Prescriptive analytics

Prescriptive analytics

Quick definition

Prescriptive analytics uses data and machine learning to provide recommendations for the next steps a company can take for growth or optimization.

Key takeaways

Prescriptive analytics uses machine learning and artificial intelligence to examine possible courses of action for a company or for improving customer experiences and then put the best course of action into effect without human intervention.

Prescriptive analytics builds on the foundation established by descriptive and predictive analytics to help companies make better decisions based on what they anticipate specific customers are prone to do.

Companies need to have a good understanding of their business goals and the particular use case for the data to gain valuable insights from prescriptive analytics. If a company doesn’t understand the problem they’re trying to solve, or the variables involved, they will not receive viable recommendations. .

Rohit Gossain is a product manager for Adobe Analytics, currently working on Customer Journey Analytics, which is an analytics service for Adobe Experience Platform and allows clients to analyze multiple data sources directly in Experience Platform. Rohit has a bachelor’s degree in computer science, an MBA in marketing, and a master’s degree in data science. Previously, he was a senior product manager for data science at Shutterfly, and has worked for Expedia, OpenTable, and Lenskart in the last 10 years.

What is the business benefit of prescriptive analytics?

How does prescriptive analytics differ from predictive and descriptive analytics?

What is a prescriptive solution?

What problems do companies run into with prescriptive analytics?

How has prescriptive analytics changed over time?

Q: What is the business benefit of prescriptive analytics?

A: Prescriptive analytics provides multiple benefits to a company's decision-making process. By using machine learning and artificial intelligence (AI), and by making sure they have clear business rules defined, prescriptive analytics can look through all the possible optimization routes and tell them the best course of action.

One of the most important use cases for prescriptive analytics is pricing strategy. Most companies define their pricing based on multiple factors. They consider the product value, the marketing costs, and the product development cost. They have their margins, including employee resources and costs. Based on these various numbers and the profit the company wants for shareholders, prescriptive analytics works backwards and optimizes all those variables to come up with the final pricing of the product.

Q: How does prescriptive analytics differ from predictive and descriptive analytics?

A: The three types of business analytics serve different roles and perform different functions, but they all work together and build on each other. Descriptive analytics is the foundation for all data analysis, and the first step of the analytics process. You can’t use prescriptive or predictive analytics until you have strong descriptive analytics. Descriptive analytics inform you about the current state of an organization. It looks at past data and provides information about how the company currently stands.

Predictive analytics is the second step of the data analytics process. It indicates where the organization might be in the future and what changes the company is likely to see based on the information derived from descriptive analytics. Predictive analytics also allows a company to anticipate customer behavior, which enables the company to make changes based on a customer’s propensity to take certain actions. It is an essential part of improving the customer experience and understanding a customer’s needs.

Prescriptive analytics is the third and final analytics stage. To successfully run a prescriptive model, an organization must be well-grounded in both predictive and descriptive analytics. Prescriptive analytics looks beyond what will happen next to show why certain actions or changes might take place. It also indicates what decisions a company can make to better ensure a specific outcome, and how to avoid potential problems with certain courses of action.

Q: What is a prescriptive solution?

A: A prescriptive solution is a solution that depends on descriptive and predictive analytics to offer a correct recommendation to a business problem by optimizing the variables involved in the decision.

There is not one single solution that works across the board for every company. Companies often create a customized solution, pulling together different tools based on the needs of individual teams and the company’s business functions. Many companies actually build their prescriptive model in Excel, but they use Tableau or another business intelligence (BI) solution for their descriptive analytics, and then they use machine learning for predictive analytics.

The solution optimally changes based on the business function. The finance team might not be using machine learning — they might be running a similar model using less data in Excel. The marketing team might be using a marketing technology platform to create a prescriptive model. The data science or engineering team might be using cloud technologies for their analytics. Every solution is different.

Q: What problems do companies run into with prescriptive analytics?

A: There are four main challenges related to prescriptive analytics. The first issue companies experience is trying to do prescriptive analytics without a strong foundation in descriptive and predictive analytics. For example, a company might want to expand their product line. If they haven’t used descriptive analytics to understand their current revenue, or predictive analytics to establish a goal, prescriptive analytics will be unable to recommend what steps the company can take to viably and successfully add more products. You can’t predict or prescribe future action if you don’t have past data on which to base your predictions or prescriptions.

The second challenge companies face is identifying the right business problem to solve. There are multiple problems you could solve, but you have to define which problem will be the best to focus on to achieve the optimal results. Once you've defined the right problem, then you can go back and see what other variables are involved and optimize for those particular variables.

The third challenge goes hand-in-hand with the second. A company has defined the problem they need to solve, but they are not aware of the variables involved with that problem. For example, when developing a pricing strategy, a company may forget to include the external factors like export and import cost. Neglecting to define all the necessary variables will negatively impact the whole model, as well as the whole recommendation. It will not provide the best answer, and the company will be left pursuing a course of action based on erroneous or incomplete information.

The fourth challenge revolves around having the right people and tools working with the data and analytics. Companies should hire domain experts who understand both the data side and the business side. Many companies work with data scientists who have a strong grasp of the data itself, but they are less savvy about the business problem they need to solve. Hiring domain experts with deep business professionality leads to more strategic success.

Companies also need to have high-quality data governance and storage to ensure they are working from the most relevant and up-to-date information. If they can’t trust their data, they can’t make helpful recommendations. And if they don’t have the right technology, they can’t successfully run a prescriptive model. But by working closely with domain experts to find the right solutions and investing in their data, they will have a strong foundation to begin optimizing their decisions.

Q: How has prescriptive analytics changed over time?

A: Because of machine learning and data science, we are able to solve problems that we couldn't solve earlier. In the past, companies had more limited data and variables. Even 10 years ago, if someone wanted to solve a basic problem through prescriptive analytics, they might be using an Excel sheet where they only had ten variables to optimize for. And now, they might have 100 variables.

If they have 100 variables, they bring more complexity to the problem and in the dataset. Problems we couldn’t solve 10 years ago have now become manageable through machine learning and neural networks, which are subsets of AI. Through machine learning and AI, someone could analyze 100 variables involved in the business problem in real time to discover the best course of action. The system can then automate the best course of action by itself, without needing human interaction to put the plan into effect..

People also view