This action might not be possible to undo. Are you sure you want to continue?
DATA WAREHOUSE, BUSINESS INTELLIGENCE AND DATA MINING
DATA WAREHOUSE, BUSINESS INTELLIGENCE AND DATA MINING
Is NEXT CHAPTER of our FREE
BOOKS in PDF at
To better understand how decisions and decision making processes impact business performance they need to be first understand and defined. So, in this book we will briefly make introduction into world of Decisions together with Information Systems because they should be analyzed together, not separated!
© Copyright Gabriel I.S. www.business‐intelligence‐secrets.com
DWH/BI/DATA MINING FUNCTIONALITIES
6 DWH/BI/DATA MINING FUNCTIONALITIES
In this chapter Introducing DWH, BI, Data mining Limits and advantages Description of functionalities There are many books that go much deeper into DWH/BI/Data mining topics. Aim of this chapter is not to compete with many excellent materials. Aim of this chapter is to look with users eyes (business side) and through users requests on named Information Systems. This approach is more understandable to parties tried to define their needs and to implement systems. Literature for this approach is neglectable. In this chapter, first of all, authors are trying to explain what DWH/BI/DM actually is, how it functions and what role has it in companies. Afterwards authors will after several brief explanations try to explain strategic advantage and role of DWH/BI/DM. Then will be shown example from practice how with data browsing tool knowledge from data is created. Chapter also describes
© Copyright Gabriel I.S. www.business‐intelligence‐secrets.com
DWH/BI/DATA MINING FUNCTIONALITIES advanced techniques of discovering knowledge from area of statistics and data mining. Since it DWH, BI and DM are separate solutions but also very integrated, in first part of chapter authors analyse them separately but in second part of chapter DWH/BI/DM are described together since they are very tightly integrated.
6.1 Myths and legends
“DWH will solve everything” „Push the button and everything will appear on screen“ „It’s fancy. Our competition has it. Why shouldn’t we have it too?“ „System will solve generated problems instead of us.“ Do following statements sound familiar? Well, might sound cheap but this are everyday statements before projects start in every level of company.
© Copyright Gabriel I.S. www.business‐intelligence‐secrets.com
But, this is… …totally…
DWH/BI/DATA MINING FUNCTIONALITIES
This should never be impulses to jump into waste projects like DWH and BI, because DWH/BI implementation looks more like operation on many internal organs at same time. It is big construction site with many workers on it.
DWH/BI and Data mining are not magical solutions. Brief demystification… DWH is central integrated data repository designed for reporting, and for keeping history. Many core business systems and ERPs are burdened with reporting requirements and because of them reduce operative performance quality. Examples are many where response time of application is critical and should not be slowed down by reporting demands. DWH serves to take data from production, store it and prepare data for reporting and analytics. Step before DWH is creating data repository and create reports. Since DWH has very low successful implementation/operation rate app. 40% it is very reasonable to stay on step before like preparing only data repository and act like DWH but with far less functionalities. BI solution is in simple words reporting and analytic interface consisting of forms, diagrams, OLAP cubes and similar and are based upon data repositories like DWH. Primary function is to publish data in user friendly form. Behind BI interface run logical data sets like OLAP cubes combining data dimensions and interconnect data from production systems. Data mining solutions serve to find hidden – new data (trends, segmentations, behaviors, patterns, tariff simulators and etc.) not visible with ordinary analytical tools. Data mining brings true value add to business.
6.2 About DWH, BI and Data mining
6.2.1 DWH Introduction
As long as owners, managers, investors exist, exists also aspiration of this persons to penetrate into core knowledge behind figures from business. This is important. © Copyright Gabriel I.S. www.business‐intelligence‐secrets.com
DWH/BI/DATA MINING FUNCTIONALITIES
With development of accounting during 19th century reports were stabilized that could serve as basis for company status analysis. During 20th century with standardization raises comparability. In latest years of 20th century, with emerge of new applications for every business process, quantity of information grew exponentially compared to previous data quantities, also grew number of reports with goal to reach core company status. Companies that manage to understand trends on markets modify business and prosper competitively on market. For example, if company wants to analyse sales of 5 articles from product portfolio for 10 customers, revenue & costs in last 5 years it will result at least with 1800 numbers (one paper filled with figures). Without computers and tools for this simple task analysers would have big work to do. PC is help in this example but number of process supporting applications multiply tremendously in big corporations. Available data quantity is also enormous that additionally complicates analysis. During 70s appeared first applications supporting data analysis. They had many deficiencies like user interfaces, integration with production systems – source systems and common lack of power to store and process and this was why they were not in massive usage. With appearance of Lotus 1‐2‐3 and Excel emerge possibilities for users to create own models for business analysis. Model is based upon sets of attributes with goal to present values of attributes in future or attributes for estimation and comparison with other attributes. In 80s appeared so called executive information systems (EIS) applications with promises to provide requested information to management for efficient business. Big problem was to fill applications with data, import time was very long. Beside initial data load in cases of dynamic market and environment time to adopt and add new data into models from sources was very long. Even today EIS products are still sold because as tools upon whose results decisions are made. People tend to make own life easier instead of making others life easier, in this case easier life of those to prepare data from sources for the tool. During 90s SQL language spreads on market for accessing data in databases. This was trigger for ETL tools to appear on market, designed to automate data import process. Interactive tools developed in parallel to access organized data for management. © Copyright Gabriel I.S. www.business‐intelligence‐secrets.com
com . Development of management information tools.business‐intelligence‐secrets. 6.2. There are two types of knowledge BI/DWH systems provides: • Knowledge resulted from aggregations of historic data (quantitative) • Knowledge resulted from models implemented on DWH and implemented through BI system. Main request on production system during data entry is to allow company operative and not interrupted work.2 What is DWH? The Data Warehouse is database of special data structure allowing relatively quickly and simply complex query performance upon larger data quantities. © Copyright Gabriel I. On the other hand DWH is designed for quick and simple access to huge data quantities. www. First implementations of BI software happened in second part of 90s. It is important to mention that BI system in sense of knowledge generation is the source system.S. This functionality makes DWH suitable for making DSS ‐ Decision Support System. Classical production system is first of all designed for data entry. High Users interaction 1992 DWH/BI/DATA MINING FUNCTIONALITIES 72 2002 Data mining – sugestions and solutions) 1996 MIS – data analysis EIS – data analysis 1985 Low Low Data agregation (sum and average) Reporting High Analytical capabilities Figure 12.
planning and making business decisions. Iterative nature of model building data warehouses and thus iterative nature of building the software system for the extraction. This is one of the reasons why the projects are building data warehouses.S.com . On the other hand. © Copyright Gabriel I.business‐intelligence‐secrets. 2000. Best is to see through wonder wheel. making the process of building extraction system takes between 70% and 90% of the total time required for the construction of warehouses. University in Zagreb. review and analysis of large amounts of data. That is periodically automated data transfer from the source to the destination of the production data warehouse. largely subject to the failure6. building DWH system is becoming the system which is very difficult to accurately determine time of the construction. and use it for evaluating the results achieved. DWH/BI/DATA MINING FUNCTIONALITIES 73 Daily stored data into production systems at the end should serve to management. Master's Thesis: Building DWH in changing environment. problems related to extraction of data represent the biggest challenge. 6 Citation ‐ Edin Hadžavdić. 6. Data Warehouse provides exactly such mode that is faster and easier access to information. Problems related to the construction of data model is quite well described in the literature and is not a problem too.3 What is BI? One non standard approach would be to see what users on Google primarily search for under term Business Intelligence. When building data warehouses implementers face specific problems that do not encounter in the construction of production (transaction‐oriented) information systems. Most of the problems are associated with the construction of systems for the extraction of data. Figure 13. as shown in practice. When combined with problems that arise because of the iterative nature of building models and data extraction systems. Google wonder wheel results for term Business Intelligence.2. www. Administrative structure of the company should be able to extract useful information from large amounts of data. Some of the problems that are encountered in the construction of the warehouse are: Gathering of different data from multiple sources (multiple production systems) implemented on different platforms. with a time measures of the reach of seconds or minutes. Quick detection of changes occurred in the source system. For this purpose it is necessary to ensure a quick and easy access to data stored in complex structures of production systems.
data mining. OLAP7. This is pretty much very good description of what standard BI solution does. DWH/BI/DATA MINING FUNCTIONALITIES 74 Figure 13. 7 Online Analytical Processing. approach invented to rapidly answer multi‐dimensional analytical queries © Copyright Gabriel I. Google wonder wheel results for term Business Intelligence Users for term “business intelligence” mostly mean and use: dashboards. analyst. warehouse..S..business‐intelligence‐secrets. www. vendors for BI.com .
