In data mining literature, clustering plays an important role in bringing insights from a dataset that is actionable and provides important business directions. In simple language, clustering aims at bringing similar observations such as similar customers, similar patients, similar users, and so on. Clustering techniques are not limited to the domain of retail but can be extended to any domain. Segmentation is no more limited to the retail or e-commerce domain; it is also applicable to all domains and industries.
In this chapter, we are going to learn the following things:
- What is segmentation?
- How can clustering be applied to perform segmentation?
- What are the methods used for clustering?
- A comparative view of the various methods
- A practical project on segmentation
In this chapter, you will know the basics of segmentation using various clustering methods. There are different methods used to perform clustering. The following examples clarify where and how clustering can be used to create segments that can be used to drive business value:
- In the retail/e-commerce industry, it is not possible to understand the behavior of millions of customers to design the campaign, marketing strategy, and sales promotion engagements. These customers actually belong to a finite group of customers displaying similar behavior within a group and dissimilar behavior between groups.
- In the telecommunications industry, the positioning of towers to optimize user experience is driven by segmentation exercise. Also the design of telecom plans, such as custom offers (especially for students, older people, professionals, and so on), based on the usage of services are also made using clustering methods.
- In the healthcare industry, to decide the size of hospital beds for different departments, such as emergency care, acute care, and so on, clustering exercise helps. Also, reporting of various diseases across different states can be used to create a cluster of similar regions or places where a set of procedures can be adopted.
- In governance, law enforcement agencies strengthen their presence in those areas where crime evidence/instance is more. So clustering techniques are used to create groups of cities such as high, medium, and low-crime areas. Based on the data, they can align their forces accordingly.
- In insurance, clustering is used to decide differential premiums by grouping different cities based on parameters such as demographic information, geolocation, and risk zone.
In standard data mining practice, customer segmentation is a way of dividing all customers into various subgroups relevant to the business and the business problem. That subgroup creation can be done either by using a subjective approach or by keeping the business objective in mind. In different industries, customer segmentation has different applications; for example, in retail, it is important to know the purchase behavior of customers, and different subgroups displaying unique purchase behavior is relevant to the business.
In the retail and e-commerce industry, offers, campaigns, loyalty programs, and discount strategies work based on the purchase behavior of the subgroup of customers. In other industries, sales, marketing strategy, and business plans run keeping in mind the customer's behavior, and the behavior drives the sales. That unique customer behavior one can be understood by performing segmentation on the data.
Having discussed various benefits of performing customer segmentation, it is important to know how to perform customer segmentation. There are two broad methods used to perform customer segmentation:
- Customer segmentation using clustering-based methods
- Customer segmentation using the recency, frequency, and monetary (RFM) model
In this chapter, we will discuss the clustering-based approach to perform customer segmentation; the second method is not in the scope of this chapter.