Oracle Information Management Reference Architecture – Examples from Oracles pre-built Analytical Products

The set of Oracle Analytical Products has become quite large in the recent time. Beginning of 2015 the list (non-exhaustive) would probably be comprised of:

  • Big Data Appliance
  • Exadata
  • Exalytics
  • TimesTen
  • Endeca
  • Oracle BI
  • Data Mining
  • Oracle R
  • Essbase
  • In Database Analytical Functions
  • Hyperion Planning
  • Essbase
  • Crystal Ball
  • RTD
  • BI Publisher
  • SmartView Office Integration

Thus, there comes not only the need to integrate but also to organize all these Products. The Integration of all these Products (on the Technical Level) does not come Out-of-the-box and can sometimes take a certain effort E.g. Importing Essbase Hierarchies in Oracle BI to allow Users to use these Entities for Analysis. Understanding the Integration and Organization of these Products, together with the Applications which contain the Business Process Execution ), Oracle does provide the Oracle Information Management Reference Architecture.

On the Left Side are all the Data Sources which could be Unstructured Data in Text Files, Master Data Systems for Customers or Products, External Data e.g. market Trend or Benchmarking data, Standard or self-written Applications. The Data is extracted from these Systems and loaded to the Staging Layer. This is a Temporary load of the Data for the purpose of taking the Data into the Foundation Layer. After the Data Loading job to the Foundation Layer has been completed e.g. for Order Data for a certain Month the data will be removed from the Staging Area.

The Data within the Foundation Layer should be normalized since it is taken over from the Operational Sources which are mostly 3NF in nature and this will also use less disk space compared to a normalized data.

The Access and Performance Layer is the Area where most data should be accessed, given that within this Area the Data will be Optimized for Analytical Processing e.g. by using Star Schema setup or any other DataWarehouse Query Optimization Technique. The pre-built Horizontal BI Applications also belong to this Area since the also use Star Schemas.

Data from any of the other Layers like the Source Systems, (Staging), and Foundation can also be access by the Analytical Tools which omits the need to Transform or tune all data within the Access and Performance Layer. The BI Server itself functions as an additional abstraction from the Physical storage of the data within the Warehouse. The Users access and gain the Insight by either accessing the BI Server (Abstraction Layer) directly by creating Reports or indirectly by receiving Alerts e.g. if certain Thresholds are met or by a weekly Business Report by mail. Data Scientists or Statisticians create their work within the Analysis Sandpit to Test their Hypothesis.

While typical Reporting data is understood by most Business Users. E.g. Revenue for a Region or a Store, conveying Advanced Analysis produced by Data Scientist or Statisticians will be more difficult. Typically a Data Scientist or Statisticians will use dedicated Tools (Desktop Client Software) which can only be operated by Trained Employees.

Thus, a Question that naturally arises is: How can the Business Insight gained within the Analysis Sandpit by Data Scientist or Statisticians be distributed to the General Business User.

Oracles own Pre-Built Analytical Products give some great Examples on how simple “System of Records” Reports and more advanced Statistical Reports can be used to provide Business Users with the required Insight.

The Oracle Communication Data model Uses the Support-Vector-Machine Algorithm to determine a correlation between certain Attributes of a Customer and their decision to Cancel (churn) their Contract or Subscription.

The above Report lists the most Important attributes (Ranked) that determine Customers how cancelled their Subscription. The Contract Left Days Attribute has a very high Importance (or Correlation) on Customers Leaving. This Information can be used to drive targeted Marketing Campaigns to offer a better retention Offer to the Customer. The Below report (again from the Communications Data Model) mixes the Predicted Churn Probability with the Contract Information for the Revenue Band:

The Sample App Image also contains some advanced (Data Mining based) Analysis mixed with normal Attributes to bring the Insight to the Line of Business:

The above Report Predicts the Life Time Value of the Customers and gives the # of Customers within the Predicted LTV Band. At last this can also be combined with the Mapviewer, allowing Geospatial Analysis. The Below Report from the Sample App shows the LTV of Customers plotted on the Map:

This can be done because Oracle Products store all this Model Information within the Oracle Database. Thus, one can use Oracle BI Metadata to build a RPD Model on Top to provide a convenient access to the Information via the web Browser. The Below Example is from the Oracle RTD Integration with Oracle BI EE:

For further readings on the Oracle Information Management Architecture itself (rather the Examples), please find the following links:

Trying to understand the Oracle Reference Architecture for Information Management

Information Management and Big Data (From Oracle)

Evolution of Information Management Architecture and Development (Co-published from Oracle & Rittmann Mead)