Business will feel positive effects of BI apply. controlling and planning.S. BI can only analyse information. Standard BI solutions can in very limited manner and in area of only non financial performance indicators support: monitoring. Business Intelligence combines technologies. Therefore BI is tightly connected with DWH or data repositories in practice. Second wave. forecasting. information needs will explode and with usage of too many queries. products to organize key data needed for profit improvement as well as performance improvement. ERP and similar. is not pleasant. not a technology nor methodology. Business performance can be boosted by certain actions and decisions based on business analyses and information focused around key business processes. different approaches. Immediately after standard BI solution is applied business will experience business first BI effect.. www. marketing. data overload.4.com/hub/what‐is‐Business‐Intelligence (26. Standard BI cannot analyse in detail and thorough financial data flow (revenue and costs). in 1958..2010) 9 ibid © Copyright Gabriel I. Special modules and solutions are needed and not standard what BI offers. DWH/BI/DATA MINING FUNCTIONALITIES 75 Hans Peter Luhn for the first time.”9 Business Intelligence is also known as competitive intelligence. Standard BI cannot synthesis information. In order to support key processes like strategy. analytics. ☺ Standard BI solutions can give approach to core data in very comfortable way.) will help to improve main intention of owners and top management and that is business performance by income increase and reduced costs. dashboard. BI effect.com . Main focus is on key business processes. data access but will not give new value to information. Information used in BI serves for Decision supporting ‐ making and acting toward business performance improvement. described Business Intelligence as „the ability to apprehend the interrelationships of presented facts in such a way as to guide action towards a desired goal”. 8 Source: http://hubpages.. standard BI solutions increase data awareness and that is excellent functionality.. budgeting. views over data consequence will be data hyper production. 8 “Business Intelligence is not a single product.business‐intelligence‐secrets. data clutter or data tsunami. BI will shorten data calculation. Common belief is that standard BI solutions as already presented (OLAP. sales and similar BI needs referential data feed from core production systems like CRM. but this is far away from insider information that is usually needed to significantly move business. effective methods.
order fulfilment and billing. operating processes – like – manufacturing. It usually covers a certain part of the company's operations. Therefore.business‐intelligence‐secrets. Many failed in belief that standard BI solutions are very powerful. 6. sales).4 What is Data Mart? Market Data (DATAMART) is a component of the data warehouse. www.g. campaign management and of course sales. data mart bookkeeping. the market data for its functionality is complete and can exist for them as standalone model. DWH/BI/DATA MINING FUNCTIONALITIES 76 revenue generating processes – for example – marketing.com . Data Mart is or is not designed as a component of a large data warehouse. All named operations are very limited and might mislead positive efforts in project starts of BI solutions. customer service.2.S. It is data and logical standalone “island”. inventory management. channel management. © Copyright Gabriel I. The rule would separate market data must be on separate computers‐servers because they themselves are in some way logically separate entities. logistics. aimed at a particular group of users (e. marketing.
Master’s thesis: Izgradnja skladišta podataka u promjenjivim uvjetima.S. © Copyright Gabriel I. 2000. www.business‐intelligence‐secrets.com .2. DWH/BI/DATA MINING FUNCTIONALITIES 77 Figure 14. Datamart example – material accounting 6. Sveučilište u Zagrebu.5 Difference between DWH and production system The main differences between production systems and data warehouses are summarized in the following table10: 10 Quote ‐ Edin Hadžavdić.
Input data Manual entry of individual Automated entry of large amounts of records from the operative data collected from the source. The non‐working hours: a small number of transactions performed by reading and enter a very large amount of data (data extraction). business. production systems. www.S. Quick detection of changes occurred in the source system. Mode / operation The working hours: a large The working hours: a small number of which is carried number of small transactions transactions performed by reading a out of the system that generally perform a smaller very large amount of data. amount of input data.) at a time when the source system is loaded. of possible errors in input Basis for strategic and everyday decisions. DWH differences 6. User type Operational companies Administrative structure of the company.3 DWH importance Advantages Data Warehouse brings to information reporting system are: Merging of different data from multiple sources (multiple production systems) implemented on different platforms. Table 1 Classical production systems vs.. Data storage. The frequency of Continuous intake during Periodically enter (once daily.business‐intelligence‐secrets.com . weekly. © Copyright Gabriel I.. monthly . DWH/BI/DATA MINING FUNCTIONALITIES 78 Classical production information Data Warehouse system The main purpose Data entry by the operative Read data (reporting) of the business. data entry working hours. administrative structure of the Organization and minimization company.
Error detection in production system Long‐term storage of data (typically 5 to 10 years) in relation to production systems (typically 1 to 2 years) Aggregation of data is an important feature of the data warehouse. The problem of timely collection of necessary data. input data in the data warehouse is done automatically. physically separated. but only read from the database storage of data. nor necessary. One day delay can make a significant difference. here users encounter the fact that the warehouse data always have the old data from yesterday or last week or month. Only the data warehouse does not allow a direct. Of course. Such non‐integrated information system is a major problem for the system of reporting within the company. DWH/BI/DATA MINING FUNCTIONALITIES 79 Iterative nature of model building data warehouses and thus iterative nature of building the software system for the extraction. controllers. Manual entry of individual records in the data warehouse is not allowed. the data warehouse database is calm ‐ not any data entry is done. and built on different platforms. Data Warehouse uses administrative structure (experts. week. which may not be the case with statements from the production system in which the data fluctuate due to © Copyright Gabriel I. In the time between two refreshes. For example. Just the process of collecting and combining data from all available sources is the most difficult task in building a data warehouse. manual entry of data into it. periodically and in large quantities. the inconsistencies among the reports obtained from various sources covering the same area of business within the company. Moreover. This may seem like a disadvantage but the purpose of data warehousing is such that the state does not seek what is precisely in real time. The information system of companies in many cases consists of multiple subsystems.business‐intelligence‐secrets. This work will perform software system that must build and run in defined time intervals. as between two refresh does not perform any input into the data warehouse database. when it comes to foreign business partners of the realization of what is charged in the same period the previous year" or "What are the most problematic categories of users in terms of return the loan and how much is the average delay in the case of married male with more than two children? ". www. management) of the company and generates following questions: "How much I earned in the last month. because the data is already entered the production information system company (it is the basic purpose of the production system). reports made at that time will certainly be consistent. On the other hand. Daily data warehouse refresh period is quite sufficient for the first question. or as month depending on how up to date data is needed. while the monthly period more than good for the second question that takes into account the historical data that can reach up to ten years ago. Data Warehouse does unite all existing data sources and makes them accessible in one place. Each component of the production information system is a potential source of data for data warehouse. can be made decision to import data at the end of each working day from any available source and to make data aggregation and transferred to the data warehouse. are reporting inadequate. While the individual data (data warehouse refresh period) may be one day.com .S.
i. it is very difficult to find information. etc. Production systems Program systems for periodical warehouse refresh DWH database External data Figure 15. Second requirement is extremely important. Availability of persons who are sufficiently familiar with the structure and content of the source system.business‐intelligence‐secrets. let us create a data warehouse ‐ it will tell everything about our business. www. if users ask question: "What is the status of bank account?” then users will not use the data warehouse but production information system that shows what the situation just in this minute or second. a list of measures and dimensions that the user wants the data warehouse database.S. Without fulfilling requirement obviously cannot go into the design phase of the reach of data.e. the desire to buy a data warehouse "out of the box" by various distributors of such software. move into the construction of the warehouse project means project collapse. First requirement means that there is at least a scratch data model. DWH/BI/DATA MINING FUNCTIONALITIES 80 the continuous input.1 Preconditions for building systems for data transfers Detailed elaboration of the process of data transfer can start only after it meets following conditions: Defined as (initial) requirements of users in terms of necessary data. Of course. When somebody knows the source system and is not always available for team building data warehouse.3. Often the case in practice is the idea. or to determine the exact algorithm to obtain information without a good source familiar with the system.com .. Unfortunately. Coarse DWH import scheme 6. without knowing what the company wants to know. team can easily ignore some important facts related to the complex structure and content of source data a © Copyright Gabriel I. Because of the complexity of the sources. The source system is not able to build or modify the logical structure.
Update data is completely automated and requires no action from the people. and thus necessarily delaying the project. The results of these programs basically are on the summary level and do not deal with the details like individual records (for example. etc. In case of any mistakes in the production system was in the process of refreshing the relevant people are automatically notified (builders warehouse developers and administrators). Tools for interactive viewing of data warehouse (which is already implemented and running) are different from tools to build a data warehouse and are commercially available products or custom applications. and the major companies. 81 6. etc.2 DWH live and analytical tools Working with data warehouse can be seen as two separate parts. which leads to user dissatisfaction. Typically. often larger than in typical applications used in company. the algorithm reach the data may be subject to frequent change.S. these reports users must expect from the production system rather than data warehouses). www.business‐intelligence‐secrets. Data Warehouse has a certain amount of time in which the data is refreshed.3 DWH as vicious cycle of quality Decision support systems like DWH have become a common tool for a better introduction into own business in most of the world. Reasons for the third set of conditions are obvious. the production system should have the better answer to the analytical and data warehouse to the question of synthetic character. If this situation is repeated for several times in the project of building a warehouse. These programs are modified to work with data warehouses and are intended as support for administrative decision‐making. and to relieve hardware resources to be carried out mostly at night and not disturb the normal operation. the data warehouse updated once a day. Initial costs might seem not so big investment. One is the automated process of daily data import and the other is an interactive work users with applications where the data source is data warehouse. DWH/BI/DATA MINING FUNCTIONALITIES result of incorrect data retrieval algorithm.3. They differ from OLAP tools mainly because they are more customized to company for which they modified and for reports what company needs. but should be counted once built warehouse does not work on its own and requires the © Copyright Gabriel I.com . 6. typical reports are by region rather than by customer reports "from first to last”. Fact that should be kept in mind is that investments in data warehouse are large. This deadline cannot be disregarded. If the structure of the source system is not stable.3. Speaking in general. consequences are constantly changing code.
