Chapter 7. Prescriptive Science and Decision Making

Predictive analytics scales the power of analytics and decision making to a paramount level. We can consider our daily life as an example. If we have an answer to the 'when' question, it can help us take better decisions to secure our future. Visibility into the future makes life easy for everyone. The same holds true for problem solving in decision science. The nature of the problem can be descriptive, inquisitive, predictive, or prescriptive. Prescriptive science or Prescriptive Analytics answers the question 'So what, now what?' and aims at improving the outcome of the problem. We often ask this question after we see an issue in the routine operations.

Prescriptive analytics is a fuzzy transition from the combination of descriptive, inquisitive, and predictive analytics. The problem reaches a point where we continuously iterate through different questions either to recover from a disaster or further improve the solution. In this chapter...

Using a layered approach and test control methods to outlive business disasters

Prescriptive science results from the convergence of descriptive, inquisitive, and predictive analytics. It is used as a layered approach and iterated till the time we reach a promising solution. To understand the concept lucidly, let's simplify it by reconstructing the abstract and ambiguous words in a layman's way.

What is prescriptive analytics?

Prescriptive analytics helps us answer the question 'So what, now what' in the problem solving exercise, that is, it helps us improve the outcome. It is the final layer in the problem solving stack that results from the convergence of the previous three types: descriptive + inquisitive + predictive.

We'll consider a very simple example to study this in more detail. Consider a telecom giant (say AT&T, Verizon, and so on) who provides multiple services such as broadband connections, IPTV, mobile telephone connections, and so on The director of the Customer Experience...

Connecting the dots in the problem universe

If we take a look at the problem universe that we designed for our telecom giant's use case, we can see that we have identified multiple problems. A few are basically the hypotheses that we might have missed while brainstorming for HDH; additionally, due to limited visibility of the domain, we might have missed out during the creation of DDH. After reaching prescriptive analytics, we would have ideally finished one complete iteration of the problem. At this point, we will have reinforced our understanding about the problem and domain better. We can leverage this to improve the problem further with another round of iteration, but in some cases, we might find out a few new problems that are completely different from the current problem we are trying to solve. For our use case, refer to the following problem universe:

The problems highlighted in red are actually new problems. Staff training and customer education are smaller problems that can be...

Story boarding - Making sense of the interconnected problems in the problem universe

The problem solving journey is very long and iterative in nature. Once we have designed a version of the problem universe, we will have a fair understanding that solving the problem will definitely take much longer than we anticipated. The process being iterative in nature doesn't mean that seeing tangible results should take time. It becomes increasingly important to evaluate the value from results derived so far and the results expected in the roadmap designed.

In decision science, storyboarding (the art of conveying the results in the most lucid way) is of paramount importance. In fact, it is in any problem, but here when we have a holistic view of the problem, we know that the problem keeps evolving. At every step, it becomes increasingly important to realize the value delivered and the value that will be delivered by solving the connected problems. Storyboarding requires drafting the results in a sequence...

Implementing the solution

The final step in our problem solving journey is to implement the solution. We discussed about the implementation that would roll out immediately as per our plan; that is, developing a solution where the agent will be notified in real time whether a customer will repeat the call in the next 48 hours. To make the solution more actionable, we can design an association rule table that will calculate the association between different call reasons (categories). This will come in handy when the agent is notified that the customer will repeat in the next 48 hours for a different call reason. The association rule table can then be leveraged by the agent to understand the most probable reason for the repeat call and then take additional steps to mitigate the chances of a repeat.

Once we have all the previous steps in place, implementing the solution is just following the steps we designed as a part of our journey, that is, solving a business use case end-to-end. When we finish...

Summary

In this chapter, we touched base on the final stage of the problem solving journey-prescriptive analytics. We borrowed a hypothetical use case from the customer experience team of a telecom industry. To solve the problem, we leveraged the layered approach in problem solving and quickly traversed through the descriptive, inquisitive, and predictive path. You then learned about prescriptive analytics and saw how it can be leveraged to enhance the outcome and answer the questions: So what, now what?

To learn the decision making process in decision science, we studied how the problems iterate in reality and how we can solve them better by revisiting the DDH and HDH matrix in our problem solving framework. We further studied how problems are interconnected in nature and how the evolution of a problem can be studied, captured, and proactively solved in a structured way by designing the problem universe and connecting the dots to make more sense of the problem. Finally, we studied how to...