The Data Journey - Part Four: Prescriptive Analytics

November 30, 2020 Claranet Limited

Mike Fowler

Every business has data. Some businesses can spin gold from their data but others merely collect it in the hopes of getting value someday. While there is no easy path to maximising the value of your data, there is at least a clear journey. In this final post of a four-part series we'll look at how we can use our predictive analysis to shape the future. 

Building a predictive model is not a one-off process. It's one that requires continual assessment, monitoring and retraining. The litmus test is how good your real predictions with never-before-seen data are. You may have built a model that measures an impressive 99% accuracy with training data, but if every prediction it makes with new data is wrong it's utterly worthless. Equally, a model that at first makes stellar predictions may degrade as the underlying patterns in the data drift. 

With time you'll develop a number of different models that make predictions for different aspects of your business. You'll have continued to track how well these models are at separating the signal from the noise and this leads you to our final stop, prescriptive analytics. 

Our predictive models are all based on historical data and if we've proven that these models are an accurate representation of our business, we have the power to start applying the scientific method. We can state a hypothesis, simulate the scenario in our data and see what the implications are through our predictive models. 

One of the best examples of this in practice is documented by Michael Lewis in his book Moneyball. It follows the manager of the Oakland A's baseball team, a team with a low budget that was able to match the performance of the team with the highest budget, the New York Yankees. In 2002 both teams won 103 games but the cost per win for the A's was $0.39 million compared to the Yankees' $1.22 million, a saving of $85 million over the year. 

How did they achieve this? Their models uncovered predictive relationships that did not align with accepted wisdom. Running many experiments using the data from historic players, they were able to further narrow down exactly what attributes they needed across the team. They were able to exploit this to hire players that the market considered to be of low value yet possessed the attributes that would lead to the wins their models predicted. 

We can achieve similar benefits in other industries such as retail. Knowing product demand is useful for forecasting, but what if you can determine what causes the demand? Could you exploit this to increase demand for unpopular products or extend the demand for high revenue items? Ultimately, prescriptive analytics allows you to optimise your business. 

While this is the end of our data journey, you can see that there is no real end once you start to be prescriptive with your data. You may even start to explore the growing field of decision intelligence using your historic data, predictive models and prescriptive scenarios to make data driven decisions for your business. The limit to the value of your data is the amount of effort you put into extracting that value. 

The journey to data transformation begins where you stand today. Whether you have started your journey or it's an entirely new adventure, the Claranet team is here to accelerate your transformation. 

Previous Article
Exposing Costs with Infrastructure as Code
Exposing Costs with Infrastructure as Code

As a discipline, FinOps is often seen as a central function of an organisation – something managed by core ...

Next Article
The Data Journey - Part Three: Predictive Analytics
The Data Journey - Part Three: Predictive Analytics

Every business has data. Some businesses can spin gold from their data but others merely collect it in the ...