That is why it makes sense to build in a warehouse a warning system that tells sale of the site YYY is out of statistical framework. Scheme of one such process of data extraction through the filter is shown in Figure 16. Filters can be more complex. Data that have not passed the filters are candidates to be the data trash ‐ error. some of them may be conditioned by the business process (non‐existent customer cannot make a payment) and statistically. DWH/BI/DATA MINING FUNCTIONALITIES attention of one or more persons depending on DWH size and number of users who use it.S.500 USD intake of the application is acceptable. Categorizing problems identified by the process of analysis and reporting in general it is possible to install additional business rules to reduce and lower incorrect entries in the production system. During the data extraction process. 110. These filters are installed in the extraction software. Data cleansing. Adding working hours of few people who are engaged in data cleansing identified as ''garbage'' and optimization of response time to queries and similar jobs for which DWH raises initial costs for several times. data can be monitored and filtered (data cleansing). but it is already known that the workers in data entry can make mistakes.000 USD.business‐intelligence‐secrets. at a certain point of sale YYY where otherwise charged 10. engaged and responsible persons for the data warehouse must be of the same companies that know how to operate production systems ‐ sources of data for data warehouse. it is likely an error if one day we have over 120. Warehouse with respect to its role in detecting not logical data (errors are much easier to perceive in the character of tools for data access) can be seen as a proof reader of production systems. For example. the simplest filter by entering the date of payment for a party looking to whether the date in the current year. Data cleansing © Copyright Gabriel I.000 USD. Depending on the number of such sources and the time and comprehensiveness of data in the warehouse such employees are often very valuable to company. Production system1 Statistical filter 82 Produkcijski sustav 1 Statistički filtar Data Skladište podataka warehouse Production Produkcijski sustav 2 System2 Skup poslovnih business pravila Set of rules Privremeno Temporary područje area Figure 16. www. Without these processes data warehouse becomes rubbish. In most production systems during the year 2000 for various reasons known to find the 1900th year. For example. On the other hand.com .
Based on sales data that is obviously outdated and unattractive service. For example ‐ service XX shall be deactivated because it creates a loss of 12% per month. reports are valued only the number of pages are printed. Data Warehouse contains insurance information back fifteen years. Data provide a picture of what happens in a particular segment of business enterprises.4 The strategic value of Data Warehouse The company in its daily operating‐production systems collect large amounts of data. Data Warehouse is a valuable tool and knowledge system for people in business decision‐making processes. This means sooner decision makers start to begin to make decision based on available data stored in the company.5 million users in some way justify DWH team being proud of the installed information system.com). but also recalls the hundreds of reports that were generated before (and are still). Unfortunately this is not possible. On the other hand. which falls under the control role. © Copyright Gabriel I.3. like the streets with riskiest insurance of burglaries and which are safe. Some services should be viewed through a long series of years to see whether the investment in the service was worth.business‐intelligence‐secrets. DWH/BI/DATA MINING FUNCTIONALITIES 83 6. In this way DWH users can establish premium paid by the insured. With such approach 4 TB of data itself is 11 Data mining ‐ A class of database applications that look for hidden patterns in a group of data that can be used to predict future behaviour. etc. Therefore imposes the need to buy almost as soon as possible data warehouse solution that is in the package.S. After the data was collected next important step is transformation into knowledge. and none of them do not really read and recognizes report structure value. i. These data can be of various structures.e. system can identify workers who often make mistakes when entering data (if added dimension for each employee entering data in the stock). (www. Say that has 4 TB of data on 1. it is possible in a narrow sense. such as the data ‐ person Y on 23. For example.com .webopedia. Knowledge enables company to base important decisions for future business. but for now lets call this finished product ‐ storage. The amount of data can contribute in shaping users knowledge of enterprise business processes through data mining11. www. Commonly thought that the amount of stored data such as measured in a Terabyte‐in data warehouses are important and quality should be forgotten. users can learn complex data that could not be assumed at the beginning of DWH project. On the other hand. It is possible to find simple things such as which services during the last n‐years were profitable and which carry long‐term loss. company will sooner benefit (benefit achieved from the project before the Data Warehouse).3. paid 200 USD for subscription service or as Z ‐ sales in continental region A has stagnated at 11% compared with the same period of previous year.
com . DWH/BI/DATA MINING FUNCTIONALITIES 84 actually slightly useful information and that is only a large amount of numbers that are in the warehouse from which there is little use if they are of low quality. Production system periodically deletes data.S. On the other hand these data can be useful in statistical processing. In company does not exist a sufficient number of people who can devote 100% of building data warehouses. www. Another often neglected aspect is the fact that the data warehouse system serve for the documents storage. It is certain that they do not have to keep data about payment of permanently disconnected customer. Why is this so and what are the reasons that company own forces are unable to build a data warehouse? Here are some reasons: There is no insurance sponsor of management structures. 6.3.business‐intelligence‐secrets.5 Successful implementation of DWH project? Researches show that about 50 ‐ 60% of data warehouse projects fail to set goals. Top management must be interested in building a data warehouse if the analysis proves the validity of its construction and it must be supported with resources. statements made without the necessary historical context for this function will never satisfy users. Top management support is manifested through the sponsors of the project. It comes to other essential functions of a data warehouse ‐ to the prediction of customer behaviour based on previous behaviour. Reporting in the classical sense and prediction of customer behaviour (such as objective analysis generally) justify the investment into data warehouse. the project simply collapses. Prediction of client behaviour production systems cannot efficiently provide. © Copyright Gabriel I. Data Warehouse together with BI may be a way to display the data from production systems that are already adapted for entry but not to the analysis of data. and if not then it is slow. Production system can meet most requirements required ‐ report (although they are more complicated to produce compared to reports on data warehouse technology) but they do not contain historical information. About 70% of the failed warehouses were built with own forces. The primary question for the management of company should be to have a clear picture of what the data warehouse will serve for. If they do not want to use to use it. BI is primarily intended for people who make decisions in the company. Of course data warehouse is usually both but the question is which role is primary. If production system does not delete data and is still fast hen it’s the case where hardware is at the time of purchase prepaid and unnecessary.
4. In this chapter authors will try with a concrete example. as well as to the nature of their business have a greater and wider experience. Let's look at the following components of an information system company. Resistance can be manifested from the leadership structure to the lowest levels. Allocating enough resources and ''do not bother them'' with operational problems through such a long period is usually impossible and if it is possible then these people definitely were not valuable for existing production systems and were in excess of the beginning. www. and it is necessary for its smooth flow.business‐intelligence‐secrets. DWH/BI/DATA MINING FUNCTIONALITIES 85 Should know that the construction of a typical warehouse takes 1 to 2 years and usually mobilize 3 to 10 people. Involvement of external vendors to build a data warehouse with employees inside the company. containing information and information (re) combination provides the content that is meaningful information.com .S. Possible resistance within the company in introducing new technologies ‐ BI in regular operation. Resistance is inevitable and if it is too big data warehouse project will fail. as illustrated in the Figure 17. As already stated the construction by own forces succeed in only 30% of cases and because there is a real danger of failure of building data warehouses if company does not engage specialists outside the company. 6. characterized by a large number of users. ERP ‐ DWH ‐ Portal. In the middle is middle management level together with data © Copyright Gabriel I. Distinguish the production system to support the business (the bottom of the pyramid. For example external specialized IT experts on the basis of these requirements often know how to recognize future problems and know what to do for their removal before the actual need arises.1 Performing knowledge from data ‐ OLAP tools Production (ERP) systems usually contain a large amount of data that follows the business. but the fact is that the external IT experts are not burdened with company problems. after brief introduction. to illustrate the difference between knowledge and stored data. especially if are for long time involved in data warehousing. Idea behind is that IT within the company knows better than the external data and processes in companies.4 Knowledge creation from data 6. Staff and management who will mostly use applications based on data warehouse technology must be open to new information technologies.
etc.) should be sufficient to manage the enterprise management. balance scorecard application. Unfortunately. and whose task is to deal with aggregated information in the monthly reports.com . etc. Well established system of key performing indicators of business are located at a single site (e. What is a key indicator (KPI = Key Performance Indicator)? The key indicator is unambiguously and clearly a number of whose growth or decline is unambiguously interpreted as a positive or a negative shift in the quality of a segment of the functioning of the company.S. Number of warehouse users is much less. but it should be the basis of short‐term top management decision‐making. At the top. since some of the indicators are common (e.business‐intelligence‐secrets. Portal. even for the industry there is no universal system of indicators. which means that they must be developed internally in the company. asking for larger and larger monthly reports in order to read only the overall result at the end of a set of reports. www. etc. or other tool to view aggregated data. © Copyright Gabriel I. at the very top senior management is not asking for a large amount of information. There is quite an important role of indicators. Figure 17. Financial indicators) but some are a result of competitive advantages (what distinguishes the company in the market) and must describe increase or decrease in the segment in which the company differs from others in the market.. which however needs to know to set up an enterprise. DWH/BI/DATA MINING FUNCTIONALITIES 86 warehouse.g. ERP ‐ DWH ‐ Portal It is often the case that management borders with large amount of information.g. department plans and analysis. the amount of information is reduced. typical users of data warehouses are heads of departments.
