Making smart asset choices from imperfect asset data
April 29, 2022
April 29, 2022
How using machine learning to fill in data gaps allows for more efficient operations and better stakeholder trust
Since the early 2000s, we have seen exponential growth in the demand for and consumption of data. That data comes from social media, smart apps, and artificial intelligence. It also includes new innovations that combine data inputs such as Flood Alerts, Constraints Checkers, and Route Selector. Data plays a critical role in Asset Management from helping determine asset criticality, to making key investment decisions regarding repair, replace, or refurbish—especially for asset intensive industries such as utilities, civils, and transport.
The importance of data-driven decisions is ever increasing as large organizations are turning more to digital solutions to keep pace with the demands of making well-informed investments.
The consequences of making the wrong investment decision can take years to rectify, impact operational performance—especially in regulated markets such as water—and have significant financial repercussions. Therefore, the management of data quantity and quality is becoming even more important for asset managers. Often organizations fail to recognize that the base asset data is an asset itself. Unfortunately, that means maintaining the accuracy and completeness of the base dataset is often an afterthought and overlooked. And it shouldn’t be.
The wide scale adoption of digital tools to inform long-term planning and various investment scenarios has highlighted the need to improve both asset data capture and data maintenance. This is particularly true for the UK Water sector where the regulator Ofwat is recommending long-term investment scenarios with adaptive pathways as part of their latest consultation. This firmly puts the emphasis on improving data and data insights to ensure better investment decisions are made in the future.
The cost to improve the quality of base data and to minimize data gaps can be incredibly expensive. Driving robust data capture can take several years to develop—if not decades for some industries.
Therein lies a key question. Can we make good decisions based on noisy data or incomplete datasets? This is not a trivial ask and a struggle for many utility companies across the UK.
To overcome the challenge of making smart, sustainable investment decisions in the long term with less-than-ideal data is difficult, however, it’s not impossible. There are various techniques to reduce data gaps and assumptions made. A value for money alternatively to collecting new data is to use various infilling techniques, this can be rules based, max/min values, or expected values. The key advantage of this technique is that it uses the existing data to extrapolate missing data, therefore not requiring further new information. Studies have shown that datasets that have been infilled up to 40%, still drives similar conclusions and results as if the complete dataset was used in the analysis.
This is very powerful. Being able to demonstrate that despite missing data, one can still make highly effective decisions—even with almost half of the base data missing—is an asset-management and a monetary-management game changer.
The wide scale adoption of digital tools to inform long-term planning and various investment scenarios has highlighted the need to improve both asset data capture and data maintenance.
I must admit, there is a stigma associated with infilling techniques, which has negative connotations regarding the accuracy of the resultant analysis. Some argue infilled data points are not genuine data points and it is only a best guess, bringing into question the uncertainty of the results. There is no denying that this is factually true, and without a massive financial investment there is no way to get around this statement.
Our teams have found that you can “tag” infilled data points and perform uncertainty and sensitivity analyses to get an appreciation of the ambiguity ranges when using infilled data vs incomplete data. When applied appropriately, statical infilling can greatly bolster confidence in investment decisions.
There are infilling techniques that are fundamentally not suited for certain datasets, so careful consideration is needed. An experienced data analyst will know what types of analysis best fits the type of available information.
With increased interest in the ability to deploy data to deliver better investment insights, I present two examples below that show the effectiveness of data-driven asset management. Both examples showcase how data and data insights played a critical role.
More often than not, large asset intensive organizations have too much asset data. It can be overwhelming. In fact, studies show companies are collecting data for the sake of collecting data. Essentially building what is sometimes referred to as “Data Graveyards.” Collecting new information is not always the most cost beneficial strategy; rather, analyzing existing data to drive useful insights is more productive.
To get started on this journey, I recommended organizations start with reviewing existing data and its quality. Try to use infilling techniques to derive outputs instead of continuously collecting new data to acquire the perfect datasets. Effective decision making can be achieved with partial datasets—although it requires a lot of patience and critical thinking. This will lead to more solid investment decisions and improved stakeholder confidence—a win-win.
Our team offers a number of industry-leading data analytic tools in growing demand in our current costs-conscious environment. Find out more about Stantec’s Digital services.