Professional Documents
Culture Documents
Fases de Implantacion Proyecto BI
Fases de Implantacion Proyecto BI
net/publication/40646563
CITATIONS READS
26 2,791
6 authors, including:
Some of the authors of this publication are also working on these related projects:
Intelligent system for trading on wholesale electricity market (SMARTRADE) View project
All content following this page was uploaded by Bâra Adela on 03 February 2014.
Adela BARA, Iuliana BOTHA, Vlad DIACONIŢA, Ion LUNGU, Anda VELICANU,
Manole VELICANU
Academy of Economic Studies, Bucharest, Romania
Faculty of Economics Cybernetics, Statistics and Informatics
bara.adela@ie.ase.ro, iuliana.botha@ie.ase.ro, diaconita.vlad@csie.ase.ro,
ion.lungu@ie.ase.ro, anda.velicanu@ie.ase.ro, mvelicanu@yahoo.com
Often, Business Intelligence Systems (BIS) require historical data or data collected from var-
ious sources. The solution is found in data warehouses, which are the main technology used
to extract, transform, load and store data in the organizational Business Intelligence projects.
The development cycle of a data warehouse involves lots of resources, time, high costs and
above all, it is built only for some specific tasks. In this paper, we’ll present some of the as-
pects of the BI systems’ development such as: architecture, lifecycle, modeling techniques and
finally, some evaluation criteria for the system’s performance.
Keywords: BIS (Business Intelligence Systems), Data Warehouses, OLAP (On-Line Analytical
Processing), Object-Oriented Modeling
1 Introduction
The main objective of our research
project that we had in one of the multination-
searches on Object Oriented modeling, we
consider it as a good option. In the last years,
there have been some proposals to represent
al companies from Romania is to develop a MD properties at the conceptual level. As we
decision support system for public institution, mentioned in our previous researches, based
so we tried to meet the requests from the ex- on the conference paper [1], we defined a set
ecutives and managers of one national com- of object-oriented extensions that can be used
pany. We based our work on previous expe- for modeling the components and require-
riences, researches, articles and studies that ments of a data warehouse. Also, we had to
we’d been developed. Thus, we followed the consider the system’s development lifecycle
next classical steps: analyze, design, develop that has to be flexible and easy to fulfill.
and applying for the project lifecycle the
framework described in the book [2]. For the 2 The concept of Business Intelligence Sys-
implementation phase we used different BI tem
techniques, like data warehousing, OLAP, With rapid advances in technology, enter-
data mining, portal and we finally succeeded prises today frequently search for new ways
to implement the BI system’s prototype and to establish value positions. Well built Busi-
to validate it with the managers and execu- ness Intelligence Systems (BIS) can provide
tives in one national company. The system the ability to analyze business information in
gathers data, using the ERP system, to extract order to support and improve management
data from different functional areas or mod- decision making across a broad range of
ules such as: financials, inventory, purchase, business activities. They leverage the large
order management or production. For the ex- data infrastructure investment, for example,
ecutives, the system is able to provide analyt- in ERP systems made by firms, and have the
ical reports and dashboards. As the storage potential to realize the substantial value
solution we designed and build a data ware- locked up in a firm’s data resources [7].
house. The major problem is that there will While substantial business investment in BI
be many more changes in the structure of the systems is continuing to accelerate, there is a
organization and the impact of these changes complete absence of a specific and rigorous
may affect the BI system. So, we need to find method to measure the realized business val-
a solution, and, based on our previous re- ue, if any. BI systems have the potential to
100 Informatica Economică vol. 13, no. 4/2009
maximize the use of information by improv- access both historical and real-time data
ing the company’s capacity to structure a through ad-hoc queries. In essence, managers
large volume of information and make it ac- at every level can have a customized view
cessible, thereby creating competitive advan- that extracts information from disparate
tage: “competing on analytics” [15]. sources and summarizes it into meaningful
BI utilizes a substantial amount of collected indicators. Executives need information for
data during the daily operational processes, strategic and tactical decision that often re-
and transforms the data into information and quires the combination of data from ERP and
knowledge to avoid the supposition and ig- non-ERP application sources. The usual re-
norance of the enterprises [14]. ports developed from daily transactions does
The main characteristics of a BIS are: the ca- not satisfy the business needs, an executive
pability of providing representative informa- cannot take a real time decision based on a
tion to the high-level management, to support hundred pages per month cash-flow detailed
strategic activities such as goal setting, plan- report. Information must be aggregated and
ning and forecasting, and also tracking per- presented with a template based on a busi-
formance, to gather, analyze, and integrate ness model. In the table below we represent
internal and external data into dynamic pro- the main differences between ERP reports
files of key performance indicators. Based on and BIS reports [2]:
each executive’s information needs, BIS can
ERP systems are transaction-processing fo- different sources. The following section de-
cused and weak on analytics. Strategic and scribes some of these techniques and the ar-
executive manager’s demand for technology chitecture of the BI systems.
