Four Types of Analytics and What They Mean to Asset Management
Everyone wants continuous improvement of their maintenance and asset management processes. It’s an admirable goal, but how do you know if those processes are really improving? You’ve probably got maintenance KPIs, but are they showing you what you really need to know? Real insight can only come from solid and reliable data. In turn, you need an analytics program to draw those insights out.
One of the challenges of enterprise asset management (EAM) analytics is making sure that the type of analytics being used is appropriate for your end goals. For example, logging and summarizing certain failure modes only requires descriptive analytics. You won’t get very far if you try to use the same techniques to unravel the root causes of those failures. You need to more complex analytics to determine that, such as diagnostic analytics.
In this blog, we break down the types of analytics so you can see how each is used in maintenance and asset management. Data scientists typically recognize four types of analytics: descriptive, diagnostic, predictive, and prescriptive. These are listed in order of complexity.
This rising complexity means that each level requires more effort to be put in place than the one before. The insights gained usually become more valuable as the complexity level rises, but this doesn’t mean that you should only aim for the top level of prescriptive analytics. The type of analytics used should be appropriate to the type of insights you’re trying to gain.
1. Descriptive Analytics
As the most basic analytics process, descriptive analytics is used to summarize and concentrate raw data so that it’s more easily understood. Descriptive analytics is concerned with facts and historical data. It typically does not try to determine cause and effect relationships, but simply describes a situation.
Descriptive analytics by itself does not make estimations, predictions, or suggest courses of action. Traditionally, doing so requires that human beings determine what steps, if any, should be taken. However, it’s possible to use the initial descriptive results as inputs in predictive or prescriptive analytics.
Many kinds of business intelligence fall under descriptive analytics. Descriptive analytics include dashboards, KPI reports, aggregations of performance data, and other familiar forms of business intelligence.
2. Diagnostic Analytics
Descriptive analytics tells you what happened. Diagnostic analytics, also known as root cause analysis, offers insight into why something happened. The process is used to discover anomalies, identify correlations, and determine cause and effect. This represents a deeper level of insight than descriptive analytics, which cannot usually identify causal relationships.
Diagnostic analytics is particularly useful in discovering hidden relationships and answering questions that descriptive analytics cannot. Diagnostic analytics can be simple if there is a single root cause that must be uncovered. However, when no root cause is immediately clear, the analyst must look at other data sets to see if a relationship can be established, even if the relationship is not always obvious.
For example, say that your descriptive analytics show you that a piece of equipment has experienced a 10 percent increase in failure modes compared to the same period the previous year. The descriptive analytics for that machine do not provide any clues: the PM schedule was the same, the production demands were the same, and no modifications have been made to the machine.
You need to look at other data sets and see if you can find correlations to discover the root cause. In this example, you discover that every unexplained failure took place at the same time as the machine next to it was running hotter than optimum. At this point, it’s worthwhile to investigate to see if the first machine was failing due to waste heat generated by the second machine.
3. Predictive Analytics
Both descriptive and diagnostic analytics give you insight into past events. Predictive analytics, however, focuses on predicting the future in terms of possible outcomes. The accuracy of the predictions is, of course, strongly related to the accuracy of the data fed into the solution to begin with.
Predictive analytics relies on statistical modeling and machine learning, a subset of artificial intelligence that improves automatically over time as it receives more data.
In maintenance, predictive analytics is usually seen as a way to improve on preventive maintenance practices. Preventive maintenance focuses on time or cycle-based maintenance practices. The downside is that crews are often performing “extra” maintenance work that isn’t necessary and doesn’t extend the life of the asset or increase its productivity.
Predictive analytics in maintenance can be used to build a predictive maintenance dashboard, which you can use to move towards a predictive model. In predictive maintenance, tasks are typically only performed when the models say it will be needed to prevent failure.
4. Prescriptive Analytics
This can be viewed as a subset of predictive analytics. Prescriptive analytics goes beyond simple prediction and ventures into mitigation. Simply put, prescriptive analytics not only tells you what the problem is and when it will occur but gives insight into why it will happen. From there, predictive analytics also shows courses of action and highlights the implications of each option.
You may think that prescriptive analytics sound complicated. You are 100 percent correct. It is an extremely complex field. In fact, it may be a mistake to think of it as just one field. Mike Gualtieri is a VP and Senior Analyst with Forrester. In his article, “What Exactly the Heck Are Prescriptive Analytics?” he notes, “Our research indicates that prescriptive analytics is not a specific type of analytics, but rather an umbrella term for many types of analytics that can improve decisions. Think of the term “prescriptive” as the goal of all these analytics — to make more effective decisions — rather than a specific analytical technique.”
No matter which type of analytics you’re using, it’s important to remember that the usefulness of any data analysis technique is strongly related to the quality and amount of data. Sparse, low-quality data will tend to yield sparse, low-quality results.