Business Intelligence Systems

Efforts undertaken to develop BI systems have resulted in many business solutions that allow for effective support of manager’s work. Practice shows that the most significant business effects are obtained while using the following analyzes offered by the BI systems,

• analysis that supports cross selling and up selling;
• customer segmentation and profiling;
  
• analysis of parameters importance;
• survival time analysis;
• analysis of customer loyalty and customer switching to competition;
• credit scoring;
• fraud detection;
• logistics optimizations;
• forecasting of strategic business processes development;
• web mining (analysis and assessment of the Internet services performance); and
• web-farming (analysis of the Internet content)

Analysis that Supports Cross Selling and up Selling

Marketing techniques of cross selling or up selling involve selling products to specific customers taking their previous purchases into consideration. Cross/up selling increases customer’s trust in the company they deal with, and reduces the risk of customer’s switching to competition. It leads to a remarkable increase in company’s incomes and customer loyalty level. Data mining model helps to select marketing campaign objectives optimally and, what is more, show the best
cross/up selling offers for customers in such a way that they correspond with customers’ present needs. There are many advanced methods that are used to find interdependencies between purchased products. One of them - Market Basket Analysis – provides knowledge on what kind of services and products should be sold together in sets or which set should be recommended to a particular customer. Using classification models to select customers who are the most susceptible to a particular offer is another practical application of the discussed solution. It allows to direct marketing activities correctly and – as a result – to reduce costs of the campaign while simultaneously increasing its effectiveness.


Customer Segmentation and Profiling

Customer segmentation and profiling is based on grouping customers in some homogeneous segments. BI systems enable both descriptive and predictive segmentation. Within descriptive segmentation the following segmentations are carried out:

• demographic segmentation (on the basis of the data including customer’s income, age,
sex, education, marital status, ethnic group, religion, etc.);
• behavioral segmentation (on the basis of the data including frequency of shopping,
amount and sort of purchased products, etc.); and
• motivational segmentation (on the basis of variables that describe reasons of customers’
purchases – this kind of data usually come from questionnaires and surveys carried out).

Subsequently, predictive segmentation is useful when it is necessary to distinguish ‘good’ customers from the ‘bad’ ones. At the very beginning, a variable that describes ‘good’ customers is determined (e.g. on the basis of total shopping they have done so far), and then, other variables that greatly influence the initial variable are determined. Such analyzes allow to create a specific approach to a particular segment of customers, and this approach is supported by dynamic updating of segmentation and analyzes of customers’ migration between segments. Segmentation and profiling of customers together with identification of potential cross/up selling offers and testing of different hypotheses enable to create a customized offer that enjoys huge potential of meeting future, new and loyal customers’ needs. Segmentation and profiling of customers provide some knowledge that is useful while designing new products and addressing marketing campaigns appropriately, as well. They allow for much more individualized customer service and optimization of marketing activities and sales, thus deriving profits from data concerning customers.

Analysis of Parameters Importance

Analysis of parameters importance allows for determination of the most important (from the perspective of company’s benefits) variables that describes products, processes and customers in the situation when there are different variables that describe analyzed objects. Knowledge obtained this way is used to identify directions to be taken while perfecting products and customer service, and planning marketing actions, etc. The Bivariate statistical analysis, stepwise regression algorithm or artificial neuronal networks are mainly used in this case.

Survival Time Analysis

Survival time analysis evaluates customer’s survival time length and a possibility that they leave during that time (leaving is understood as customer’s switching to other supplier of a particular product). The analysis describes a distribution of survival time for individuals of a given population, monitors strength of other parameters impact on the expected survival time, and additionally, it enables to compare distributions of survival time between different sub-populations. Taking advantage of this method, a company may be given an invaluable insight into customer behavior and find some ways to prolong customer’s survival time.


Analysis of Customer Loyalty and Customer Switching to Competition

Analysis of customer loyalty usually concerns four categories: time of co-operation, amount (volume) of co-operation, closeness of co-operation and quality of co-operation. It is strictly related to analyzes of customer’s switching to competition. That results in identifying customers who are inclined to leave a company and join competition. Discovery of factors that result in switching to competition enables a company to direct – appropriately - its activities that aim at retaining customers. Moreover, distinguishing groups of customers characterized by different risk levels of leaving allows for construction of effective loyalty programs and more attention paid to loyal customers.