solutions that can extract, analyze, and vi-
sualize information from ERP and stand- 3 Business Intelligence Systems’ Architec-
alone systems, and this has provided the mo- ture
tivation for a new type of information sys- BIS architecture is structured on three dis-
tems like BIS. The components of these sys- tinct levels [2]:
tems are based on innovative technologies Level 1. Data Management – is represented
such as data warehousing, OLAP, data min- by relational databases, data warehouses and
ing, friendly graphical user interfaces, inte- other type of data sources. At this level is
grating tools capable of collecting, common to use a data warehousing solution
processing, storing and retrieving data from that collects and organizes data from both in-
Informatica Economică vol. 13, no. 4/2009 101
ternal and external sources and makes it (on-line analytical data mining) systems are
available for the purpose of analysis. A data OLAP systems used for data mining, used to
warehouse contains both historical and cur- discover new information from multidimen-
rent data and it is optimized for fast query sional-data.
and analysis. Data warehouses extract, trans- Level 3. Data Visualization Tools - provide a
form and process data for high-level integra- visual drill-down capacity that can help man-
tion and analysis. agers examine data graphically and identify
Although a data warehouse can make it easi- complex interrelationships. BIS attempts to
er and more efficient to use the BIS, it is not present data in a form that is relevant for stra-
required for a BIS to be deployed. Organiza- tegic decisions. At this level, one can find
tions can extract data directly from their host tools for reporting and presenting data in a
system database for their analysis and report- friendly manner. A very efficient solution
ing purposes, but in a more difficult way. that can be used also to integrate data is to
Level 2. Model Management – is the level of develop a business intelligence portal [4].
data extraction, transformation and The main purpose of a BI portal is to inte-
processing. This level is based on different grate data and information from a wide range
type of models for statistic interpretation, of applications and repositories, in order to
analysis and forecasting data. At this level, allow visualization of a multitude of systems,
we can find technologies like OLAP, data either internal or external to organizations,
mining and analytical reporting. The OLAP through a simple Web interface [5]. There-
engine is a query generator that provides us- fore, a BI portal can be seen like a Web-
ers with the ability to explore and analyze based, secure interface, which can offer a
summary and detailed information from a unique integration point for the applications
multi-dimensional database. Traditional rela- and services used by employees, partners,
tional database systems handle this situation suppliers and clients of the organization. The
by using multiple queries. In many cases, the main advantage of the information portal is
queries become so complex that even the de- that it can be easily offered as a service to the
veloper finds them difficult to maintain. wide public [16].
OLAP overcomes this barrier by enabling
users to analyze multi-dimensional data. 4 Business Intelligence Systems’ Develop-
Managers can use an OLAP engine or typical ment Lifecycle
operation like “slice and dice” data by vari- There are some major differences between
ous dimensions and then drill down into the transactional systems’ lifecycle and BIS life-
source data or roll-up to aggregate levels. cycle which depends on decision systems’
OLAP provide tools for forecasting data and characteristics, but the same traditional tech-
“what-if” analysis. OLAP can only mark the niques and stages are used for development:
trends and patterns within the data that was initial study, project planning, analysis, de-
requested. It will not discover hidden rela- sign, construction, and implementation (fig-
tionships or patterns, which requires more ure 1).
powerful tools like data mining (DM). These In these stages there are many steps used for
tools are especially appropriate for large and modeling BIS characteristics such as [6]:
complex datasets. Through statistical or − orientation towards business opportuni-
modeling techniques, data mining tools make ties rather than transactional needs;
it possible to discover hidden trends or rules − the implementation of strategically deci-
that are implicit in a large database. Data sions, not only departmental or opera-
mining tools can be applied to data from data tional decisions;
warehouses or relational databases. Data dis- − analysis based on business needs, which
covered by these tools must be validated and is the most important of the process;
verified and then to become operational data − cyclical development process, focused on
that can be used in decision process. OLAM evaluation and improvement of succes-
102 Informatica Economică vol. 13, no. 4/2009
sive versions, not only building and ma- jor delivering of a singular a final version.
