Prescriptive Analytics is quickly becoming a critical tool for businesses of all sizes. It allows real-time analysis and provides valuable insights into customer data, trends, performance metrics and more. With the ability to analyze billions of data points instantaneously, Prescriptive Analytics can help organizations better understand their customers’ needs in order to make better decisions. As disruptive technologies continue to reshape the marketplace - such as artificial intelligence (AI), machine learning (ML) and cloud computing - Prescriptive Analytics will become even more essential for companies looking to remain competitive in today's ever-changing landscape. In this blog post, we'll explore what Prescriptive Analytics is, why it matters and how its applications are changing the way businesses operate.

What is prescriptive analytics?

Prescriptive Analytics is a type of analytics that uses advanced and complex algorithms or techniques to recommend the best course of actions for an organization to achieve specific goals. These techniques make use of data mining, machine learning, statistical optimization and data visualization to provide decision makers with actionable insights. The purpose of performing Prescriptive analytics is to find out what actions are required for a specific business to achieve certain goals which makes it more valuable than ‘descriptive’ and ‘predictive’ analytics that only evaluates what has happened in the past & what might happen in the future, respectively. Prescriptive analytics is often considered as the last stage of business analytics, the first two being the descriptive and predictive analytics respectively. The observations and analysis performed in the first two steps provide a basis for advanced decision making using the recommendations generated in the final step that is ‘Prescriptive analytics’.    

How prescriptive analytics works?

As defined, Prescriptive analytics is based on techniques developed around data analytics, machine learning, mathematical optimization (statistics) and data visualization (to showcase the trends and results). All these processes combine together to provide recommendations or suggestions that help businesses in data driven decision making to achieve their goals. 

Defining a problem statement

As an initial step, it is very important to understand the challenge or a problem that a particular business is trying to solve. As it is often said, ‘it is crucial to know what you want to achieve in the end’. A clear problem understanding is must for Prescriptive analytics to be helpful and worth investing time and effort. 

Data 

In any form of analytics, data is the key. There are many phrases that underline the importance of data in today’s world like ‘Data is new oil & Machine Learning is new electricity’. Well, it cannot be denied. Same is the case for Prescriptive analytics. 

It all starts with ‘Data Collection’ where data relevant to the identified problem statement is collected from all the possible mediums (both online & offline) and in all possible forms (text, image, audio). ‘Data Preprocessing’ is the next step in the data journey where the collected data goes through a series of steps and is made ready where ‘Data Analysis’ could be performed. In data analysis, the main goal is to find relationships among different entities of data and derive some key insights using statistical concepts mainly and simulating different scenarios. 

Optimisation

This is the step where Prescriptive analytics plays its role to solve the business challenge. After going through the turmoil of data collection and analysis, the process of optimization starts which uses rigorous mathematical models and Machine Learning to define the most efficient solution to a described business problem.

The key terms of this process are

Objective 

A metric(s) evaluation of the business output. In business terms, it could be total revenue, profit or number of acquired new customers etc.   

Decision variables 

These are input variables that a company can change and it has a significant effect on the objectives. For example, the price of products, opening and closing time of business, choice of mediums to promote their products and/or services etc.

Constraints  

These are some rules or boundaries that are defined by businesses that are considered before making the final recommendations for business actions. These constraints could be related to profit margins, product prices, inventory management or other processes involved in businesses and often have two boundaries i.e. maximum and minimum. 

These steps (parameters) provide a basis for mathematical optimization - a key concept in Prescriptive analytics, by using machine learning algorithms and coming up with optimized steps that a business can take in order to solve the presented challenges, which is also known as ‘recommendation generation’ in ML terms. 

Decision making (data driven)

This final step in Prescriptive analytics involves critical human skill i.e. decision making. Stakeholders need to understand the results provided by mathematics and machine learning, evaluate those options  and select the best course of action to achieve their business objectives. 

