An Inquiry into Business Intelligence Application in Telecom BOSS

1 Introduction
With the construction and development of Business and Operation Support System (BOSS) database, comprehensive applications of mass data accumulated during telecom operations become a key problem. Such applications are far beyond traditional statistical reports, and can serve as a powerful analyzing platform to provide statistics, operation analysis and data tapping and data warning through comprehensive use of data warehouse, Online Analytical Processing (OLAP) and other technical means. Scientific marketing management, construction planning and business decision-making are realized through wide application of Business Intelligence (BI) in BOSS[1,2].

    The objective of BI construction in BOSS is to establish a unified business data information platform. By adopting advanced data warehouse technologies and analyzing and tapping tools, BI is able to extract valuable information from enterprise data, provides scientific and effective support for enterprises’ customer services, marketing and network construction, and enhances enterprises’ operation level and competitiveness.

    In system construction, BI has close relations with other service systems of BOSS. It uses production data in the BOSS as main data sources that are supplemented with other enterprise information systems and external data sources. Therefore, it can establish a unified operation analysis view and form a specialized data warehouse for operation analysis.

    At the same time, BI analysis results can be fed back to the production system directly or indirectly according to actual situations to realize process improvement and management optimization and to provide customers with better services.

    By and large, with the accumulation of a massive amount of valuable production data during BOSS operation, the demand for integrated data applications will become more and more urgent, providing a strong impetus for the application of BI.

    China Mobile, China Telecom and other Chinese operators have set about its standard formulation and large-scale popularization on the basis of their pilot projects, although there still exist many problems.

2 Technical Framework and Strategy of BI Realization
For BI technologies, since there are rich development and application experiences in other industries, BI construction in telecom BOSS can fully make use of the external experiences and adopt mature technologies to make high-starting-point planning and construction possible to create economic benefits for enterprise development as soon as possible.

    The technical framework of a BI system can be divided into the data integration layer, data storage layer and data application layer, as shown in Figure 1.

    The data integration layer extracts, clears and converts data out of the Customer Relationship Management (CRM) system, accounting and dispatching system and other BOSSs, as well as Network Management (NM) system, Management Information System (MIS) and other external data sources, and then loads them into the data warehouse. The data integration layer is generally divided into three sub-layers, namely, Source Data, Data Extraction, and Clearing/Converting/Loading.

    The data storage layer realizes centralized storage and management of data and meta-data in the data warehouse, and can establish sector and subject-oriented data fair on demands, and can store some pre-aggregating data in the online analytical processing (OLAP) server.

    The data application layer conducts analysis and processing of data in the data warehouse with the help of diversified front-end analyzing and displaying tools, forming scientific, accurate and timely service information and knowledge for market operation and decision-making. The data application layer can internally be divided into two parts: OLAP Analysis and Data Tapping.

    As major Database Management System (DBMS) manufacturers all provide their own data warehouse solutions, and have had a batch of independent and mature software products in data application and data conversion, filtering and loading, present BI technical solutions can integrate such software.

    In formulating and implementing their data warehouse solutions, enterprises shouldn’t blindly choose product suppliers, but should give a comprehensive consideration to many factors including manufacturer strengths, product functions, past experiences, brand services, development prospects, cost-effectiveness, etc.

    And at the same time, attention should be given to the following respects:
    (1) Data Extraction Strategy
    Data extraction must fully meet the demand of BI system’s analysis and decision support, and ensure not to affect the performance of the service system. So, in data extracting, full consideration must be given to these factors to work out an appropriate strategy, including the way, time and cycle of extraction. The ways of extraction include increment extraction, full extraction, and so on. The right time of extraction should avoid the peak period of the system as far as possible, e.g., choosing the idle period of the service system at night. For different data sources, comprehensive consideration should be given to service demands and system costs, and then as a result, a reasonable cycle extraction works out.

    (2) Data Conversion Technique and Strategy
    When file mode processing is needed in adopting data load, full consideration must be given to the storage capacity of the intermediate disc and the coordination of the entire flow of Extract, Transform and Load (ETL), as well as the programming of a large amount of non-SQL (Structured Query Language) statement. Loading performance must be considered, when data transition is adopted during data loading. When adopting processing after data is loaded into the data warehouse, the mass data processing capability of the data warehouse engine must be considered.

    (3) Data Loading Technique and Strategy
    Data loading strategy should consider two aspects: the loading cycle and data addition. According to the actual condition of telecom service data, the loading cycle should comprehensively consider service analysis demand and system loading cost, and different loading cycles should be applied for different service systems. However, the integrity of service data in a certain period of time must be ensured. Data addition strategy should be defined according to data extraction strategy and service regulation. Generally it can be divided into three types: direct addition, full coverage and updated addition.

    (4) Data Storage Method and Strategy
    The data storage layer includes the enterprise-level data warehouse and reproduction/propagation modules.

    The data warehouse module is used to define the logical and physical storage of information. The reproduction/propagation module creates a slave department-level data fair and OLAP database from the data stored in the enterprise-level data warehouse. The enterprise-level data warehouse organizes and stores data in accordance with an enterprise’s overall information model and by the smallest service unit as far as possible. This can ensure the flexibility of data access and minimum data redundancy as well.

    During the implementation of a data warehouse, for some subject-service analysis, data may be further organized in a way of the data fair or OLAP database based on subjects. The corresponding slave department-level data fair and OLAP database will be created by demand analysis on the basis of enterprise-level data warehouse.

    (5) Data Application Technique and Strategy
    The data application layer should provide a unified portal inlet to realize seamless connection of predefined statement, instantaneous search and multi-dimensional dynamic analysis, and provide integrated authentication, information release and management environment. This aims at enabling system users to realize access to and analysis of system data without paying attention to the specific technical implementation approach.