According to the top of the pyramid rules become less determined. How this process works in practice is best illustrated by a series of images from one example system developed for the local distributor of food products. modelled information becomes knowledge. because unfortunately can not be said that increase in marketing costs by 25% mean (necessarily) an increase in sales of 10%. 6. www. at the bottom of the pyramid one rule can be defined as "With dispatch and loading goods invoice is issued for the customer”. DWH/BI/DATA MINING FUNCTIONALITIES 87 As moving towards the top of the pyramid business model increase undeterminable state compared to the bottom of the pyramid. price. From the product overview it’s easy to spot that cities A and B distribute more Ice cream products than others. In the event that can enlarge the amount of information (e. Selling articles can be seen as selling items at distribution centres where goods come out from all the warehouses for Ice cream products. employee. So. Users must be able to choose the customer and the product and see a sale. Data Warehouse is the place to meet these influences and therefore should be consolidated into a meaningful set of all the information generated by the company as a whole. and it should be presented transparently to the user without cryptic names characteristic for design of RDBMS. enterprise influences are increasingly coming to the fore. © Copyright Gabriel I.2 Taking knowledge from DWH Data Warehouse is a database with denormalized structure as described in previous chapters. It is knowledge that can be derived on the amount of the statistical basis of its characters.) amount of information is created in the user's mental model. quantity sold. At the top of the pyramid prevailing external influences on the model (ie cost substitutes on the market) towards the bottom of the pyramid. item.com .g. It's sort of claim that comes on the basis of large amounts of information. a claim that working on the printing invoice in any business unit (for the impossibility of comparison). customer. In Figure 18. The model of the warehouse must be translated for the user (if not out of the design of certain table) on a business language. Thus.4.business‐intelligence‐secrets. This policy is relatively easy to implement in application. where after confirmation of take over these documents invoice is printed in application for the goods in shipping. etc. Can be said only (and not for sure) that increasing marketing means increased sales. all accounts in last 3 to 4 years) to display proper grouping by subjects in the business process (department. we cannot impose because the horizon of information is relatively limited at the bottom of the pyramid. or any other formula defined percentage.S.
For more serious conclusion deeper analysis of historic data is needed.com . WH – Milk industry sales 2004 © Copyright Gabriel I. Selling articles Can be stated that conclusion has been made from converting knowledge about sale. In a longer period of time and in the picture below for example Milk industry products can be seen that the sale during the summer is much better. data view should be expanded for a longer time interval in order to test the hypothesis that we have just presented. can be concluded sale during the summer goes about 40% better. WH – Milk industry sales 2003 Figure 19. More to say. As the graph also present.business‐intelligence‐secrets. www. DWH/BI/DATA MINING FUNCTIONALITIES 88 Figure 18. Ice creams are better sold on north coast compared to other selling regions. it can be seen under the Figure 20.S.
www. where can be also noted significant increase of sales in August. WH – Milk industry sales 2003 The conclusion can be made (perhaps even wrong) that all the products of company are sold more during summer. DWH/BI/DATA MINING FUNCTIONALITIES 89 Figure 20. Here's also a version by the daily distribution. but through a longer period of time. © Copyright Gabriel I. Sales drop in winter is noticed! Drawn curve describes the seasonal oscillation.com .S. and in some ways follows the mental model ‐ a better summer – worse winter. Now can be tested same thing for meat industry.business‐intelligence‐secrets.
WH ‐ Example of digging deeper ‐ query. DWH/BI/DATA MINING FUNCTIONALITIES 90 Figure 21. WH ‐ Example of digging deeper ‐ "to see a deepening asked to answer what is selling so well.business‐intelligence‐secrets. In Figure 22.S. © Copyright Gabriel I.com . but the model provides considerably more and that is possibility of dynamic deepening inquiry into the details. and where the most goods are sold. and then under Figure 23. WH – Meat industry sales What is shown here is a time dependency of selling brands. Now it is known where to ask and what is sold there. users could be interested where in RIJEKA is a better sale. Of course. www.
DWH/BI/DATA MINING FUNCTIONALITIES 91 To see what are prerequisites for DWH.business‐intelligence‐secrets. www.com .S.business-intelligence-secrets.com/business-intelligence-pdf © Copyright Gabriel I. Business Intelligence and Data Mining download our FREE BOOKS in PDF at http://www.
S. WH ‐ Example of digging deeper ‐ query Figure 23. Data © Copyright Gabriel I. although by its nature is not related to each element of the business process).com . which is relatively easy to display in a simple report.business‐intelligence‐secrets. WH ‐ Example of digging deeper ‐ query One of very common analysis of universal character for each company is ABC 5 analysis (of customers in the observed case. DWH/BI/DATA MINING FUNCTIONALITIES 92 Figure 22. www.
The fact that most of the revenue brought small number of customers (large accounts) and that they therefore should be given more attention and consider them important customers. Can be looked for important customers upon such KPIs like customers whose rank (RUC) ‐ rank (total)> 5 during the last year. for example customer on 24th place. WH ‐ ABC analysis of customers (with deleted names behind third place) can be seen that 60% of traffic make first twenty customers. Quality key indicator tells about how many of these customers have gone. and its increase means that company feel problem directly in the profit and it can be built into the system of indicators. It is also a comparison of the ranking by total income and profit. Can be also noticed those who are at relatively high‐ranking RUC. In this illustrated example large amounts of data stored in the model allow interactive analysis of customers and work with such a large data set. 93 Figure 24. www. Those customers are very important. WH ‐ ABC analysis of customers (with deleted names behind third place) revenue company will be compromised.S. ABC analysis targets the relative importance of observed elements classified in business processes. and that their purchases have declining trend. DWH/BI/DATA MINING FUNCTIONALITIES storage system should provide by its features simple creation of this report. What can be concluded from the analysis.business‐intelligence‐secrets. WH ‐ ABC analysis of customers (with deleted names behind third place) In Figure 24. The buyer is in the eighth position of bringing revenue to company. specifically in this example by customers.com . Such KPIs. what knowledge can be made on the basis of that? Can be concluded that leaving of any of the customers from group A (see last column in Figure 24. on the strength and weaknesses © Copyright Gabriel I. Loss of any of them would significantly and visibly reduce income of the company. Users come to knowledge of business processes.
more or less easily come to a conclusion without too much help of mathematical apparatus. DWH/BI/DATA MINING FUNCTIONALITIES of sales.1 Introduction into data mining processes What is data mining? The term data mining (DM) is considered a class of applications processing a large amount of data looking for hidden patterns and regularities that can be used to predict future behaviour. Some typical examples are: Decision tree constructed from the history of the membership. and thereby attract new customers "Diapers and beer" ‐ looking at transactions from the retail environment to conclude why consumers often buy diapers and beer. and about how relatively easy can use this data structure for senior management reports. Area of machine ‐ learning is used to enable software to learn some of the models themselves.S. Data mining term is relatively wide and covers a larger set of methods arising from mathematical statistical methods.5. but also other processes which people used without computers assistance. and visualization techniques are important to prepare data. diabetes genes responsible for its formation) Data mining has resulted in several scientific disciplines whose multidisciplinary synergy achieved combining the effects of which are important: Statistics Artificial intelligence ‐ especially the so‐called resort. www. 94 6.business‐intelligence‐secrets. Finding regularities in the behaviour of tourists in order to provide them different models of discounts. "Machine learning" Research algorithms for clustering Visualization techniques Databases Statistics is at the heart of most data mining methods.5. but to find more familiar types of customers and offer them something "more". and that they are part of the process © Copyright Gabriel I. with purpose to decide whether a potential member will get a credit card or loan or not. especially in the case of neural network etc. Why is it so and its cause? This is not the aim of sellers.3 DM clustering. Algorithms for clustering are described later in Chapter 6. Fact is data sources in this case are almost always located in the database. and some believe other data mining methods are also part of standard statistical analysis.5 Advanced methods – data discovery (Data mining) 6.com .g. In search of human genome DM methods helped in discovering the causes of many hereditary diseases (e.
its preparation may require a long period. Extracting details in this process often requires a very good knowledge of ERP systems. www. Data collection Longest process in time perspective. Typical examples of sex ('M'.g. even a few months. Take for example credit card house. Rare are transactions with a low amount of purchases and other types of point of sale. ie to recognize the fraud on the cards (preferably before they occur). Are results relevant or not of results should be decided by expert for area of analysis. With the example described below. and considering that they are object of work. mostly by mistake. Most of the goods were purchased in stores such as "retail chains”. Later results are tested and compares with the second set of results that are known well. Data mining process can be divided into several important steps: 1.S. bigger set of data has to be prepared in order to get more relevant results. 'F'). Data cleansing Cleaning of garbage data is also a long process and it has to be done on a set of rules with the attributes that are used for analysis. 2.com . 100) where are evicted all rows that contain some obviously false attribute values appeared for any reason. can be seen that the buyer based on habit from past transactions to buy goods from 50 to 400 USD. part of the preparation of data. It is necessary to mention here that this is less interesting but vital part of data mining. the results of mining are not relevant. These are all large quantities data that even only handling with them is relatively big problem.. Very often due to lack of a testing data and visualization set. and the second part group set to test the hypothesis. Data mining most important functions Data mining software can help companies of different industries in the prediction of behaviour of their customers. Pattern DM in the narrower sense ‐ the execution of the algorithm recognition 4. The transaction (although it may be entirely legal if the buyer intends to engage and buy engagement ring) © Copyright Gabriel I. Goal is to judge how well DM algorithm learned and foresees results. Evaluation Not every discovered fact is true. Table 1. DM software is often used for so‐called "fraud detection".. How does it work? Take into account the historical behaviour of members. Creating a In the domain of machine learning data set is divided into 2 groups ‐ set for test set learning.1. DWH/BI/DATA MINING FUNCTIONALITIES 95 preceding analysis.business‐intelligence‐secrets. If having in the event a production data source. 3. So the customer buys food and similar goods. Suddenly appears in the type of trade transaction is marked as a luxury goods with a very high amount of the transaction. the distance from the position in graph square where most purchases are also very large. Can be concluded that the difference from the average amount is high. neural networks) learns. 2. years (18 . which has its own habits in n‐number visible from his past transactions. With the first group computer ‐ DM algorithm (e. Also it’s very usual it comes to transactions such as payments for POS terminals must be also prepared.
