subject: Mensur Boydas Lecture On Data Mining Techniques For marketing [print this page] Mensur Boydas Lecture On Data Mining Techniques For marketing
1- What is Data Mining?
Data mining is the explorations and analysis of large amount of data in order to discover meaningful prototypes and rules. The aim of data mining is to help organizations to improve their marketing, sales, and customer maintaining procedures in analyzing their customers. In reality first data mining algorithms were invented for marketable applications. Data mining in marketing consists of two ways; directed and undirected. Directed data mining tries to categorize some target fields such as income or consumer response. Undirected data mining is for pattern recognition or similarities among groups of data without the utilization of a target field or classes. The main purpose of data mining is to build models. A model is an algorithm that connects a collection of inputs to a particular outcome. A model can result in imminent by providing an explanation of how outcomes of particular interest, such as placing an order or failing to pay a bill. Models are also used to depict scores. There are six tasks can be performed with data mining that are;
Classification
Estimation
Prediction
Association Rules
Clustering
Description and profiling.
Classification consists of investigating the characteristics of a newly vacant item and assigning it to a predefined set of classes. Simply discrete results: yes or no. Estimation deals with continuous results that given some data, estimation comes up with a value for some unidentified continuous variable such as income or credit card balance. Prediction is no different from classification and estimation except that the records are classified according to some predicted future behavior. The aim of association rules is to determine which items go together. As an example determining what items go together in a shopping cart at the supermarket. Clustering is segmenting a various population into a number of more identical clusters. Clustering does not rely on predefined classes. Profiling is describing what is happening in a complex database in a way that increases our understanding of people, products, or processes that produced the information. Decision trees, association rules, and clustering can be utilized for profiling customers for a particular outcome in marketing.
2- Data Mining Applications in Marketing
Data mining application utilized in business context to be of value. The topics covered in this section are prospecting, data mining to choose the right place to advertise, and data mining for customer relationship management.
Prospecting
A prospect is a person who may be an expected to be a customer in the future. Data mining is used accomplish the organizational goal of finding people who will be important customers in the future. Comparatively, few of earth's more than six billion people are in fact prospects for a company. The exclusion based on geography, age, ability to pay, need for product/service, etc. A newspaper wants to target people who read the appropriate language and will be of interest to its advertisers. Data mining can be utilized in many ways in prospecting and some of the important ones are as the following;
Identifying good prospects
Choosing a communication channel for reaching prospects
Picking appropriate messages for different groups of prospects.
The most widely used role here is the identifying good prospects.
Data mining to Choose the Right Place to Advertise
One of the ways of finding prospects is to look for people who are similar to current customers. Majority of the companies that sell products and services need to advertise and promote their products and services in order to maximize their share in the mean market. There are mainly two approaches to advertising and promotion; mass and direct marketing. Mass marketing uses TV, radio, and newspapers, broadcast messages. Direct marketing analyzes the customers' characteristics and needs and choose certain customers as the target for the promotion and advertising. As an example, when something is going to be advertised in one of the mass media such as TV, publication should look for channels whose watchers match their prospect profile. Mass marketing is an effective way of promotion when there is a high demand for the promoted goods or services. In today's perfectly competitive business environment mass marketing is becoming less and less effective that expecting to sell a product after advertising it with mass marketing techniques. Direct marketing is more promising in gathering responses from the prospects of particular goods or services. Today, there is a huge amount of information is kept in databases about customers, so data mining could be very promising in direct marketing. From the databases patterns can be discovered through data mining to predict the creditable customers for promotion. The aim of marketing campaign is to change behavior. Thus, reaching a prospect who is going to purchase anyway is less effective than reaching the ones that are not likely to purchase. Whether a customer fits the profile is up to the similarity measure also called distance between the customer and the profile.