The rule should be: share one coordinated an organization from different functional
ETL process. areas or modules such as: financials, invento-
Step 10: Design metadata repository – if it is ry, purchase, order management, production
used a pre-defined solution for metadata re- we need to analyze and design the business
pository then in this step it is adjusted for model and strategic requests. This model
project requirements, otherwise a metadata have to be mapped on a logical model and
repository is designed in terms of metadata physical model in the data warehouse and al-
logical model depending on data model: rela- so used for extracting and presenting data
tional, object oriented or multidimensional. through OLAP technology. These models are
Stage 5: Development known as multidimensional models and basi-
Step 11: ETL development – filtering tools, cally, they represent an extension of the rela-
procedures, operators are used for building tional model or ER schema or a multidimen-
ETL process. Data filtering and transforma- sional view over facts.
tions depends on data sources quality. These Multidimensional models are classified in
sources are different like: files, databases, e- two major types: models that are an exten-
mail, internet, unconventional sources. sion of ER model are based on a star schema
Step 12: Application development – after and consist in the relationship between some
prototype validation, building the final appli- dimensions and facts or measures and n-
cation may be a simple process. Procedures dimensional cube based models that use a
templates and interfaces are re-built; user multidimensional view over an individual
rights and privileges are granted. situation or data.
Step 13: Data Mining – executive systems In Business Intelligence Systems, the multi-
have to implement data mining capabilities in dimensional model that is used has to be able
order to succeed and accomplish manager’s to overhear the business requests. All we
requirements. This step involves testing algo- need is a business vision over data structure
rithms, data mining techniques like clustering, so the star schema or the n-cube based mod-
predictive and organizing methods. els have to design and incorporate business
Step 14: Developing metadata repository – if aspects or demands not only the facts or the
the metadata repository has to be built-up relationship between data. The managers and
then metadata dictionary and data access in- executives request a synthetic view over facts
terfaces are developed. and indicators and these key performance in-
Stage 6: System implementation dicators are built from the entire organiza-
Step 15: Implementation – it is the delivering tional data or even external data.
process in which the development team or- Also, the system have to provide a friendly
ganize training sessions for managers, final graphical interface with advanced capabili-
documentations and technical support are ties of slicing and dicing through data and
prepared, data loading process and applica- easily get a new perspective over data by ro-
tion setup is accomplished tating dimensions and drill down or roll up
Step 16: System testing – after system im- over hierarchical levels. So we need a multi-
plementation preliminary conclusions are dimensional model in which these operations
made, costs are estimated and the develop- can be made easily, in real time and that can
ment team build a final report in which are it overhead the entire business model with re-
describe system performances and also some lationship between dimensions, facts and hie-
parts which have to be improved or re-built- rarchies and it is based on the entire organi-
up. zational data at operational level, tactical lev-
el and strategically level.
5 Business Intelligence Conceptual Design Based on these considerations we propose an
Model extension of the star or the constellation
In order to gather data from various sources schema but with aggregate data and hierar-
and ERP systems those are implemented in chies in fact tables not only in dimension
104 Informatica Economică vol. 13, no. 4/2009
tables. The model is structured over three model can be mapped on the each level of the
distinct levels and we can call it a pyramidal pyramid;
model with the following structure (for de- Performance in the drill-down or roll-up op-
tailed description see the book [2]): erations – because the dimensions and facts
− Organizational level (or the base of the are separated at each level we can easily na-
pyramid) – containing dimensions and vigate through hierarchies from a level to
facts with an organizational scope, at a another;
general level, that shape and are common Incremental development – the model can be
to the entire activities. Such dimensions built in stages and each stage can be vali-
can be: <time>, <zone>, <product>, dated and used before the next stage;
<currency> and facts: production, pur- EIS, MIS and DSS support – the top level can
chasing etc. Data are at a detailed level be used to implement an Executive Informa-
with multiple hierarchies over each di- tion System (EIS), the bottom and middle le-
mension table. vels can be used for design and realized a
− Departmental level – containing dimen- Management Information System (MIS) or a
sions and facts for the departmental le- Decision Support System (DSS) because
vels of the organization and particular ac- these systems can use the specific dimension
tivities in these departments or field of in- and fact tables from these levels.
