Data to Decisions – Don’t Forget the Analytics

Eric Hansen ??01.30.23

Last week at the ASMC Army IT Day in DC, I heard the Undersecretary of the Army, Gabe Camarillo , describe the FY2023 service-wide initiative to unlock the Army’s data to support data-driven decision-making. This renewed commitment to data and a deliberate, quantitatively supported decision structure is music to my ears. As a former Army Operations Research Analyst, I’ve seen the focus and clarity that data-driven decisions can bring to the warfight. Yet, in listening to Undersecretary Camarillo’s comments, I was struck by the absence of the final ingredient necessary to complete the data-to-decision recipe: analytics.

Although the definition of analytics may differ widely, from my perspective, analytics (or analysis) is the transformation that must be undertaken to turn data into a meaningful resource to support any given decision. This transformation may be instantaneous such as when you decide whether you have time to drive through a yellow light before it turns red, or it may take months or years such as when the Army seeks to make major force structure adjustments. Regardless of how quickly the analysis takes place, it’s a step that can’t be excluded.

As students of military history, we know this. Sun Tzu, Jomini, Clauswitz and Boyd all include analysis as a critical step in the decision-making process. During World War II, the US Navy’s Tenth Fleet pioneered many of the techniques that we’ve come to know as Decision Science in their quest to eliminate the German submarine menace in the Atlantic. It’s time to revisit our use of Decision Science and re-focus on how to transform raw data into a resource to support meaningful decisions.

Below are four of the major lessons I’ve learned during a career in applying data analytics to DOD challenges. This list is not exhaustive and there are many others who have valuable insight into the analytics needed to support the Army’s data-driven decision-making initiative, but I view these as some of the basics:

1. Data Provenance. Most people today intuitively understand the phrase “Don’t believe everything you read on the internet”. Applying this lesson to data provenance is equally important. We must be highly critical of any data we use to support decisions that will ultimately affect the security of our nation. Where did it come from? How was it gathered? Has the data been changed in any way? Is the data complete? These are just some of the questions that must be addressed before we can have the trust that a data resource is suitable to support vital decisions.
2. Suitability. Does the data at hand support the decisions you’re making? Some data is simple to collect. Think about the time it takes you to run a mile on a treadmill at the gym. That data, your mile time, is right there in front of you and difficult to miss. Other data, like the intentions of an opposing force, is far more nuanced and difficult to collect. Regardless of the difficulty in collecting data, we must ensure that the data we’re using is germane to the decision to be made. This mistake is made more frequently than you would suspect when data that is recorded and collected for one reason is misapplied in the decision-making process. Not all data surrogation is problematic, but this practice should always be undertaken with caution.
3. Correlation is not Causality. At one point in my career, I conducted an analysis which resulted in the conclusion that the occurrence of roadside bombs in Baghdad was somehow related to swimming pools! After re-checking my math, my conclusion was the same: the data showed a strong positive correlation to roadside bombs and swimming pools. My error was in assuming that swimming pools somehow had an impact on the presence of roadside bombs. Reality was that the two simply existed in the same geographic space. Was their presence correlated? Yes, but a causal relationship simply did not exist. In using data to support decisions, we need to ensure that causality is not ascribed solely because a correlation exists.
4. Interpretation. In 1954, Darrell Huff wrote a fun little book called How to Lie with Statistics. While Huff’s approach to the subject was lighthearted, his message was not. Numbers and metrics can be presented in such a way as to influence the audience’s perception of the results. For example, my company, Significance Inc., was recently named to the INC 5000 list of fastest growing companies in America. We were informed that we were number 1,073 out of 5,000. The headline we internally generated in response, quite appropriately, announced that we were selected in the top 25% of that list. If you think about these two equivalent metrics, number 1,073 and top 25%, you’ll realize that your brain automatically perceives them differently. Everyone wants to be in the top 25% (or top 10%, top 5%, top 1%), but you’ll find that no one readily admits the desire to be number 1,073.

It’s an exciting time as the Army’s Digital Transformation turns toward unlocking the Army’s data to support data-driven decision-making. There is no doubt that the Army will be better for this effort and that our nation’s security will be enhanced. We just need to ensure that, as we begin to realize the value in data as a resource, we don’t forget the analytics needed to transform that data into something useful.

I’ll leave you with an analogy. How useful would it be if I gave you a barrel of crude oil to help you move your car down the highway?

Oil –> Refinery –> Gasoline is no different than Data –> Analytics –> Decision.?In both cases it’s the middle step that’s necessary to transform the raw resource into the valuable end result.

Significance has invested in an expert team to support its clients in the DoD across a wide range of Decision Support and Analytics initiatives.?Reach out to me if you’d like to learn more.??[email protected].