- Today Business Intelligence (BI) is proving to be an important tool using which organizations are able to provide the right people, the right information at the right time, leading to
- Increase in operational efficiency
- Eliminate costly, delayed report generation
- Enable end user to analyze and create reports from available information without assistance from Information technology teams
- Analyze root cause and take corrective actions
- Maximized revenue
- Improve Marketing Analysis
- Targeted marketing campaigns
- Real time fact based decision making
- Improved customer experience
- 360 degree customer analysis
- Personalized marketing
- Faster response
- Compliance achievement
- Assess risk position
- Maximize available funding
- Provide social and public information
- A separate BI system is recommended as unlike usual business applications, an individual BI application encompasses continuous enhancement and improvement based on feedback received from the business community.
- In contrast to business applications, a BI system can be built on cross organizational / functional activities to sustain an enterprise wide decision support environment. With minimal dependency on business application, a BI application would exhibit enhanced performance and efficiency.
- Customer satisfaction
- Problem resolution
- Up-sell & Cross-sell
- Customer Churn
- Product / Customer segmentation
- Cost control
- Revenue & Profitability
- Performance management
- Defect Minimization and Tracking
- Market Analysis and effectiveness
- The need of a BI application arises from the operating departments like Marketing, Sales and Finance, wherein analysts need the actionable insights of data in timely fashion. Such requirements are usually not generated by IT departments.
- BI applications are usually, needed by the top management and strategic planners of the organization and drills down in the organizational hierarchy.
- Top management, Strategic planners and Key business users
- Line workers
- Customers / Suppliers
- It is a known misconception that BI projects belong to the IT department of an organization, whereas any successful BI / Data warehousing project is driven by both - IT as well as functional departments.
- A few Critical Success Factors in any BI project:
- Involvement of business and BI experts, with the right team structure
- Well defined business cases to get an apt solution to the business problem
- Sufficient funding and Return on Investment (RoI) plan
- Bottom up approach - to start by building the data mart first
- Continual and adequate involvement of functional experts
- Definition of correct and comprehensive KPIs
- Thorough and efficient project management
- Adequate metadata design
- Availability of latest technology and tools
- Thorough end user training
BI is much more than just management reporting. It provides real time, rapid and easy access of the actionable insights of business conditions - about customers, products, finance and market. Let us take one scenario:
- Product acquisition cost
- Return on Investment (RoI)
- Plan for user growth
- Agile BI support
- Ability to connect to any data source / database including NoSQL, Big Data
- Ease of integration with third party applications (Embedded Analytics)
- Data blending support
- Single solution for entire spectrum of BI requirements like Adhoc Reporting / Analysis / Dashboard, ETL, Predictive Analytics, Support for SSO / LDAP etc.
Multi-tenant BI refers to a single BI instance that serves multiple organizations / clients. Multi-tenant BI solutions provide the file / folder / row / theme level secure multi-tenant environment.
Self Service BI facilitates the end user / analyst to design and deploy their own analytical reports and dashboards without taking any help from IT department, within a multi-tenant BI platform.
Metadata is data about data e.g. if we receive any file in a data mart, the metadata will contain information like number of columns, file type (fix width / limited), ordering of fields etc.
Cloud BI refers to BI solutions deployed in the cloud. It is considered ideal for SaaS (Software as a Service) based BI solutions.
- In theory, Predictive Analytics and Data Mining both use Mathematics to get the desired results.
- Data Mining is an analytical process designed to explore data to identify useful patterns and relationships between attributes.
- Predictive Analytics is an analytical process that uses useful patterns identified in Data Mining to forecast or predict the future. In essence, Data Mining is a part of the Predictive Analytics.
- A Predictive Model is a function that takes the input variables, applies a formula and / or rule to predict an outcome.
- Types of the Predictive Models
- Classifiers: Classifiers construct models for each class based on historical data. It can be used to predict the category for given data set for future events.
- Recommenders: This model can find similarity between pairs of users by using similarity formula and is useful to predict what the user wants to buy based on buying patterns.
- Clusters: This model aims to identify the similar data objects embedded in a complete data set. It helps to combine the predictions for similar data objects.
- Time Series: This model is meant to analyze large historical time frame data and find similar sequences which help to identify trends and behavior over a certain period.
- Analytical database is a MPP (Massively Parallel Processing) based, column oriented database specifically designed for large scale data warehouses.
- Analytical database can be deployed in multiple machines in a distributed, MPP-enabled, high availability environment with unlimited scalability.
- E.g.: HP Vertica, Amazon Redshift, GreenPlum
- Dimensional Modeling consists of fact and dimension tables.
- Fact Table is a primary table in Dimensional Modeling where quantifiable measurements of business are stored.
- Dimensional tables are an integral companion to a fact table. The dimension tables contain textual descriptions of the business. Dimension describes the “who, what, where, when, how and why” associated with the measures.
- Star schema: The star schema consists of fact tables referencing any number of dimensions.
- Snowflake schema: It closely resembles the star schema but here, dimensions are normalized into multiple related tables.
ETL stands for Extraction, Transformation and Loading. ETL refers to methods involved in accessing and manipulating a variety of data sources and loading into target databases.
- Uniform and less coding owing to graphical user interface
- In built scheduler
- Availability of advanced and continual features
In Data Analysis and Predictions, human eyes can easily detect abnormal and out of range patterns and deviations. BI is complimented with advanced visualization tools like Bubble, HeatMap, Tree, Scatter etc. These tools provide a drill down facility for advanced visualization and easily integrate into BI solution. Data Visualization presents data in a visually appealing form and assists business users to take quick decision.
An OLAP cube consists of numeric facts called measures which are categorized by dimensions. The cube metadata (structure) may be created from a star schema or snowflake schema of tables in a relational database.
- ROLAP is Relational OLAP. In ROLAP the data resides in a relational database. This model allows the multi-dimensional analysis of data and user can perform slicing and dicing of data.
- MOLAP is Multi dimensional OLAP. In MOLAP data is pre-summarized and stored in an optimized format in multi-dimensional cubes instead of relational database. In this model, data gets stored in proprietary formats based on client’s requirements, with calculations pre-generated on cubes.
MDX is a language that allows you to query OLAP cubes in the same way as SQL allows you to query relational database. In addition, MDX expression can be used to add business logic to cubes.