terests, group by data marts or data cen- Disadvantages of the model:
ters. Such dimensions can be: <account>, High complexity – because it is containing
<client>, <vendor> and facts: stocks, three different levels, the business model
payments, sales etc. Data are at a detailed need to be careful analyzed and designed in
and aggregate level with specialized hie- order to identify the proper and suitable di-
rarchies over each dimension table. mensions and facts and also the hierarchies at
− Strategically level – containing dimen- each level. An inadequate choice can have a
sions and facts derived from the base di- major effect on the performance of the entire
mensions and facts, with specific ele- system;
ments for the strategic analysis, like <in- Moderate performance of the interrogation
tercompany>, <plan>, <budget> and process – in order to perform a complex
facts: cash-flow, KPIs. Data are at an ag- query the model need to establish many rela-
gregate, synthetic level with specialized tionships and joins between the fact and di-
hierarchies over each dimension table. mension tables and this can reduce the per-
The main characteristic of the model is that formance of interrogation;
between the dimension tables and the facts Top-down and bottom-up development – In
from different levels of the architecture can order to overhear the entire aspects of the
be establish a relationship and also the fact business process we need to build the sys-
tables can have hierarchies and class tems in two directions: first top-bottom to
attributes that can be used for drill down or model the strategic requirements and second,
roll up. bottom-up for validating and setting up the
Advantages of the model: hierarchical flux of data.
Flexibility – new elements or objects like The pyramidal model is suited for business
new dimensions or facts can easily be in- needs and can be developed and imple-
cluded in the model without affecting the ex- mented through an object oriented approach,
isting architecture or remodeling the system defining classes for dimensions and facts,
and the loading process for a specific level following the rules of the OO design. The
can be made without refreshing the whole da- prototype will have the main functionalities
ta; of the business model and when there will be
Real model of business requirements – the any change in this model and a new business
three level architecture is based on the real requirement appear then new functions or
model of business requirements thus this new attributes can be added to the main
Informatica Economică vol. 13, no. 4/2009 105
classes to complete the demands. changes in any type of business process and
functions like personnel, services, and
6 Criteria for Evaluating Business Intelli- processes, as well as new mandates, laws,
gence Systems and regulations requiring the capture of dif-
A problem of BI systems is measuring suc- ferent types of data. BIS should be expanda-
cess. There are some case studies evidencing ble to accommodate data growth and changes
benefits generated by organizations that are to organizational structure. EIS also should
successful with the use of BIS like the one allow contributed content to grow without a
shown in [8] and [9], but with limited empir- slowdown in performance.
ically validated measures. There have been Integration
several studies calling for the development Integration involves two types of issues: data
of a measure for evaluating the business per- integration and system integration.
formance effects of BIS ([10], [11], [12]). Data integration is the ability to access data
In deploying a BIS there are many risks in- from much different type of systems, so BIS
volved: system design, data quality, and will be particularly effective if it can over-
technology obsolescence. System design come the challenge of information fragmen-
risks stem from poor conceptualization of an tation, allowing executives to measure fea-
enterprise’s true business needs before the tures of business processes that involve in-
technology is deployed. Data quality risks re- formation from inside and outside of the or-
late primarily to whether or not data has been ganization. System integration refers to two
properly cleansed. Technology obsolescence things: the ability to extent the BI software
refers to the failure on the part of the vendor with new capabilities and modules and the
to anticipate new technologies. system’s ability to coexist with other enter-
Large budgets and strategic information are prise solutions.
involved in deploying BIS systems – this is Friendly user interface
the reason to establish rigorous criteria for BIS should be designed to allow managers
evaluating such systems. These criteria are who are not trained to use query languages
discussed below. and advanced technologies, a fast, easy, and
Decisions based on business process understandable way to navigate into data and
BIS should not be viewed only as a data re- identify trends and patterns. BIS should per-
pository or a large set of data. Instead, sys- mit the user interface to accommodate differ-
tem’s implementation should be concern on ent degrees of technical knowledge.