As part of further analysis there will be a brief overview of some important methods and possibilities of their application. etc? 96 Luxury stores Restaurants Amount of transaction Type of store Other stores Retail stores Figure 25.. Mapping the credit card transactions in the two‐dimensional system Can be concluded that many DM techniques have their application in various aspects of company business such as: Fraud Detection Customer Segmentation Sources for business decisions . and methods that may not fall in any group of Knowledge Discovery ‐ Visualization.S. DWH/BI/DATA MINING FUNCTIONALITIES can be considered (at least spoken) suspected.. Is it really the result of theft and the thief attempts to quickly and easily buy goods payable in the form of gold / jewellery. www.business‐intelligence‐secrets.com . Special attention will be paid to describe clustering methods. © Copyright Gabriel I.
1844 received the Royal College of Surgeons of England. In 1849 published the paper "On the Mode of Communication of Cholera (way cholera spread). the London doctor after graduated school was admitted for an assistant surgeon in Newcastle‐on‐Tyne in a private school. provided the conditions for research.5. Unfortunately. 6. DWH provides very easy access to data. J. DWH/BI/DATA MINING FUNCTIONALITIES 97 6. © Copyright Gabriel I. Site is today in the picture below. One of them was Figure 26. With the help of the marked area and discussions with the ill and their families successfully locate the source of infection at the pump in Broad Street. Snow had no opportunity to confirm his theory. Snow made the folder where he marked incidence of cholera for the part of London. In the case of WH systems. and accidents came very quickly. London received water from the two companies. At that time.5. what is not the case in production systems. Cholera epidemic in 1854 in London. Official theory was cholera spread through the air. We can threaten to dash the very common one in the London area.2 Role of DWH in data mining WH systems can be seen as typical and very good source for the DM software. DM team members no longer have to know the specific organization of the production company systems.com . but have a lot of data structured and prepared.S. Since the data are structured in tables only needed thing is a team to start DM.3 DM clustering John Snow (1813‐1858). where he explains that cholera is spreading through contamination of drinking water.business‐intelligence‐secrets. Otherwise. One work that he made is interesting in the context of clustering. but appear in distant places. After that. The Incidence of cholera for the fourth London 1854th water pump the Thames upstream of London and the other downstream. www. the process of DM preparation is significantly shorter. attended Huntierian School of Medicine in London. this part of the job consumes a lot of time. or breathing in the vicinity of patients. J. DWH shortens and eases DM implementation.
www. Mathematical clustering basis Set of points can or cannot be considered as cluster. y n] can state: © Copyright Gabriel I.S... which is similar to reiterate today with the help of computers. . y 2. of address. Snow was then a manual clustering. etc. ..business‐intelligence‐secrets.. Imagine cases of cholera as transactions in the system. daily movement. x 2. The intention was to draw an interesting parallel.. In short if x i = [x 1. x n] and y = [y 1. DWH/BI/DATA MINING FUNCTIONALITIES 98 Today his work may not seem revolutionary.. each with its own attributes. Figure 27. Cholera frequency in London blocks 1854. What has worked J. Take a set of points and determine their mutual distance in n‐dimensional space. but the way the data set was made upon ideas later helped in the early stages of the suppression of countless epidemics. spread the knowledge of scientists of that time and created a significant foundation for the development of science.com . and show that one method.
Then. taking into account the distance of points in the cluster from the centre of the cluster is smaller than other points. but very good technique for spotting patterns of behaviour and anomalies of these forms bounce. There remains the problem of how to find the centre of clusters. it is necessary to form n clusters. Of course at the beginning of the algorithm for a large number of points can be defined and the expected large number of clusters. It is possible to conclude that clustering is not new technology that came with appearance of powerful computers. 5). Let’s see the picture: Figure 28. points 1 and 2 assign the role of centerioda in two corresponding clusters. www.com . L = DWH/BI/DATA MINING FUNCTIONALITIES 99 ∑ (x i =1 n i − yi ) 2 Therefore. Adding 5 points which is closer to the point "a" than "b" and the centeriod moves to a point "c" and point 5 is assigned to the cluster that now contains (1.business‐intelligence‐secrets. can easily recognize the anomaly and to know what to expect in the statistical © Copyright Gabriel I. Internet is a model of clustering represented as 10 8‐dimensional space). determining clusters ‐ a framework for behaviour.S. Problems that need further steps to solve in the n‐dimensional space (for example. When inserting next point 3. so called centeriode. point 4 is closer to point 2 and allocate the cluster containing the point 2 with time to adequately centeroide moves to point "b". which is close to point 1 centeriod cluster moves to a point "a". This is an example of the assumed two clusters. Once determined patterns of behaviour. Points and clusters In the first step. 3.
Decision tree give values from which it is possible to conclude a probability. © Copyright Gabriel I. 100 6. Basis for computation is usually historical data. as well as other methods of data mining can be conceived as a proposal for later models build on it own. simplified and suitable for the human brain. Demonstrated by the example of Figure 29. Decision trees for simpler examples can be drawn on paper. What can be get is new knowledge about the behaviour of elements in the system.S. really a large amount of techniques powered with computers are trying to be presented as data mining.5 Decision trees Decision tree is not necessarily related to computer data mining techniques.4 Other methods Today. How do they affect knowledge discovery? Methods are improving mental model. Various programs help to decide on different ways to handle decision trees. or evaluate on ourselves in case result does not provide probability.business‐intelligence‐secrets. DWH/BI/DATA MINING FUNCTIONALITIES proving standards. Of course it is a computer help here. either by taking it from the computer. www. and only a tree is an excellent tool that helps to determine which way to go even if the decisions are similar and not easy to immediately see what is optimal. Basic concept consists of the initial questions which are then in detail split with sub questions into branches. Description of these methods require deep analysis and thats why according to authors opinion most important methods will be described in following chapters in order to show that there are many applications that support various forms of decision‐making and the various models of suggestions.5.com . as well as technology of visualization mainly oriented to the handling of large amounts of data. 6.5. Decision tree (Vanguard software) is case of deciding price strategy with the introduction of new products.
such as parts of the sensory receptors (for example motor output neurons responsible for movement) and hidden ones which are vast majority and are in fact cells in the brain. © Copyright Gabriel I.com .6 Neuron nets Neural network consists of large amounts of cells associated with a large number of connections.business‐intelligence‐secrets. the input cells are those that accept information. and between them are hidden cells.5. If analogy is made with human brain. which receive information to be processed. www. DWH/BI/DATA MINING FUNCTIONALITIES 101 Figure 29. the output returns from the information system network.S. Decision tree (Vanguard software) 6. Cells are divided into three main groups of cells: input units.
The computer repeated signal gives the same output. the connection weight values. DWH/BI/DATA MINING FUNCTIONALITIES 102 Input cells Ulazne celije Hidden cells Skrivene celije Figure 30. Even simple organisms learn to ignore repeated stimuli. and speed of thinking does not depend in any way on the number of connections (if ignoring signal propagation). where the contribution to the account depending on the value ponder connection that it contributes with activation multiplied by the value that carries the connection at that moment. Process in the human brain occurs in parallel.S. How to account for all cells in the same way. etc. in principle. which can be © Copyright Gabriel I. www. Computer must calculate weighted (pondered) values in each cell one by one. Sooner or later. They even calculate the activation value based on all given incoming values. the signal propagates to all the output cells. while the concept of the computer is completely reversed. This signal is transmitted to countless times through the levels of hidden cells. Described network is called a "feed forward" network and in order to make it more realistic model should include a large number of hidden cells. Activation algorithm is dependent on the strength of the connection. The speed of the brain in the analogous process is the slowest connection speed and the speed calculation is the number of cells x speed connection (calculating weight values). These cells send the activation value of the hidden cells with which it is associated. while in living organisms somewhat different exit. That number is usually calculated to adjust to being in the limits between 0 and 1. representing inhibition of signal coming from elsewhere. Activation functions. and basically summarize the contributions of all incoming connections. It is possible to restore the value of certain rules to 0 if the threshold value of touches. which of course in the case of a large increase in hidden cells results with slow work. filling them with output values. it is sequential.business‐intelligence‐secrets. Example of Neuron net Output cells Izlazne celije Each cell has a front entry value representing a input value to the network. the connection pondered values have in the process crucial role. It is possible to accept negative values. and propagating in the next step further.com . It is thought that in this way is possible to achieve a certain form of cognitive behaviour of the computer. are not too complex mathematical models.
etc. characteristics of the engine and look like the following figures: © Copyright Gabriel I. and not described with math as before mentioned method of clustering. Interest to work with neural networks show.com .S. although built with the idea of imitation of the nervous system. www. Good examples are the graphs as a way to display numeric values from a table where a lot faster distinguish jumps. How is human nature to want to penetrate into what will be. but also can be displayed for the attributes having something that is not part of the wider population standard knowledge. Can be concluded that. DWH/BI/DATA MINING FUNCTIONALITIES achieved by various levels of neural networks. among other things calls for help and computer. e. let's look at how can conclude something visual way? 6. 103 6. How is the visualization of method character best method to display images. „after the crystal ball“.7 Visualization Visualization is the process where conclusion is based upon the properties of a large set of data visually presented with the help of computer tools.1 DM – Visualization – Data collection Available are information about different attributes of the car. neural networks computers are still not able to replace the functioning of even primitive nerve system. dimensions. philosophers who believe that based on such models can be partly explained by the flow of cognitive processes in humans.g. This is a relatively familiar for most people. etc. beside IT experts and analytic users. price. Attributes of the models are typical attributes by which customers decide on the car. number of horsepower. information on the cultivation of vegetables.business‐intelligence‐secrets. First step in the process of DM is definitely data collection. anomalies.5. here will be shown a way of categorizing and knowledge discovery in a very simple example. Core idea is about the idea of mental models which are based on relatively unrelated attributes concludes somewhat unknown. So.5. possibly a future based on the past. Stated that the neural network shows the ability to learn is particularly important when considering and discovering new knowledge and the legality of the information with which every company has.7. Visualization is a very important process and gives very good results because human eye in a relatively well‐presented material quickly reveals the rules.