Types of data analytics

Business analytics (generally data analytics) is commonly viewed from three major Prescriptives i.e. descriptive, predictive and prescriptive analytics. Let’s have a detailed look at each of the types.

Descriptive analytics

Usually, this is the first step in any business analytics scenario. Descriptive analytics is considered to be the most initial form of business analytics. This analytics makes use of the past data to understand processes that happened in the past and what results they generated and depending on that, it makes informed decisions. Descriptive analytics helps in understanding the existing data by converting it into structured form where it becomes possible to understand it and visualize it in the form of charts, graphs (dashboards) to get a summary of events that happened in past, how frequently those happened and who played key role in those events etc. 

Following figures showing a basic sentiment analysis of different tweets gives an idea of how descriptive analytics can be helpful in initial understanding of data in the business analytics process.     

‍

Fig.1. Number of Neutral, Negative & Positive sentiments                            Fig.2. Percentage of Neutral, Negative &  

                                                                                                                       Positive sentiments.                                           

Fig.3. Word frequencies in tweets sentiment analysis.(Image source: - Kaggle)   

From a business point of view, descriptive analytics can help in baseline understanding of historical data, can help in gap analysis in the existing market, can help in understanding customer segmentations and provide in depth analysis of past trends that already have happened and what had been the impact of those events. All this information and charts help in developing a clear understanding of past events and can help stakeholders in devising a certain course of actions or provide the direction for their future decision making. 

Predictive analytics

Predictive analytics is the next stage after descriptive analytics where the purpose is to predict the future events by looking into historical data. Future events can vary as per various businesses and also as per the required goals e.g. predicting the sales figures, or number of new customers acquired as result of conversion campaigns or predicting the churn rate or expected growth economy for the next year etc. All these questions or scenarios represent the type of analytics that is ‘predictive analytics’. 

This is an advanced form in business analytics as state-of-the-art tools and technologies are required to find out the hidden patterns in data and how accurately those patterns can help in predicting the future outcomes. Predictive analytics make use of statistical and machine learning concepts like linear regression, decision trees, support vector machines (SVMs), Neural Networks and gradient boosting - to name a few.

All these algorithms require models trained on the training datasets and are being evaluated using validation and test data before deployed for practical use cases. Also the training of these models require significant computational power to process big amounts of data being fed to these models. In the end step various evaluation metrics are used to check the performance of these models including accuracy, p-values, F -scores and confusion matrix etc.     

‍

Fig.4. Confusion Matrix, showing actual vs predicted labels (Image Source:- Kaggle)

‍

With the amount of data being generated in today’s age, predictive analytics has become an essential part of businesses' decision making process. As per a report by MarketsandMarkets, the global predictive analytics market is expected to reach $23.9 billion by 2025 growing at a CAGR of 21.7% from 2020 to 2025. The stats show the clear adoption of predictive analytics that is on the rise and businesses are investing in it to gain competitive edge over the others. 

Prescriptive Analytics 

As described before, Prescriptive analytics is the third and final step in the business analytics process with the purpose of optimizing the best alternatives to minimize and maximize some objectives. Predictive analytics is great in providing multiple options that can be selected but optimizing those options is key as businesses focus on maximizing their profits by taking decisions ensuring them to hit ‘bulls eye’.

For example devising the best pricing or best advertising strategy to maximize the number of sales or minimize the loss percentage.  

Business analytics is the convergence of these three analytical approaches. Therefore,  In business analytics journey, it is important to understand how these three analytics work altogether thus aiding to devise a data driven strategy for businesses with minimum chances of failures. The figure below summarizes how these approaches work altogether in an organization in different time frames.

‍

Fig.5. Business value of analytics with respect to time (Image Source:- Krumeich)

Use cases of prescriptive analytics in action

In today’s world, it is hard to think about any sector or business that is not relying on the power of data that it generates or is able to use (open-source) and trying to use any form of analytics (descriptive, predictive, prescriptive) to not only compete but also maximize their profits. When we talk about applications of these analytics approaches, prescriptive analytics is relatively new and hence a less mature approach. It, though, has been considered as a next step in increasing business maturity and optimizing the decision making. 