conceptualizing new data models, processes,
and indicators that form the content of BIS; 7 Conclusions
also it should provide extensive understand- BI systems have a powerful impact on stra-
ing of the benchmarks that are useful to eva- tegic decisions quality to reduce the time for
luate business processes. making decisions and thus these systems
Performance must have the ability to allow managers to
This feature typically refers to the response view data in different perspective, to drill-
time that a system provides to its users. Most down and roll-up to aggregate levels, to na-
responses should range from a few seconds vigate and on-line query data sets in order to
to a maximum of 30 seconds for routine que- discover new factors that affect business
ries. Response times depend on the com- process and also to anticipate and forecast
plexity of the database and the queries being changes inside and outside the organization.
requested. BIS improve the quality of management in
Flexibility and scalability organization through new type of technology
Flexibility determines whether a BI solution and techniques for extracting, transforming,
can continually adapt to changing business processing and presenting data in order to
conditions after the system has been deli- provide strategic information.
vered. BIS should be able to accommodate One of the major risks in the process of de-
106 Informatica Economică vol. 13, no. 4/2009
veloping a BIS is the system design that [4] A. R. Bologa, R. Bologa and A. Bara,
stem from poor conceptualization of an en- “Technology vs Business Needs in Busi-
terprise’s true business needs before the sys- ness Intelligence Projects,” Proceedings
tems is deployed and for every change in of the International Conference on e-
these requirements the prototype must be al- Business (ICE-B 2008), 26-29 July, 2008,
so revised. A solution for covering this risk Porto, Portugal.
is object oriented modeling of a data ware- [5] V. Diaconita, I. Botha, A. Bâra, I. Lungu
house that helps us to improve the designing and M. Velicanu, “Two Integration Fla-
phase and the development cycle and also vors in Public Institutions,” WSEAS
we can re-use some parts of the prototype Transactions on Information Science and
that it was implemented in an organization Applications, May 2008.
in order to design and implement another [6] L. Y. Liang and R. Miranda, “Dash-
prototype in other organizational environ- boards and Scorecards: Executive Infor-
ments. So, object types can store structured mation Systems for the Public Sector,”
business data in its natural form in object Government Finance Review, December
tables and then allow applications, such as 2001.
OLAP applications, to work in a multidi- [7] M. Z. Elbashir, P. A. Collier and M. J.
mensional way using the object oriented Davern, “Measuring the effects of busi-
properties and facilities. ness intelligence systems - The relation-
ship between business process and orga-
Acknowledgement nizational performance,” International
This paper presents some results of the re- Journal of Accounting Information Sys-
search project PN II, Ideas Program, Code tems, pp. 135–153, 2008.
820: Informatics solutions for decision sup- [8] J. Hesford and K. Antia, A process-
port and for the development of knowledge oriented view of competitive intelligence
management in public institutions, financed and its impact on organizational perfor-
within the framework of IDEI research pro- mance, Working Paper. Richard Ivey
gram. School of Business, The University of
This article is a result of the project Doctoral Western Ontario, 2006.
Program and PhD Students in the education [9] M. Edwards, “Best practices and leader-
research and innovation triangle. This ship in data warehousing award winners
project is co funded by European Social Fund 2002,” Bus Intell J, Vol. 7, No. 4, 2002.
through The Sectoral Operational Program [10] A. Lönnqvist and V. Pirttimäki, “The
for Human Resources Development 2007- measurement of business intelligence,”
2013, coordinated by The Bucharest Acade- Information Systems Management, Vol.
my of Economic Studies. 23, No. 1, pp. 32–40, 2006.
[11] J. Marin and A. Poulter, “Dissemina-
References tion of competitive intelligence,” Journal
[1] S. Luján-Mora and J. Trujillo, “Extend- of Information Science, Vol. 30, No. 2,
ing UML for Multidimensional Model- pp. 165–80, 2004.
ling,” Proceedings of the 5th Interna- [12] L. Davison, “Measuring competitive in-
tional Conference on the Unified Model- telligence effectiveness: Insights from
ling Language, Germany, 2002. the advertising industry,” Competitive
[2] I. Lungu and A. Bara, Executive Infor- Intelligence Review, Vol. 12, No. 4, 2001,
mation Systems, ASE Printing House, pp. 25–38.
Bucharest, 2007. [13] Y. C. Lu and H. Cheng, Automated op-
[3] L. Moss and S. Atre, Business Intelli- timal equity portfolios discovery in a fi-
gence Roadmap – The complete project nancial knowledge management system,
lifecycle for decision-support applica- computational intelligence in economics
tions, Addison-Wesley, 2004. and finance, Springer-Verlag, 2003, pp.
Informatica Economică vol. 13, no. 4/2009 107