    Meantime, the portal site can also accomplish such functions as convenient and fast customizing, coloring and guiding according to different analyzing and decision-making personnel’s demands, so as to meet the need for personalized information services.

3 Typical Applications of BI in Telecom BOSS
Presently, the construction of Chinese operators’ resource management systems has just started, and data shortage universally exists in the management of network resources while accounting data and customer data of the business operation system are more complete thanks to the relative long time of accumulation. Therefore, currently most applications of BI in BOSS are customer analysis and income analysis.

    With advances in the construction of the resource management system and fault management system, network resource analysis, fault analysis and network optimization will become the target of next BI construction[3].

    Compared with the technical platform, a major problem in service applications is the shortage of service experience for various analysis items. Based on the author’s work experience, this paper discusses service contents of two typical applications: customer analysis and network resource analysis:

    (1) Application in Customer Analysis[4]
    The deepening of customer management is the base of CRM. The focus of CRM is how to retain existing customers and attract and develop potential customers. For this purpose, we must have a full understanding of customers’ demands, define customers’ value to the enterprise, and determine specific customers’ opportunities and risks to the enterprise. All this can be hardly done without unified collection, analysis and tapping of customer data that are available.

    Therefore, the analysis of customer data is an important content of present BI. The following points are considered in terms of service contents:

• Customer type analysis: Critical natural attribute characteristics of customer subgroups are tapped through the analysis of the customer background attribute, customer status attribute, customer account attribute and customer behavior attribute. Customer type analysis is conducted on this basis, and customer loyalty levels, customer service grades, customer credibility and customer’s accumulated points are set up. And based on the customer type analysis, telecom enterprises are able to choose different marketing channels and provide differentiated products and services for different customer groups.
• Potential customers tapping analysis: Potential customer groups are confirmed with the analysis of customer types, business income changes, customer service types, consulting and inquiry behavior, and the prediction of customer change trends under certain conditions. Based on this, an enterprise can make relevant marketing activities to expand customer groups and improve its market share.
• Loss analysis: Fundamental reasons of the customer loss and the value loss caused to an enterprise are found through the analysis of the loss and abnormal moving of different types of customers (like the ranking analysis, comparative analysis and multi-dimensional analysis). Measures and methods of minimizing the loss caused to the enterprise may be found, which enables the enterprise to conduct customer holding specifically and to reduce such losses finally.
• Customer service income analysis: The total income and its change are analyzed from different angles, and then the change trend under certain conditions is predicted to provide a scientific and effective basis for working out a rational marketing strategy, which is necessary for raising the income of a telecom enterprise.
• Customers’ communication behavior analysis: Customers’ traffic and its increment and the use of new services are analyzed to obtain the customer composition of different services, and traffic change and growth. It aims at providing a basis for enterprise’s service prediction and corresponding network construction as well as investment optimization.
• Customer service and satisfaction level analysis: By combining statistical analysis with sample investigation from the angle of an enterprise itself, finding whether improvement has been made in customer service quality and the speed of improving are found out. In a way of combining sample investigation and statistical investigation from the customers’ angle, the enterprise can get to know customers’ satisfaction and demands on communication networks, service developing, tariff and service quality.
  

    (2) Applications in Network Resources Analysis
    Based on the construction of network resource management system, a statistical analysis can be made with a large amount of accumulated network resources data. By analyzing the status and use of telecom network resources from different angles including network resources occupation, use of leased resources, resource warning and network bottlenecks to predict the change trend under a certain condition.

    It can provide a scientific and effective basis for network construction and network optimization, and will be an important direction of BI’s application in BOSS. Its main data source is the resource data provided by the resource management system and NMS. The following points are considered emphatically in terms of service contents:

• Resources use analysis: By analyzing different telecom network resources and their bearer service types, a correlative model between the resource and the service is established. The telecom enterprise is able to predict and plan the use and development of network resources through the service development to avoid the blindness of network construction investment and improve the economic benefit.
• Resource warning: By setting a threshold of a network resource, warning management of the use efficiency of a resource within a region is realized. Examples include port warning, circuit warning and line warning.  It is possible to automatically trigger warning and start relevant processes. It then provides a scientific and timely management means for telecom enterprise’ network capacity expansion and allocation of resources.
• Network bottleneck analysis: Through correlation analysis of various telecom resources during service operation, network bottlenecks and its fundamental causes are found and defined. They are a scientific basis for telecom enterprise’s network optimization and rational allocation of resources.
• Third-party resources analysis: Through data tapping and statistical analysis of such factors as the existence, use, service bearer and cost of the third-party resources used in enterprise’s network resources, a scientific decision-making basis for the use of
  enterprise’s third party resource is provided. Besides, it is also a basis for the replacement of the third party resources with self-constructed networks.
  

4 Summary
BI construction in BOSS is a comprehensive application of a massive amount of accumulated operation support data. By establishing a unified enterprise data information and analysis platform, a scientific and effective support for enterprise’s customer service, marketing and network construction can be achieved. Therefore, it can be forecasted that BI will have broad development prospects.

References
[1] Wu Bin. Technologies and Applications of Business Intelligence[J]. China Computer Users, 2004(2).
[2] Sun Ding. Development and Applications of Business Intelligence[J]. China Computerworld, 2003(1).
[3] Ye Yun. Development Trends of Telecom Operation Support in Market Competition[J]. World Telecommunications, 2003(2).
[4] Service Specifications of Operation Analysis System of China Mobile[S].

Manuscript received: 2004-09-02