Price – revolution per minute for given horse power ‐ height Figure 32.com . Same as previous figure.business‐intelligence‐secrets.S. DWH/BI/DATA MINING FUNCTIONALITIES 104 Figure 31. www. rotated © Copyright Gabriel I.
Visually segmented graph As shown on the graphs can be recognised several important things. it is not the case with most others. In particular the highest car is shown to the right. after finding that it was a Hummer. Imagine to replace the role of the axis and the frequency of purchase become the type of trade. for people transport. It is relatively unusual car. Additional check. DWH/BI/DATA MINING FUNCTIONALITIES 105 Figure 33. and are not among the higher vehicles. Same as previous figure. Otherwise. Figures can be further analysed with the coloured regions on the graph and can be distinguished into three to four regions. These models have high price. among other things. Should take into account that the whole process is only visual and consists of the rotation graph of visual recognition and legality. and users may conclude not to consider Hummer as "cars".business‐intelligence‐secrets. Visually segmented graph which behaves according to rules that senior cars for maximum engine power demand less number of revolutions.com . marked with blue and orange. as a niskoturažni. At the top are different vans intended. Most of the cars are somewhere in the yellow box marked in the figure Figure 33. It is obvious that there are some cars that are in the red box set ‐ segment of low. it is common for the observed four Bentley models. Hummer in the example as can be noticed that eye easily isolates gatherings meaning something suspicious for analyst. On the graph marked with white dot is the Lamborghini Diabolo. Orange‐marked cars are distinguished by their height and the price.. it is Hummer. The blue marked set of data is Jeep. where some may conclude that highest cars are those with maximum power for a relatively low number of required rpm. Differ now red marked area where have scattered a few obvious specific models. This example was created with the aim to show a completely different analogy. and one axis can be projected as a set of goods sold in a particular trade. the amount to be paid in the store.. It is expected that the "cluster" or clusters are created and © Copyright Gabriel I. On the computer it is only adjustment of data visualization.S. which has already been mentioned. Can be said that the most expensive cars are actually low on the graph. rotated around and marked with red curve. Most of the cars to conclude from the Figure 32. What still catches the eye are two round recorded segment that also do not fall so to speak anywhere. www. From this we can conclude that the height and number of rpm are correlated.
based on the model of data warehouse.6 DWH and Decision supporting system What is very important to mention that the custom data model. www. etc. allows a simple application of commercially available tools for data mining and visualization tools that are used for data aggregation. Those points that are at the top of the list are obviously suspicious because they were farthest from the groups (mold) in set of normal transactions. human brain with more or less modestly prepared set of data is capable very quickly to give quite good results. In the first phase formed the final. DWH/BI/DATA MINING FUNCTIONALITIES 106 recognized visually more easily. fraud. as part of recognition is left to mathematical algorithm. Why is this interested? Because visual anomaly is more odd than anomaly presented with other methods.business‐intelligence‐secrets. Careful examination of similar cases in the bank it is possible to detect12 money laundering. number of clusters (clusters.com . © Copyright Gabriel I. Highest car in analysed set of cars In this example. the whole problem should be seen in two important aspects that describe the basic way to discover knowledge. 12 In fact it is a step further. if we look at them visually). What will most certainly be interested in such analysis is to recognise person who often (usually) purchase consumer goods in stores with low prices pays a very high amounts.S. 6. Each of these clusters calculate focus. seeking to draw attention to the fact that despite the relatively large and expensive programs and algorithms. to expect a small (3 ‐ 5 ‐ 15). indicating the process of research data using visual method. As the knowledge discovery is concerned. then calculate for all points of the total distance from the cluster center of gravity. Figure 34.
1 Effects of DWH system as IS subsystem DWH has been described so far in terms of Information System subsystems. That is something users only imagined or suspected. and more or less successfully on the basis of analogy with the existing model of creating idea of what is worth and what is not worth to invest in the broadest sense. Models are not implemented by users but by application itself (neural network). and display such data. phase of the product and all those associated with marketing) and product attributes.6. implementation model in practice. what is better sold and etc. Assumptions for the implementation of this subsystem in the work depends largely on the © Copyright Gabriel I. what was sold and what was not. Persons involved in the process have built a mental model. It just means that on the basis of learning from history. which is of course always management wish of each company. which simplifies the process of creating such mental model. Knowledge resulted from historical data DWH/BI/DATA MINING FUNCTIONALITIES 107 Knowledge resulted from historical data is kind of knowledge managers collect and it is more often called experience.com . This mental model is later applied to new products. where model if it is good gives information about something what previous knowledge did not had access to. based on the attributes of the market (of circumstances. Knowledge resulted from the model based on the DWH database and implemented through a BI solution Knowledge resulted from the model builds on previously mentioned experience. it is possible to automate some of the typical tasks that are performed by users. After the mental model is created it is transferred to the computer and the computer learns. Having experience in sales is to know how certain products had performed with their typical set of attributes in the market in some point of time of sale. 6. In the case of the model (implemented through the application program) in the state to perform its task in real time. learned knowledge is put into operation to improve work and thus reduce cost. www. saturation.S.business‐intelligence‐secrets. Another aspect of the model is testing the model with a completely new combinations of attributes. Finally its worth to mention that techniques described here are similar to techniques used to display data before the computer usage era. Where computer actually started helping is relatively easy development of models (especially in the data mining) where from large amounts of data models began to discover regularities. and explain improvement of model where the typical answers to the questions ‐ where is sale better. Such a model is still used in the prediction of behaviour of the market. What data warehouse in this case provides ‐ it is structure optimized for reporting and data aggregation.
IS with DWH subsystem and applications on DWH Significance of this system is primarily of strategic character. Despite it is a strategic decision making tool and development. data mining etc. This tool is a powerful to support manager for decision‐making. which has wide range of aggregated data from different subsystems. © Copyright Gabriel I. www. data warehouse and advanced tools that use data from data warehouses. Together they make standard Decision Support System. Should be distinguished structure of the database (data warehouse in the true sense of the word). FCBI will be analysed in detail in next chapter. BI and FCBI are marked here for better understanding relation with DSS. and to use data collected in the data warehouse structures. DWH/BI/DATA MINING FUNCTIONALITIES primary purpose of the system.S. BI FCBI Operative reporting Planning and what‐if modeling BI DSS and strategic decisioning Data mining 108 DWH Application1 Application3 Application n Application2 Application4 Company Information System Figure 35. it is in some way strategic advantage (or disadvantage) in company business. With regard to the strategic nature of the DWH subsystem. and tools designed for reporting. it is possible to implement a range of intelligent solutions that help in strategic decision‐making. In addition to the tools that serve for ad‐hoc reporting there are numerous tools that allow what‐if modelling. This implies that DWH system consists of database and is essential for feed of BI applications (call it the DSS system. based on data from the subsystems. Upon this structure. Using same data can at higher levels make completion with external information available on the market like data about competition and similar. Figure below shows a complete system. it is used primarily in the operational reporting. DWH structure is adapted for reporting. or MIS system.business‐intelligence‐secrets. which was also employed).com .
Characteristics of DSS Strategic The primary strategic tool that helps in decision making. During operative decision making process attention has been focused on the immediate future. IS role is supporting strategy. print invoices. Release production systems for their purpose ‐ data Operating entry and not for demanding reporting. although it should be a supporting role. card specification. This system should be differentiated from operational reporting. Decisions are based either on historical analysis of the observed information necessary to issue a decision. debt buyer. Automation of reporting is much easier to recognize as a direct benefit. Large systems are trying to solve non arrangement in the system (entropy) with reduction of information quantity about the non arrangement. Such system is adjusted to give a report on how well company keep within the course target of business strategy. either on the basis of more sophisticated tools that allow what ‐ if simulation. one of the tools of its implementation. for tomorrow. Automation of reporting. etc. entering new markets.) © Copyright Gabriel I.3 How can DWH / DSS systems be strategic tools? Main objective of strategic planning for the area of information technology is to connect systems with business strategy of company.S. Users can quickly recognise that for some reports that were prepared “manually” now produce for less time. which is intended for everyday work and used to perform primary functions of the company (print payment slips. DSS characteristics 6. Attention of tactical decision‐ making is focused on growth and on the period of the fiscal year focuses on efficiency. They cover period long enough to encompass the development of certain new products. Role of DWH and DSS in this scheme is primarily for informational purposes as feedback. Strategic decision focused on the prediction and the consequences of such potential and actual changes in the environment that could significantly affect the behaviour and activities of the organization. DWH/BI/DATA MINING FUNCTIONALITIES 109 Should not also be ignored fact of supporting character of DWH system in terms of automation of reporting. the next few days or weeks.6.com .2 DWH as source for strategic decision making Strategic decisions are based on longer‐term forecasting than tactical.6.business‐intelligence‐secrets. Hopefully Through a strategic decision company should gain a competitive advantage and thus a better chance of quality survival in the market Table 2. 6. www.
we only now that a buyer claims this is the first time that late. and so on. Information System is always late in relation to the business requirements.business‐intelligence‐secrets. Knowledge Management Systems. shaping. IS was composed of modules for billing and general ledger (bookkeeping module). What does it mean anyway? It is a fact that manager/analyst in analyzing data investigates causes of a situation (for example situation is why sale of units A is non‐profitable) wants to get as much as possible data about unwanted situation. reports." This is the most common cause that some series of business events are recorded collectively and not separately. Decision Support Systems Decision Support Systems (DSS) are set of Information Systems and processes that support decision‐making activities by accessing.S. Data © Copyright Gabriel I. we do not know exactly. DSS uses raw data. Example: Many companies grown during of 90s introduced IS. we have the old state debt to the buyer. As IS was growing spontaneously and not as a result of the company planned expansion IS developed after business requirement pressured much management. www. This fact leads to the impossibility of obtaining accurate information. because of the constant deficiency. DWH/BI/DATA MINING FUNCTIONALITIES 110 Good strategic plan is development of the base system as a whole. Frequent situation was that the financial module began to take care of booking various aspects of basic services by introducing more and more analytical recording. and thereby supporting role is reduced. “We cannot make this and this.com . preparing. often colloquially mashed up with the term ERP. DWH system will first point to deficiencies in the IS and the inability to provide adequate information. there is no data for the costs of xxx by business units. Digging into data is very common situation that leads to the answer "we do not know. External data. After the introduction of the DSS system deficiencies quickly (at first at all important analysis) arise to the surface.” DSS is not all mighty and book is full of notes what to be careful about. etc. documents. experience. presenting and delivering data important and relevant to make decisions. knowledge from Business Performance Management Systems. Excellent DSS is possible to make but there are many things that have to be considered. Data repository. BI. what is possible and what is not with each module.
wikipedia. Major benefits of DSS is in creating competitive advantage and generating evidence for decision making. 13 Source: http://en. DWH/BI/DATA MINING FUNCTIONALITIES 111 mining and similar information subsystems. These decisions have some of assumptions. 13 Management Information Systems A management information system (MIS) should be treated same as DSS but some authors consider it as a separate Information subsystem. and easily reachable. www. For the last.business‐intelligence‐secrets. folk‐wisdom and similar experiences. which are part of the decisions in company. In addition DSS also uses management feelings.org/wiki/Decision_support_system 14 Source: http://en. helps in discovery of new approaches. and procedures by management accountants to solve business issues like costs per single product or service. premises and they are based on information such as sales.com . In addition DSS automates managerial processes increases organizational control.org/wiki/Management_information_system © Copyright Gabriel I. According to them main functionality is to apply internal controlling mechanism on other information subsystems engaging people. etc.wikipedia. encourages analysis and exploration.S.14 In addition to these scenarios using the standard DSS provides a quantification of simple analytical assumptions. sales growth. documents. This information from DWH system must be available. market share of company sales in total sales market. costs of new strategy and similar.
) has a crucial importance. and therefore the data structure in DWH system (aggregation. This is never the case for complex IS consisting of hundreds of production systems.business‐intelligence‐secrets. etc. Choice of data storage technology that can also be crucial. many are not needed for reporting to management but are for operations and processes © Copyright Gabriel I.e. use of multidimensional methods of information storage. Easily achievable means everything is stored into DWH in proper way in practice.S. DWH/BI/DATA MINING FUNCTIONALITIES 112 Enlargement of procurement IS? Procurement Strategic plan DWH IS xxx Figure 36. Example of IS‐development projects emerging from the strategic plan Term easily achievable has strategic importance in the enterprise. It is “impossible” to store everything into DWH due to: • • • Too many tables in production systems (i. thousands) Too many data are exchanged between production systems Not all information are core information. because certain technologies allow mentioned benefits. www.com .
S. Ad‐hoc reporting is a term that refers to the way of making reports that are acceptable and simple for analyst and middle management in form of fast retrieval of information regarding some of the concepts in the business enterprise (sales. debt. DWH can be in case of limited resources in company main source and not the only source. data are relatively easily attainable. © Copyright Gabriel I.com . 6.business‐intelligence‐secrets. On contrary when saving data in multidimensional OLAP database retrieval is almost instantaneous. cost. time. This investment can be considered as poor investment. customer.) In choosing to save the data in relational model. For companies that do not store large number of transactional data it is enough to apply relational model.7 Example of DWH as standard Setting DWH as standard in company means to choose and use one source for reports. etc. so‐called star model (star schema). Expected increase in business (access to other markets as a result of a strategic orientation towards regional expansion of the company) can justify the investment in licenses of multidimensional databases. plan and other processes for management. but getting data in the institutions that collect large volumes of transactions in their business can be painfully long and therefore really unproductive. choice will be wrong and chosen technology will quickly be needed to be replaced. DWH can never be up to date and store everything for easily achievable information in complex environment. DWH/BI/DATA MINING FUNCTIONALITIES 113 To conclude. www. and the business / strategic unit. In case of buying what is currently not needed and not in alignment with corporate strategy. item.
Example of DWH as a standard in company Rules: DWH is unique source system for all data projects from finance and business area. Changes on definitions and that can have consequences on historic data must be realized according to predefined process in order not to lower credibility of historic data. www. direct extractions/queries. Users reach data from DWH through OLAP cubes.S. © Copyright Gabriel I. standard and ad‐hoc reports for portal.com . Process of consolidation is realized through Business Dictionary. DWH/BI/DATA MINING FUNCTIONALITIES 114 Business (BI) Finance (FCBI) Consolidation Analysis DM modules / reports BI modules / reports ERP cross modular reports Profitabi lity ABC MDM Planning Reporting DWH ERP CRM Call center Cash system Billing … Figure 37. Business and data definitions must be consolidated and verified. Power users from finance and business make OLAP cubes.business‐intelligence‐secrets.
com . Companies estimate 'one‐stop‐shop' is best for them and in matter of price. Trying to make substitution in DWH/BI/Data mining is always temporary. Gartner predicts many current companies with DWH/BI solution will have limited acceptability or will be total failure as a direct result of data quality problem neglecting. On the other hand project is expensive and without solid hard support on top level project will stay sooner or later without resources. Typical DWH implementation lasts at least 1‐2 years and engages 3‐10 experts.S. © Copyright Gabriel I. Engagement of only own resources without external experts and consultants since they know best own processes can lead to pitfall. BI is first of all intended for decision makers and their experts.business‐intelligence‐secrets. External IT experts are not loaded with company problems and have much broader knowledge and experience especially if they are specialised for DWH and what is very important they know what problems will appear in future and how to prepare for them now. If company had this people then those people were not valuable for existing production systems and were just a bourdon. And off course those people have to be „no border“ from everyday activities. CRM. Other approach will most certainly destroy DWH project. „Our vendor of corporate applications will provide best solution“. Top management needs to be interested for DWH building. Make substitutions of certain functionalities in DWH/BI instead of buying finished solution. This is very hard and almost impossible for many companies. from top management levels to lowest administrator levels. DWH/BI/DATA MINING FUNCTIONALITIES 115 6.8 Failure‐success factors of DWH/BI projects Research show low implementation rate of DWH systems and many existing DWH projects going out of date (about 40%). Solution from vendor has to be compared with solution of best company in this industry. Master data management or similar. Resistance from inside company. limited solution and therefore will collapse after each change. There is always a great threat that DWH/BI tries to resolve certain problem that is already much better solved in certain system but not implemented in company like product management. Company does not have enough own people that can be 100% dedicated to project. Although this solutions have to be installed before DWH/BI implementation. Users that will mostly use DWH have to be open to new technology. Result is business users percept it of little or no value solution. „If we build it users will come“. www. DWH/BI build from IT side has little or no business connectivity with real world. „We do not have data quality problem“. Here are some reasons: No sponsor secured for project on top level management. If they don't want it project will collapse.
S. Out of the box solution are expensive sometimes more then building DWH from scratch. Important is to have solutions that meets it’s important expectations. not a solution no one wants to use. but BI projects have to start and to be finished. linking with documents. prediction models.9 Basic functionalities Doesn’t really matter does company has BI or even data mining solution. Many books and articles are already written about it.com . alarm and threshold system. Described functionalities grouped by modules serve as idea catalogue for output KPIs and are data feed for FCBI and TCC. aggregation/grouping. In this chapter authors try to reveal best of BI/DWH/Data mining functionalities and are not going into analysis of technical architecture. so more or less expenses can be compared if not bigger. graph manipulation. Some of following functionalities can be bought as independent solutions and therefore are much better to be implemented in independent applications then in DWH. Authors approach is to describe functionalities from user side. DWH/BI/DATA MINING FUNCTIONALITIES 116 „BI projects have to evolve“. and never underestimate resistance to change in mankind. Outsource only non‐core DWH/BI business not everything. www. In addition is a list of requirements toward DWH that can give quite beneficial results. e‐mail notification. diagrams. BI has to evolve. „Can be outsourced“. easy managing and analysis. geographical data mapping actual/forecast/variance comparison. Also standard DWH/BI is mostly business world and far less financial world in mean of processing data – platform used for reporting upon transactions in productive systems and for analytics. is temporary solution better resolving functionalities in DWH and BI or permanent solution – buying module. Wrong. Some of functionalities that DWH/BI/Data mining might be used for: central repository place. benchmarking. theory and similar. and that is most of the job in creating BI solution.business‐intelligence‐secrets. data has to be transformed in such a way that it fits out of the box solution. © Copyright Gabriel I. visualization of trends. Decision makers should count with following dilemma. 6.
