Cluster, classification and subdivision techniques for specific marketing

The terminology segmentation has become ubiquitous and has many different meanings in different contexts, which are often confusing. This is not uncommon for businesses, and there are often several subdivision functions that appear in different departments at the same time. Generally, most experts agree that subdivision as a term encompassing Vientiane refers to the general division of dividing a whole into subsets of similar units. In addition, however, the subject may also be a controversial subject.

Suppose that in an organization, the following subdivisions work at the same time for a given period:

The

Research and Development (R&D) Division develops a customer segmentation to better understand consumer preferences and buying behavior, thereby driving customized product improvements. R&d may also develop a product breakdown to understand the similarities and types of products that are usually purchased together. The Finance department identifies customers and prospect segments to help with revenue forecasts. The data in this case may be profitability, acquisition costs, lifecycle values, demographics, retention, and advertising costs. Market research segmentation forms the basis of service and quality perception, thus promoting brand strategy and advertising investment. Traditionally, market researchers performed segmentation through survey instruments and customer feedback data. The marketing department also has another subdivision to understand who responds to various marketing channel activities to refine target market choices and improve market feedback. Market analysts usually use the original customer purchase behavior and demographic data as the basis for subdivision.

This kind of scenario is quite common, the enterprise lacks a common subdivision strategy, and disparate (usually contradictory) subdivisions are developed across departments and used in very different ways. This practice is prevalent in many industries that use subdivision. To provide a limited snapshot of how each industry handles subdivisions, consider the following applications: Insurance companies use subdivision to determine risk allocation and set pricing criteria and premium levels. The power industry uses a bottom-up approach to load forecasting and execution phase forecasting to aggregate overall demand. The automotive industry uses a breakdown to understand the design and performance preferences of the target market. The bank subdivides the credit card market foreground to realize the direct mail service. Biologists classify animals according to their physical structure and growth area by dividing them into different things. The Pharmaceutical enterprise deployment segment to maximize product innovation lifecycle. The field of image processing (including facial recognition) is one of the most complex fields, which relies on the complex subdivision of parameters, region growth and edge detection algorithms. Regardless of the industry (and possibly all enterprises), it will benefit from trying a more unified and matching enterprise segmentation strategy.

Marketing Segmentation

The differences listed above describe in detail the various methods and objectives for the breakdown of the project. Market researchers and marketing analyst professionals typically handle the process with disparate goals, input data, and methods. Let's explore the standard method of market segmentation further.

The first step in any subdivision is to understand the goals and motivations of the study. Who requires subdivision? What will the subdivision be used for? Why do I need to subdivide? What information is required but not yet available to consumers? Who will use output? What data will be used to support segmentation? How will the subdivisions be manipulated and deployed? How to measure the success of a project? The answers to all these questions help determine the most appropriate technology, data, and algorithms to solve the problem. In the next section, we'll look at a specific use case, outline two possible approaches, and discuss similarities and differences between customer segmentation and predictive modeling.

Data entry and standard subdivision methods

Data is the key input to any subdivision. In general, as long as the data source can be accurately associated with a personal or family ID, the more data the better. The list of available data is almost limitless, but there are several key data categories:

survey data can be collected by customers or consumers in the general population around product and price concessions, channel sales, customer experience satisfaction, and improvement recommendations. Transaction data is typically stored in a relational database that includes purchases, returns, discounts, payment methods, purchase dates, and times in a retail environment. In a financial environment, this information will become deposits, withdrawals, reconciliations, savings and mortgage-backed products, and details of each product. In the energy environment, this information includes use, storage and transportation loss, subsidies, reserves, devices and intelligent instruments. Behavioral data includes Web browsing behavior, store navigation, eye tracking, speech recognition, search, mobile usage and device information, positioning, frequency, and import and export volumes interacting with brands. Social media interactions, such as "like", forwarding, and attention are also part of this rich data type. Demographic data can be collected directly from customers, or from a population data provider, and suppliers can provide 300 to 900 information about individuals, families, and postal codes. These third parties have additional data sets, many of which come from the US Census response data. Other data categories include call centers, chats, information seeking, price comparisons, reviews, participation in peripheral programs and communities, and product information.

After the initial business goal refinement and the data discovery are completed, we can start to consider the feasible subdivision method. You can choose from a variety of traditional methods, each with its advantages and limitations. For example, many cluster options produce clusters of the same size; although this is feasible from a deployment perspective, enforcing the same size cluster may weaken the advantage of similarity metrics in the cluster.

There are 3 basic options for determining the best segmentation method, and Figure 1 shows 3 common methods: Non-quantitative, interdependent, and related.

Figure 1. 3 Basic Subdivision Options

The first option is a qualitative (or not quantitative) approach that involves the collection of fragmented information by comparing dimension information that is obtained by interviewing and focusing on the business stakeholders. These dimension information reflects empirical data about consumer behavior and is used to specify a subjective subdivision for the target processing policy. Although useful in one way or another, these are often less robust than the other two data-driven subdivisions (interdependencies and dependencies).