When is it Important to Operationalize Data Science? – A March Madness Example

As decision-making velocity has become too fast for people to manage, businesses are achieving gains using Decision Intelligence to operationalize data science and, ultimately, automate their decisions. Decision automation is challenging because of the pace required to harmonize data, run data science models, and develop insights that engage users to make business decisions.

With March Madness about to begin, I wanted to follow up on the basketball analogy for operationalizing data science that I shared last year by detailing some of the predictive decisions that AI can make around basketball, then extend that as an analogy to business.

To quickly recap what I wrote last year: My favourite time of year is mid-March when I decide on my predictions for the NCAA men’s basketball tournament. About ten years ago, I developed a predictive model to select the national champion. I review the latest conference tournament data; look for injuries and recoveries among key performing players; rerun the predictions on the adjusted data; and fill out the bracket to make 63 decisions, as there are 63 games to predict in a 64-team single-elimination tournament.

Over the years, I’ve found the important predictors of success in the tournament to be:

• Free throw shooting ability
• Great team defense
• Solid backcourt players

Currently, UCONN has predictive KPIs that support the best performance categories in shooting, defense, and taking care of the basketball. Other teams showing strong promise in these categories include Arizona, UNC, Auburn, and St. Mary’s. If I were to fill out a bracket today, I would list UCONN as the winner.

Let’s extend this approach to business and suppose that the best performance will be predicted by demand planning ability, great overall equipment effectiveness, and solid customer service. There are many KPIs that support these predictions and their different planning horizons.

My predictive modeling considers the relative strength of schedule and adjusts for schools like Princeton that achieve high predictive KPIs against weaker competitors. Don’t get me wrong though, the model does indicate Princeton as a high performing team. Extending this to business, we find that some customers are harder to serve than others because of geography, or the horizon or volatility of their ordering patterns. So the modeling adjusts for harder strength of schedule as your business unit KPIs are likewise adjusted for more demanding and higher volume or margin customers.

Finally, keep in mind that I might change my selection of UCONN as the winner depending on their performance in the final weeks before the tournament. Likewise, if we extend my predictive modeling approach to business, we find that customers can change their orders, raw material can be delivered late, and variations in quality can produce higher or lower yields than planned. This would in turn change our decision-making basis.

The pace and scale of decision making has changed

Operationalizing college basketball predictions is fun. Operationalizing data science for your business is infinitely more challenging and significant. While these 63 Final Four predictions are made once a year, your business predictions include tens of thousands or even hundreds of thousands of decisions that are made on a daily, hourly, or real-time basis.

This new status quo makes operationalizing data science critical, because the demand for business decisions is simply outpacing the ability of our organizations to make them. This is due to changes in:

• Velocity: Customer expectations, shorter planning cycles, and market volatility are accelerating decision velocity at an unprecedented rate
• Volume: The scale and frequency of decisions have increased exponentially
• Variety: Driven by complexity, the type and nature of decisions to be made now varies greatly versus past experience
• Variables: The number of factors to consider in making decisions continues to rise
• Volatility: The uncertainty and the variability surrounding decisions has continued to increase, from major events like the Covid-19 pandemic to localized events such as weather or variability in customer demand

Fortunately, Decision Intelligence — the digitization, augmentation, and automation of decision making — has allowed companies to transform how they make and execute decisions.

Business operates in real time. Decision making needs to catch up.

Decision-making processes need to occur in near-real time to match the pace of business transactions occurring in real time. When those processes don't keep up, cracks in the business appear — and cash falls through those cracks.

Transactions occur in real time, supporting the process of cash inflow to the business. When decision making processes don’t keep up with transaction volume, raw material purchasing is suboptimal, material movements priorities aren’t understood, and overall operational effectiveness is reduced, ultimately resulting in lost sales. This slows the inflow and increases the outflow of cash.

These gaps are growing increasingly bigger and bigger over time because of the five “Vs” mentioned above. Integrating data science into your decision making processes is a key step to repair the cracks and prevent them from recurring.

The best approach to starting this process is to examine your decision making workflows:

1. Find the “cracks” (decision-making gaps) in your business: Where are siloed, inefficient, manual/time-consuming, or disconnected processes resulting in lost value or missed opportunities?
2. Measure the value that is being lost due to these gaps
3. Create a strategy to repair these gaps by: a. Understanding the decision making processes that can be digitized b. Working with data scientists to augment those processes with predictions
   c. Deploying a Decision Intelligence platform like Aera Decision Cloud™ to digitize and automate these processes
4. Monitor the decision lifecycle, including measuring the value saved or gained with Decision Intelligence vs. the dollars that were falling through the cracks
5. Operationalize data science to learn how to improve your use of Decision Intelligence and increase the value of the recommendations delivered by your AI platform for decision making

Our Aera Technology Professional Services team works with businesses and system implementation partners to follow these steps every day — designing a value-based strategy, developing AI skills, and deploying the Aera platform to help you keep up with the pace of today’s decision making reality.