Healthcare   

Prescriptive analytics find extensive applications in healthcare. One of the main problems in healthcare is the allocation of healthcare resources because without efficient management of healthcare resources it won’t be able to provide suitable healthcare resources at affordable cost. One such issue is the allocation of beds. This problem can be solved with the help of prescriptive analytics by using genetic algorithms to optimize bed allocation at affordable costs. 

Similarly the problem of appointment cancellations & no-show ups that causes huge financial and time loss can be resolved using the combination of predictive and prescriptive analytics. Predictive analytics can help in predicting the patients likely to miss the appointments using ML models (discussed in predictive analytics section) while prescriptive analytics using grid search methods along with cross-validation can help in optimizing the predicting results hence causing minimum loss of time & money.     

Supply chain

Prescriptive analytics techniques like game theory play an important role in supply chain management. Vehicles path optimization, inventory shortage predictions, market’s supply-demand trends, pricing of products are some of the key challenges that companies are solving using combined predictive and prescriptive analytics.

As an example, Waste Management, Inc. (WM) - a leading waste collection and disposal company saved  $44 million by combining predictive and prescriptive analytics techniques, hence optimizing their vehicle paths, time management and potential risks of speed-limits. These analytics play an important role in decision making at all levels including tactical, operational, marketing and strategy in supply chain management. 

Finance

Prescriptive analytics is helpful for the investors to select which investments to purchase and decide maximum and minimum investment cost range. Techniques and methodologies used in prescriptive analytics provide strong case studies for financial institutions in mitigating their risks of investments and maximizing their profits in the investments that they have made.

Stock market makes use of both predictive as well as prescriptive analytics in order to help investors as well as companies to invest logically and maximize their profits. Banks are using prescriptive analytics in their strategic decision making as well as developing customer relations by analyzing what targets they have to set in limited time and budget constraints to maximize their profits and limit the risk of losing money. Banks can also benefit from Anti-money laundering models based on predictive and prescriptive analytics.  

In short, there are many more sectors that are relying on the power of prescriptive analytics (alongwith predictive analytics) in their daily decision making processes. These include retail companies, power-distribution businesses, marketing companies, automobile companies and many more. 

Prescriptive analytics challenges

All the above mentioned applications, prescriptive analytics do come up with some challenges. 

The first and foremost one is the data itself. As described above data is the main driving force behind all the mathematics and algorithms used in prescriptive analytics and we also have heard of a famous phrase ‘garbage in, garbage out’. As per Gartner, poor quality data is the primary reason for 40% of business initiative failures in data analytics based on machine learning modeling. 

Therefore, it is really important to maintain the quality of the data collected as data is constantly changing and results of prescriptive analytics also changes based on that. 

The second challenge is the availability of skilled resources to perform these analytics for businesses. As a report from Mckinsey shows, there will be a shortage of up to 250,000 data analytics people in the US alone by 2024. As the field requires an extensive knowledge of mathematics, business and programming, therefore, it is very challenging to get good resources with the combined knowledge.

Another challenge is the privacy and security of the data involved. As prescriptive analytics require access to the sensitive data whether it is related to healthcare, finance or supply chain etc. Therefore, it becomes the utmost priority for any organization to be vigilant in their data protection policies, employees' access hierarchy and be aware of any potential cyber security attacks.       

Apart from these challenges, it is also a main challenge to define the problem statement clearly before initiating analytics roadmap. Failure to understand why and when to use predictive as well as prescriptive analytics is as important as any other step in this process. 

Prescriptive analytics advantages

Talking about the advantages of prescriptive analytics, we can summarize some of those in the following points. 