Credit Scoring

Credit scoring models enable to determine financial risk that is related to particular customers. Such a process may be performed at the very moment a contract with a customer is concluded, and it is based on the data that come from application forms provided by a customer subject to analysis. Appropriate dealing with customers who are characterised by high risk of stopping payments makes it possible to reduce losses effectively. Credit scoring finds its application in, inter alias, banking (cash loans, assessment and tolerance of late payments) and in many other sectors related, for instance, to renting or leasing property and machinery. A good example of a credit scoring application may be also provided by contracts concluded to render telecommunications services connected with selling cellular phones. Credit scoring may be performed according to different models. Correct selection of the models depends on the analysis objective and specifics of the analyzed data:

•application scoring – used in case of new customers; information on them is available
only on the basis of the completed application forms;
• behavioral scoring – paying attention to additional information on customers’ track records;
it predicts customers’ future behavior; and
• profit scoring – expanding of the basic scoring model; it pays attention not only to probability of paying credits back by customers, but also helps to assess what sort of profit may be expected as a result of co-operation with a particular customer; it is a more sophisticated model because it considers several additional economic factors.


Fraud Detection

Fraud detection is a well-tried and incredibly efficient method due to which a company may save vast amounts of money, and keep good relations with customers. Fraud detection means identification of suspicious transfers, orders and other illegal activities that target a company in question. Fraud detection models may be divided into application assessment and behavioral assessment. The former is used to detect suspicious customers at the early stage of signing a contract with a company in question, and is based on data derived from submitted applications. However, the latter is formulated on the basis of all data gathered during ‘lifetime’ of customer’s activity including, inter alias, transactional data, use of services or performance track record. Fraud detection is frequently applied in order to prevent credit card frauds (e.g. Internet transaction frauds, telemarketing frauds or identity thefts), breaching of computer systems security, ‘money laundering’, telecommunications frauds, etc.


Logistics Optimizations

Logistics optimization problem involves offering the best possible plan of logistics activities (including transportation or distribution), simultaneously taking already known limitations and available potential into consideration. Wrongly prepared plan of logistics optimizations may result in huge delays of e.g. production or distribution that would consequently bring about a necessity for bearing higher costs - thus decreasing potential profits to be obtained. Employing advanced data mining techniques, it is possible to show the best available solution for actual and complex optimization problems. Quality of such solutions is usually much higher than the quality offered by traditional solutions of optimization methods.


Forecasting of Strategic Business Processes Development

Abilities to understand and forecast development of strategic business processes make up a foundation of the correct planning of any business activity. That is why, modeling of multidimensional forecasts based on historical, present and anticipated data is so important. Analyzes of time series make it possible to identify and analyze hidden trends and fluctuations (e.g. in marketing data or sales data). Taking seasonal nature and other marketing factors into consideration, it is possible to foresee potential behavior of market or customers, developments of customer expectations and customers’ purchases.


Web Mining

Analysis and assessment of the performance of Internet services (web mining) helps to obtain knowledge who uses services, when, why and how. Application of advance classification, grouping, matching and regression models with reference to data on service performance and its users (‘log’ files), business data (databases), and data on users’ experience with the service (questionnaires and interviews) allow to perform identification of customers and their preferences. Analysis and assessment of the performance of Internet services is limited to discovering and analyzing information that is stored in the service (web content mining), discovering and analyzing patterns of using the service by its users (web usage mining), and analyzing service structure (web structure mining). This way, valuable, dialectic knowledge is acquired – the knowledge on company’s offer attractiveness and its formulation in such a way that it corresponds to individual needs of particular customers. As a result, it is possible to customize service, automate navigation, shape pricing and promotional strategy and develop ‘intelligent’ e-business.


Web-Farming

Web-farming involves systematic analyses of the Internet content in order to provide a company with themes and issues of fundamental nature for company performance. Internet is more and more frequently treated as a powerful resource of important economic information on potential customers, suppliers and competitors, information on the latest market bargains, technological trends and development of the world economy. Therefore, each company that wishes to remain its competitiveness must explore the web that is understood as a valuable source of knowledge. Web-farming offers a possibility of constant analyzing of the Internet; finding important business information there; acquiring such information; saving it in data warehouse of a company; and delivering processed information to adequate persons or departments in a company. Major benefits obtained while carrying out web-farming include permanent monitoring of strategic business information sources, extracting of essential facts and their fluent matching with the internal system of company data storing. All these operations may be performed by means of advanced data mining tools.