Data Mining for Customer relationship management
Customer Relationship Management (CRM) mainly deals with well-known customers. For data mining established customers is the main source. Does the customer pay bills? What kinds of products does the customer purchase? How many times have we contacted the customer? These kinds of data could be utilized to evaluate the value of customers and predict the risks that the customer will cut the relationship or stop paying bills. The customers that you have today could disappear tomorrow. Interacting with your customers is also not as simple as it has been in the past. Customers and prospective customers want to cooperate on their terms, implication that you require to look at various criterion when evaluating how to proceed. At the end of customer value determination data mining comes into the picture to estimate prospective customer value that is approximation of the revenue a customer will bring and the customer's remaining lifetime.
3- The Usage of Association Rules in Marketing
Association rules in marketing mainly concentrated on purchases of different customer types and what they purchased. Every customer purchases different products in each trip to the supermarket. This analysis uses this information predict which products tend to be purchased together and which are appropriate to promotion. The information gathered from this could be utilized to designate when to issue coupons, which products to put on special and so on. Association rules represent patterns in the data without a special target which is an undirected data mining technique.
Here are the real rules generated from the real data; "
Wal-mart customers who purchase Barbie dolls have a 60 percent likelihood of also buying one of three types of candy bars.
Customers who purchase maintenance agreements are very likely to purchase large appliances.
When a new hardware store opens, one of the most commonly sold items is toilet bowl cleaners" (Berry, Linoff, p.296).
Given a set of records each of which contains some number of items from a given set of purchase; produce dependency rules which will predict incidence of an item based on occurrences of other items. For example
Table1: Supermarket Transactions
The table shows the five transactions made by customers. From this data using association rules we can come up with purchasing rules such that;
{Milk} - {Coke}
{Diaper, Milk} ---{Beer}
Thus, whoever is buying milk could also buy coke and the probability of this recognition could also be calculated and whoever is buying diaper and milk could also buy beer.
After the rule discoveries, there are applications to take care of such as; Marketing and Sales Promotion, Supermarket shelf management and Inventory Management.
Marketing and Sales Promotion:
Assume the discovered rule is {Milk} - {Coke}
Coke as consequent could be used to establish what should be done to boost its sales.
Milk as the antecedent could be used to see which products would be affected if the store discontinues selling milk.
Milk in antecedent and Coke in consequent could be used to see what products should be sold with Milk to promote sale of Coke.
Supermarket shelf management:
Objective: Categorize items that are purchased together by satisfactorily many consumers.
Approach: Practice the point-of-sale information composed with barcode scanners to find reliance among products.
Inventory Management:
Objective: A customer household renovation company wants to predict the nature of repairs on its goods and services and keep the service vehicles prepared with right parts to decrease on number of visits to customers.
Approach: The approach to this is to perform the data on tools and parts necessary in earlier maintenance at different customer locations and determine the co-occurrence models.
Dissociation Rule:
Dissociation rule is like association rule exept that it can have the connector "and not" in the condition to "and". Dissociation rules look like:
If A and not B, then C. Dissociation rules could be created by simple variation of the market basket analysis algorithm. Variation is to bring in a new set of items that are the inverse of each of the original items.
Association Rules offer ways to evaluate item level detail, where the associations among items are determined by the consumer transactions in supermarkets.
4- Practice of Data Mining in Marketing
The methodology of data mining in marketing has eleven steps. They are "
Translate the business problem into a data mining problem.
Select appropriate data.
Get to know the data.
Create a model set.
Fix problems with the data.
Transform data to bring information to the surface.
Build models.
Asses models.
Deploy models.
10. Asses results.
11. Begin again" (Berry, Linoff, p.54, 55).
The business problems are translated into data mining problems by the utility of the data mining tasks. Then, the data is visualized to see some informative patterns. The model creation comes next that is the algorithm to achieve a solution that is numerical value. Data mining engages the method of translating business needs into a problem which has to be evaluated, recovering the database with large amount of data for analysis, and applying statistical or machine leaning system with the final aim of accomplishing important results for taking a strategic judgment to solve the business problem.
5- Conclusion
Data Mining is very practical in the business world where gaining competitive advantage is critical for good business presentation and survival. Evaluating and modeling customer transaction information activities can provide information that leads to better decision making such as product promotion to the right segment of customers. Data mining is generally used in intention advertising, fraud detection, credit scoring, consumer segmentation, and merchandise prologue.
