The Architecture of Business Intelligence

Business Intelligence Architecture

The business intelligence (BI) architecture is a set of systems, applications, and governance processes enabling sophisticated analytics by facilitating the flow of data, content, and analyses to the appropriate users when needed.

Key Elements of BI Architecture:

1. Data management: How data is acquired and managed.
2. Transformation tools and processes: How data is extracted, cleaned, transmitted, and loaded.
3. Repositories: Organization and storage of data and metadata.
4. Applications and analytical tools: Software used for analysis.
5. Presentation tools: How information workers access, display, and manipulate data.
6. Operational processes: Addressing security, error handling, auditability, archiving, and privacy.

Data Management

A well-designed data management strategy ensures the organization has the right and accurate information and uses it appropriately. Key considerations include:

• Data relevance: Identifying data needed for analytical competition.
• Data sourcing: Determining where to obtain the data.
• Data quantity: Deciding how much data is needed.
• Data quality: Ensuring data accuracy and value.
• Data governance: Defining rules and processes for managing data throughout its lifecycle.

Characteristics that increase data value:

• Correctness
• Completeness
• Currency
• Consistency
• Context
• Control

Data management life cycle stages:

• Data acquisition
• Data cleansing ($25-30$ of BI initiative costs)
• Data organization and storage
• Data maintenance ($500,000$ ongoing maintenance per $1$ million spent on new capabilities)

Transformation Tools and Processes (ETL)

Data undergoes ETL (extract, transform, load) to become usable. Cleaning and transforming data is a significant part.

• Cleaning and validating data using business rules.
• Standardizing data definitions.
• Deciding how to handle missing data.

Repositories

Options for organizing and storing analytical data:

• Data warehouses: Integrated data from different sources, regularly updated, containing historical data.
• Data marts: Support a single business function, containing predetermined analyses.
• Metadata repository: Contains technical information, data definitions, and information about data reliability and accuracy.