Archive for the ‘Architecture’ Tag

Why MDM is improving BI?

MDM_BI Master data has long been managed through functionalities embedded in operational business applications, which is not a bad strategy if a complete set of enterprise business processes are supported on the same ERP platform.  In reality, we know today that very few enterprises succeed in decommissioning legacy systems and keeping only the ERP platform as a single platform supporting enterprise business.

In the DW/BI context, MDM related issues are quite important since, by definition, DW/BI is based on a consolidated view of enterprise data. Traditionally, DW integrates many heterogeneous source systems with different structures, different business entity definitions and different levels of data quality. In order to integrate these sources in a consolidated view of enterprise data, DW best practice recommends the implementation of ETL processes.

These downstream processes are aimed at solving problems related to inconsistent data structures, lack of data integrity and data quality issues.  The ETL process ends with the creation of a consolidated view of enterprise data in a DW/BI environment. However, the consolidated data view is not always a “single version of truth” since the consolidation process depends on the DW/BI integration rules and goals. Frequently, not all data sources are integrated in DW, or some data is rejected due to the DW/BI approach of managing data quality issues (usually two approaches: the “all or nothing” or the “fix it in ETL” approach) making the DW/BI owner of a “consolidated version of truth”.

The BI/DW approach tries to address the MDM issues through a system integration process and create a local version of master data through a concept of “conformed dimensions”. This approach will, without exception, result in important discrepancies in MD between the BI/DW environment and operational systems since there is no synchronization process back to OLTP.

The main goal for any MDM initiative is to solve problems related to master data upstream and to create a consolidated, unified and clear picture of MD, also known as, the famous “golden record”.  This must be done through a solid process of governance (policies, procedures, ownership) supporting the actions (data capture, integration, sharing) and assuring MD quality (accuracy, timelines and completeness). When the “single version of truth” is available upstream and shared across the enterprise operational systems, DW/BI is sitting on a goldmine. Finally, DW/BI and OLTP speak the same language and report the same results.

Therefore, DW/BI will need to revisit ETL architecture, which should be simplified since the main integration issues are now transferred to the MDM process. ETL as a data quality improvement process will remain part of the DW/BI landscape, some DW/BI specific issues will continue to be treated there.

Your enterprise MDM and DW/BI initiatives must be aligned or part of the same information management program to provide you a best return.  It is not necessary to envisage a “big bang” 3-5 years MDM program. Smaller, bottom-up initiatives can deliver a great value quickly, increase your program visibility and, ultimately, help you obtain the necessary support to go forward with your program.

Cube Groupe Conseil Inc. – Novembre 2012

Advertisements
%d bloggers like this: