Docstoc

Table 1. DWH vs OLTP

Document Sample
Table 1. DWH vs OLTP Powered By Docstoc
					 Multidimensional Data Model for Marketing Information System
                                Zlatinka Svetoslavova Kovacheva
                     Centre for Information Technologies in Communications of
                          Bulgarian Telecommunications Company (BTC)

                                          ABSTRACT
       The present talk deals with the basic moments in the process of design and development of the
Marketing Information System (MkIS) of BTC. The MkIS analyses the evolution of marketing
indicators such as capacity, usage, revenue of the services, etc. on the base of monthly information
from BTC regions.
   The Multidimensional data model design is considered. The program environment for developing
the model is based on the Data Warehouse technology and includes OLAP (On-Line Analytical
Processing) tools for structuring and analysis of the data into multidimensional arrays. This model
provides representation of the information in a lot of interconnected tables and graphs, which can be
viewed in different aspects according to the defined dimensions and their hierarchical levels. Data can
be easily aggregated, disaggregated and rotated according to the requirements of the experts and
managers. A generation of ad-hoc reports is available. It provides the users a fast direct access to that
part of the comprehensive data structure, which is useful for their concrete purposes.
   One of the most interesting features of the multidimensional data model is what-if-analysis. It
provides managers creating hypothetical situations by changing the values of variables in the
multidimensional data model. These changes are temporary and concern all formulas including
corresponding variables. This is the way to observe the influence of changing some parameters to
other ones. It is particularly important for the marketing decision making.
   Another advantage of the model is forecasting facility. The following basic forecasting methods
are available: linear trend, exponential trend, single, double and triple exponential smoothing,
percentage change, moving average, Holt-Winters.
   The Multidimensional data model provides a powerful tool for the decision makers in the field of
marketing and other activities concerning the firm management.
 MARKETING INFORMATION SYSTEM

  The key areas of competitiveness in today’s market place are:
   Market awareness;
   Speed of response;
   Adaptability;
   Innovation;
   Efficiency.


  Marketing information system (MkIS) is an ongoing, organized
set of procedures and methods for creation, storage, retrieval,
dissemination and analysis of information for marketing decision
support.
  DATA WAREHOUSE (DWH)
   The data warehouse (DWH) is a process supported by products,
services and partners, that collects, integrates, stores and delivers data to
the organization (From a report produced by IDC: A Study of the Financial
Impact of Data Warehouses (1996)).

  DWH is an enterprise structured repository of subject oriented, integrated,
non volatile, time variant data.

  Types of Warehouse Data:

     Fact data – Measures of the business (detail data);
     Dimension data – Query drivers (an attribute by wich data may be analyzed);
     Reference data – Text look up (contains relatively small volume of data);
     Summary data – Precalculated data;
     Metadata – Warehouse “map”.
                    DWH vs OLTP:

     Property             Operational               DWH
User activities     Operations             Analysis, forecasting,
                                           etc.
Response Time       Sub. sec. to seconds   Sec. to hours
Access              Read and write         Primarily read only
Nature of data      Current data (30-60    Historical data
(time period)       days)                  (snapshots over time)

Data sources        Internal               Internal and external



Database Size       Small to large (<100   Large to very large (50
                    GB)                    GB to 2 TB)

Types of Decision   Production             Strategic decisions
Making              management
              DWH vs. DATA MARTS

                                          DM1      Marketing


  Legacy data
                                          DM2
                     DWH
                                …..
 Operational data                                  Finance
                                          DMk




                                          DMn
  External data                                    Personnel
  sources


DWH scope – enterprise          DM scope – department single
            multiple subjects               subject
DWH size – 100 GB               DM size – up to 100 GB
           to more than 1 TB
  EXPRESS SERVER
  Express Server        is a multidimensional engine for online
analytical processing (OLAP) with the following features:

   Multidimensional analysis;
   Measures with different dimensionality;
   SQL support;
   Robust development environment;
   Open API;
   Distributed;
   Scalable.
                         EXPRESS SERVER
                          APPLICATION




Product Manager view                      Regional Manager view

                              Multi-
                            dimensional
                             Data Base




Financial Manager View                    Ad-hoc view
EXPRESS SERVER APPLICATIONS

   Applications:

    Performing in-depth competitive analyses;
    Tracking new product introductions and promotional
     response rates;
    Conducting pricing, distribution, and promotion
     comparisons across regions;
    Analyzing income and expense ;
    Tracking manufacturing inventory.
EXPRESS SERVER
OBJECTS

 Dimensions
 Relations
 Variables
 Formulas
 Programs
 Composites
 Valuesets
 Worksheets
MOST TYPICAL DIMENSIONS :
 dimension time periods: years – quarters – months;
 dimension geographical regions – regions in the
   country
 dimension countries, etc.


FOR THE MARKETING PURPOSES:
 dimension products or services
 dimension clients or types of clients
 dimension distributors, etc.
  VARIABLES
  FOR THE MARKETING PURPOSES:

 variable products or capacity – contains quantity
   characteristics of products or services;
 variable sales or usage – characterizes the realization
   of the products or services;
 variable costs – describes the expended resources;
 variable revenue – describes the financial results of
   the firm activities;
MARKETING INFORMATION SYSTEM of BTC


             Marketing manager




   Queries      MkIS             Decision making
                                   information




                 DWH
THE MAIN FUNCTIONS OF THE MKIS SYSTEM


 user friendly reports and ad hoc studies generation;
 historical and up to date data integration for the
   purposes of tendency analysis and forecasting;
 real data mathematical models representation;
 what - if analysis.
THE MAIN VARIABLES IN THE MkIS

periods                       periods
     regions                       regions
           exchanges                        exchanges
               subscribers                    subscribers
                lines                         changes


    capacity                   changed lines

periods                      periods
     regions                      regions

          exchanges                    exchanges

               subscribers                  subscribers
               services                      services


     usage                      revenue
FORECASTING METHODS:
 LINEAR TREND – models the data as a straight line;
 EXPONENTIAL TREND – models the data as an exponential
   curve;
 SINGLE, DOUBLE AND TRIPLE EXPONENTIAL SMOOTHING –
   a system of weighted averages which effectively smoothes the
   data;
 PERCENTAGE CHANGE – applies a variable’s observed period-
   to-period percentage changes directly to the user-defined set of
   forecast time periods;
 MOVING AVERAGE – calculates a moving average of a set of
   data;
 HOLT-WINTERS – decomposes data into three related
   components: a “smoothed” series, a seasonal series, and a trend
   series.
      CHOOSING A FORECASTING METHOD

     Method           Time Horizon       Data Pattern        Minimum number
                                                              of observations
Single              Immediate, short   Stationary           2
Exponentioal
Smoothing
Double              Immediate, short   Linear               3
Exponentioal
Smoothing
Tripple             Immediate, short   Non-linear           4
Exponentioal
Smoothing
Moving Average      Immediate, short   Stationary           3
Holt-Winters        Short to medium    Seasonal             2 seasons
Linear Trend        Medium, long       Linear               3
Exponential trend   Medium, long       Non-linear           3
Percentage change   Medium, long       Stationary, linear   2
  USER ACCESS TO MARKETING DATA BASE

Windows Client                         Oracle Express
 Applications     User id              Server Instance
                            SNAPI
(Oracle Express              Calls       Cached
   Objects)                             data cubes




                                     Marketing data base
                                           Stored
                                         procedures
       REPORTS LIBRARY


               LIBRARY
  MAIN LIBRARY        USERS LIBRARY



 LOAD REPORT      LOAD REPORT
                            SAVE REPORT




35 FREQUENTLY USED REPORTS

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:3/28/2012
language:
pages:18