2 Campaign management and monitoring It’s always great to include campaign management and monitoring processes into DWH. different sales channels customer inquiries. invoices 16 mails. calls. www.com . customer types. segments. different campaign response separation. fax. faults and complaints in any type of form16 Newsletter do not contact option etc. Campaign management supports following functionalities: automation of all retention programs resolves conflicts predicts marketing costs for budgeting purpose keeps reasonable RoI for marketing campaigns 15 Calls. customer behavior. For analytical purposes named processes consume lot of resources from production systems and therefore it’s much better to transport them into DWH or similar data repository system for reporting.9. direct mail.9.business‐intelligence‐secrets.S. combination with campaign costs data and similar. © Copyright Gabriel I. personally and etc. Good to be included are past contacts.1 Contact management Contact management should be integrated into DWH by giving Business units analytic information ‐ figures base for future actions. Although Contact Management should be supported with information at customer level multidimensional view over: inbound/outbound contacts15.. 6. campaign response history. DWH/BI/DATA MINING FUNCTIONALITIES 117 6.
business‐intelligence‐secrets.S. scoring list and similar. Functionalities supported in DWH: Standardised data model. transformations. derived variables and data preparation (ETL) procedures for churn analysis Predefined reports for churn analysis Advanced analysis functionalities like retention offer estimation. finding changes in product portfolio and tracking trends (what happened last month. This should result in identifying patterns of behaviours that can predict a change in profitability or possible churn of customer to competition. Example of one Campaign Process Flow Chart. This functionality should enable looking for changes in profitability over time. finding anomalies. year. © Copyright Gabriel I.3 Customer behaviour recording and predicting One of very useful feature of DWH and Data Mining can be customer behaviour recording and prediction.com . customer responses and similar. comparing forecasts and actuals). comparison. 6. churn prediction. www.9. Following elements are part of customer behaviour and influence churn: complaints. DWH/BI/DATA MINING FUNCTIONALITIES 118 Audience Treatement Update Update Output processing Automated Fulfillment campaign activities Response analysis Call center printing Load response data Capture response data Contact made Figure 38.
) Behavioural segmentation support Segmentation verification mechanism and reporting © Copyright Gabriel I. DWH/BI/DM is supporting cross‐sell and up‐sell activities in sales and marketing and allow fine‐tuning of the way how products are packaged and priced to fit customer requirements per customer segments. cross sell‐ up sell. Results should be making fine tuned and profitable campaign activities. Central analysis of customer data and making parallel hierarchy grouping. Results and reports analytically supported in DWH/BI/DM based on different criteria like customer behaviour. revenue assurance and etc. Usage in advanced analytical functionalities (like: payment risk. customer demographics and other variables are reference for determining which customers are potential for specific product types and offer. purchase history. DWH/BI/DATA MINING FUNCTIONALITIES 119 Implemented quality monitoring mechanism 6. quarter.9. www.4 Lifetime value of customer analysis Refers only to highly complicated industries like telecommunications and banking. Time remaining in the customer's life cycle Retention costs Profit from average customer Understand the combinations of behaviour.5 Customer segmentation/customer clusters Customer segmentation is essential part of CRM solutions – production systems but propagation of its data is of crucial importance in ordinary work of DWH/BI/DM which should afterwards provide base to analyse data per customer segmentation. different customer segmentation. DWH/BI/DM should provide customer point of view.S. DWH should support with supply of basic information mostly for FCBI solution: Customer acqisition cost Average amount each customer spends per period (month.business‐intelligence‐secrets. 6. year) Products and services purchased Average time with and characteristics of customer life cycle.9. grouping and grouping based on current needs. cost and revenue that are profitability drivers. churn ratio.com . Compatible usage with other stored data in data repository.
9. • Order lifecycle tracking.9.9. DWH/BI/DATA MINING FUNCTIONALITIES 120 6.business‐intelligence‐secrets.8 Other functionalities DWH/BI/DM should be also capable to give answers directly or by support FCBI with data in following segments: • Pricing Analysis and optimization. high‐value accounts and subscribers. 6. drill down and profile accounts at multiple levels—across various attributes—to determine such factors as: • Total revenue at risk and total • High‐risk accounts and subscribers • Low‐risk. • Accounts in various age bands (30 days. 60 days.S. 6. • Demographic or organizational profile of the most delinquent customers. • Analysis of re‐pricing impact (WHAT‐IF analysis and simulation) • Parallel Multi‐level granulation – using predefined groupings from production systems and possibility to make own. 90 days past due).6 Payment risk DWH/BI/DM has to support complete analysis or in limited range. parallel hierarchies. measuring the impact of pricing on buying behavior. • Account types that contribute to the most debt and how they were acquired. www.7 Cross‐sell and up‐sell Finding links between customers in order to increase sales and identify customers that are more likely to adopt new products or services by concentrating on the customer products formed by using direct interactions (communications) between consumers. • Proportion of default cases by payment method. • Market share estimation for new price plans.com . tracked orders trough different systems in end‐to‐end process. © Copyright Gabriel I. • Patterns and trends in usage and credit ratings across various segments.
DWH/BI/DATA MINING FUNCTIONALITIES 121 6. generates permanent problems and is not functional. system costs. After sometime Board could say. Danger appears form persistent negative perception and rumours for majority of employees influencing Board. “DWH is guilty why we cannot get required figures!” Its more than 99% possibility that production systems in common data exchange do not generate required data and DWH cannot produce data that is not stored in production systems.com . Whatever it means… Practically this attitude of Board could mean silent death of DWH/BI/DM. In eyes or customers that do not work directly with DWH and do not understand completely its role many myths will appear and unfortunately they can be dangerous and funny. They simply cannot support numerous functionalities. „I cannot believe that you super advanced and sophisticated new system do not support simple operation as data mapping”. Same is for functionalities from IS are not applicable on table calculators. But perception is that DWH delivers bad data. Functionalities that are so simple in Excel are not so easy to implement into BI tools or other IS systems.S. do want to make damage to others and similar reasons.10 Myths and legends after implementing DWH/BI Myths will appear as soon as DWH/BI/DM are implemented or are partly implemented. like to make jokes. This is not comparable since Excel is IS black hole – it is not a database. Further it’s very likely that production systems have trash data and need to be cleaned. They do not perceive that hole in ship can sink complete ship. www. Reasons why rumours are made could be mostly because people don’t understand it. like to make bad buzz in company. Actually they will appear together with production phase. If process‐reporting mistakes on level of production are not fixed then DWH cannot completely fix problems. Well obviously here customers do try to transfer functionalities from table calculators to DWH/BI/DM. It is not ok to compare them… © Copyright Gabriel I. Can make limited temporary solutions like code patches but this is purely temporary and to be advised as good.business‐intelligence‐secrets.
business‐intelligence‐secrets.2 but worth once again to say that people tend to talk about only very general descriptions of KPIs and compare not comparable. DWH/BI/DATA MINING FUNCTIONALITIES “DWH is guilty for KPI definitions. It is hard to fight with this rumours without help of board member. www. It's a myth. © Copyright Gabriel I.S. Already described in Chapter 5.2 and 5. utopistic expectation since it’s only a system with many limitations. they have everything” It is utopia. press button and get everything. Same myths can be applied for FCBI as specialised BI solutions.1. It is a problem of wrong decisions terminology and very rarely of system. Once again “DWH is guilty for everything. sponsor and to make positive internal marketing.” DWH is regular guilty system for process problems in reporting and for process problems in production systems. 122 There is no absolute automation. “DWH/BI/DM have needed data. we give once again several figures for same KPI”.com .
In this case DWH should serve only as repository. Only simple extractions should be processed in DWH/BI. Therefore it is reasonable to transfer demanding data extractions out of ERP. Almost immediate. Best examples are SAP modules. ERP is slower and its primary design is not for reporting but to gather and combine data from many input points.com . www. Catch is where to process ERP data. ERP has powerful reporting but in many cases reports with queries for huge data quantities extract sometimes in hours. cardiovascular disease or other disease and slowly remove it out. otherwise reporting will face collapse. Some have very quick ad‐hoc query support for larger number of users... where to prepare data for report… In production system or in DWH… It is wrong to process complex17 ERP extractions in DWH/BI. Don’t forget to implement „old diseases ‐ old reporting deviations“ back into reporting logic. Many relations between data in extractions means data from many different tables. core business applications. 17 Complexity in sense of restoring huge numbers of relations between exported data. DWH/BI/DATA MINING FUNCTIONALITIES 123 Instead of Summary – Important notes Never.S. © Copyright Gabriel I. Implement once again asthma. like ERP. “Complex” report may relay upon ten of thousands tables. Complex ERP extractions should be processed in FCBI or ABC module. Some DWH solutions don’t have such speed.business‐intelligence‐secrets. Many managers have related own goals with data from production systems and any sudden significant changes are not welcome. never… Implement DWH/BI in parallel with big changes or upgrades of production systems. Not all DWH solutions are same. Catch is not where to transfer it… best is to DWH (sometimes directly to FCBI depending on process requirements and specifics).
business‐intelligence‐secrets. It’s a temporary solution.com . With growth of requirements temporary solution will collapse and it is upon decision makers to weight such action. www. Change in production system requires additional effort to maintain code in DWH/BI. Do not restore data logic – process logic in DWH/BI unless there is no other option. DWH/BI/DATA MINING FUNCTIONALITIES 124 DWH/BI solution can serve for processing of complex extractions to certain level.S. BI/DWH/Data mining solution is mostly supply of non financial data for FCBI and TCC with aggregated numbers – KPI’s. not for creating data. This is permanent solution. © Copyright Gabriel I. Best is to force data consolidation and reconciliation on production system level. Action results in significant DWH/BI customization supported with lot of expert/finance resources. DWH/BI system is for reading and delivering data.
com/business-intelligence-pdf © Copyright Gabriel I.S. www. DWH/BI/DATA MINING FUNCTIONALITIES 125 To read more about prerequisites for DWH. Business Intelligence and Data Mining download our FREE BOOKS in PDF at http://www.business-intelligence-secrets.business‐intelligence‐secrets.com .
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.