• Data Driven Decision Making - Given that the quality of data is maintained and right mathematical tools have been used, prescriptive analytics can provide useful action items for businesses as the examples in healthcare, supply chain and finance are discussed above.
• Competitive Advantage - Despite the familiarity with the prescriptive analytics concept, the majority of businesses are still struggling to benefit from it. Therefore, it provides an opportunity to gain a competitive edge by exploiting this cutting edge technology. 

‍

• Optimization - This might be the biggest advantage of prescriptive analytics that even predictive and descriptive analytics can not offer. Businesses can not only optimize their resources, funds and efforts by utilizing this methodology but also minimize their risks of losing investments by making informed and proactive decisions.  

Prescriptive analytics disadvantages

• Ethical Consideration - As described, prescriptive analytics is dependent on data and availability of data has its ethical angle as well that is needed to be dealt with carefully. It is important to consider the biases that could generate biased results and cause issues across different genders or nationalities. 

• High Cost - The processes involved in prescriptive analytics are quite expensive. They involve collection of data - involves a lot of resources, time & effort, training of data that requires high computational power, and huge time and effort investment from all the stakeholders. Still there are chances of not getting the desired outputs given the complexities involved at each step. 

• Complexity - The mathematical and machine learning concepts involved in predictive as well as prescriptive analytics are complex to understand and require focus and time along with the right skill set. Therefore, businesses struggle to implement all the steps successfully.    

Despite the mentioned disadvantages, it is quite evident that prescriptive analytics has been able to solve some of the very complex problems faced by businesses and continue to improve as the availability of more and more data is becoming possible. 

Prescriptive analytics vs predictive analytics

Prescriptive and predictive analytics are two stages of analytics where prescriptive analytics uses the output of predictive analytics and optimizes it. If we look into depth, predictive analytics uses machine learning models to output predictions about different scenarios that are important for businesses like market trends, inflation trends, climate predictions etc. Have a look on this figure illustrating detailed technologies that are part of predictive analytics: -

‍

Fig.6. Classification of predictive analytics methods (Image source: - Lepeniote)

Now, have a look at the tools and technologies involved in the prescriptive analytics: - 

‍

Fig.7. Classification of Prescriptive analytics methods (Image source: - Lepeniote)

Without going into the details of mentioned tools and techniques, we can observe that prescriptive analytics make use of outputs generated through machine learning (predictive analytics) and perform rigorous mathematical computations and analysis to optimize those outputs. Therefore, the decisions that businesses formulate based on prescriptive analytics often result in efficient results and minimize risk factors. 

In the same research, we can see some of the common steps and techniques that are the part of both predictive as well as prescriptive analytics. 

‍

Fig.8. Representation of common methods in predictive and prescriptive analytics (Image source: - Lepeniote)

We can have a look at some of the common techniques involved in both predictive & prescriptive analytics including some level of machine learning & probabilistic modeling. It is important to emphasize though, that usage of each method mentioned varies as per the problem statement.  

Prescriptive analytics can help organizations optimize their operations and hence can help in reducing in-efficiencies and saving costs. From a business point of view, it is important to understand the challenge and what problem statement could be defined to solve that specific challenge. The selection and use of the technique or technology is the secondary step in this regard.     

Leveraging prescriptive analytics at your organization

Depending on the sector, prescriptive analytics can have multiple use cases and can solve complex problems. The common ground though is the availability of quality data as well as large amounts of data also known as ‘Big Data’. 

Prescriptive analytics is the most sophisticated and powerful type of business analytics and requires domain expertise to be carried out properly. Prescriptive analytics can be leveraged in many ways depending on the business type but few common ways in which businesses can leverage prescriptive analytics are: - 

• Improved customer targeting.
• Price Optimization 
• Supply Chain management
• Process Optimization 
• Risk Management 
• Fraud Detection 
• Predictive Maintenance

Conclusion

To summarize, prescriptive analytics is very important for businesses in today’s age. It gives businesses the competitive advantage to outgrow their competitors and helps in minimizing their risks at the same time. It has its own challenges but in the end it is worth the effort!

‍