Data mining wrhousing-lec
Download as PPT, PDF0 likes326 views
The document discusses data mining and data warehousing. It describes data mining as a technique that enables companies to discover patterns and relationships in data with a high degree of accuracy. Typical tasks for data mining include predicting customer responses, identifying opportunities for cross-selling products, and detecting fraud. The document also discusses why companies build marketing data warehouses - to more efficiently and profitably serve customers by integrating customer data from various sources and analyzing purchase histories. Key considerations for ensuring success include having the right support team, quantifying benefits, and prioritizing deliverables in a phased approach.
Data mining wrhousing-lec
- 1. DATA MINING
- 2. • In the past, there was never a high level of precision when it came to marketing your products • Trying to compare traditional marketing strategies with data mining is like trying to compare a precision guided missile to blindly throwing darts at a board. There is no comparison. • The advancement of artificial intelligence, neural networks, and computer algorithms has led to a world in which the technology is capable of learning. Because of this, it can look for relationships that my not be obvious to the user
- 3. • Data mining is a technique that enables the company to discover connections, patterns, and relationships in a set of data with a high degree of accuracy and precision. • It is not a business solution • However, the patterns that are discovered by data mining tools are only relevant if they are directly related to solving a specific problem. Otherwise, any patterns that are discovered is useless. Also, you have to look at the coverage of your data mining tool.
- 4. • Typical tasks addressed by data mining include: • Rate customers by their propensity to respond to an offer • Identify cross-sell opportunities • Detect fraud and abuse in insurance and finance • Estimate probability of an illness reoccurrence or hospital re-admission • Isolate root causes of an outcome in clinical studies
- 5. • Determine optimal sets of parameters for a production line operation • Predict peak load of a network • Without proper analytical tools, discovering useful knowledge hidden in huge volumes of raw data represents a formidable task. The exponential growth in data, diverse nature of data and analysis objectives, the complexity of analyzing mixed structured data and text are among the factors that turn knowledge discovery into a real challenge.
- 6. • Data Mining provides tools for automated learning from historical data and developing models to predict outcomes of future situations. The best data mining software tools provide a variety of machine learning algorithms for modeling, such as Regression, Neural Network, Decision Tree, Bayesian Network, CHAID, Support Vector Machine, and Random Forest, to name a few. • Yet, data mining requires far more than just machine learning. Data mining additionally involves data preprocessing, and results delivery. • Data pre-processing includes loading and integrating data from various data sources, normalizing and cleansing data, and carrying out exploratory data analysis.
- 8. • We build a marketing data warehouse for the main purpose of more efficiently and profitably servicing our customers and prospects today and in the future • Example-- a publisher of several titles that also offers special online content such as webcasts or paid reports.. • Without such a database a publisher would not know how to cross sell the various titles or online products or services most effectively. The fulfillment files for each would most likely be separate and distinct and not allow for an efficient usage of information for promotional decision purposes. And, even if all were fulfilled from the same source, the data would likely not be integrated at a customer level nor easily accessible for marketing.
- 9. • Although a fulfillment file contains a wealth of information, its structure is rigid (built for fulfillment rather than marketing) and it lacks complete information about the customer and typically deals with only one product. Thus, alone it will not meet all of the needs to efficiently target communications to your customers or prospects across the company. What are the key points that must be taken into consideration to ensure success in the build of a marketing data warehouse. Why Build a Marketing Data Warehouse Top Reasons Why We Fail What Functional Areas it Must Support Key Profit Drivers Who Will be Using the Data Warehouse and what it must deliver What should be stored on the Data Warehouse How often should the Data Warehouse be updated Should the Database be Built In-house or Outsourced Quantifying Profit
- 10. • Why Build a Marketing Database • • • • • There are three key functions of a marketing database: To most efficiently maintain your data in an organized fashion To better support corporate functions To better support marketing functions All of which will allow you to better service your customer and hence maximize revenue • The pinnacle is of course to be able to calculate and forecast customer lifetime value • The foundation of the database build is the customer contact information. From this we accumulate historical purchase and promotional data called our marketing data.
- 11. REASONS FOR FALIURE • • The Major Reasons For Failiure Underestimating the time and resource commitment to build or maintain the database • Not having the right support team in place once the database is delivered even if outsourced • Not having a plan in place regarding how you will use the database once delivered and how you will quantify the benefits. • Inappropriate in scope -- too broad or too narrow • Not properly prioritizing deliverables – phased in approach • • • Failure to shift the paradigm at the organization to a informationbased decision approach
- 12. • Thinking that if you build the database profits will come • Failure to realize that your number one priority in the build is getting the data right. • Failure to fully assess costs of “add ons” relative to total database costs versus their benefits. It is not any one reason that causes failure but a combination of the above.
- 13. KEY PROFIT DRIVERS --The major profit drivers can be --Demographic Profiling, --Mkt Research support --CRM Strategies --Lifetime Value analysis --Customer Acquisition Models --Retention Models --Strategic Corporate Reports Some functions can be performed just by virtue of having all of the data readily available and in one place, others require the application of quantitative sophistication to complete.
- 14. WHO WILL BE USING DATABASE • Keep in mind that there are many individuals throughout the organization who will utilize a marketing database. However, typically many of these needs are related in one form or another. As such, you will want to involve all in the decision process of what the database should ultimately accomplish. Many individuals will need similar access to the database and what is has to offer. For example, all functional areas/individuals will want some form of “dashboards” or other reports. To ensure success and keep the cost of the database in check, you must understand the relationships between the individuals so that priorities can be properly set regarding deliverables and features. It is best to build in phases ---start out small and build on each success. Remember this is a marketing database. The primary focus must first be to meet marketing’s needs. Other divisions can then follow