power Administration (BPA), were involved from the very beginning—helping us understand their challenges and validate our design prototype.”
Q: What excites you most about this product launch?
Russ: “If there’s one thing we’ve learned recently, it’s that the world is constantly changing around us, so being able to predict the impact of uncertainty and understand how likely a plan will be to consume resources and deliver the expected outcomes is really critical. When we tested Performance Prediction on our clients’ investment portfolios, the predicted spend accuracy was 99.8% compared to their actual 2019 portfolio spend. I’m super excited about these results and am really looking forward to helping more organizations benefit from this innovation! I’m also looking forward to continuing conversations with our clients so that we can incorporate their input into future enhancements of this product.”