Intela launched BI at the beginning of 2011 with ambition of a) consolidating multi-source data at the backend with sophisticated processes of data cleansing, standardization, and transformation, b) presenting business a single version of data with high level of accuracy, c) powering business with automated processes and tools for analytics on dynamics of business based on close-to-near-real-time data including customer segmentation based on demographics and behavior, and d) providing power to business to make quick business decisions of campaign management.
Intela made this decision to embrace the primary challenges we were facing:
- There were all kinds of data we possess but there is no easy way to access it and integrate it.
- There were possibly multiple versions of the same data, along with possibly different ways in which the business users referred to such data.
- There were data integrity issues that were not resolvable at the time when data is needed for business, due to volume, format and sources of data.
- Manual work on collecting and analyzing data done by business people were usually very time-consuming and error-prone.
BI at Intela consists of Presentation Layer and Storage Layer.
- Data can be viewed in
- Dashboards for high level views with the capability of linking to specific reports to provide more detail information
- Canned reports that provide flexibility for business users to scope reports with parameters like a time range and/or other variables.
- Self-served reports for analytics that allow business users to build their own reports through drag and drop. It also allows business users to drill through from a high level all the way down to a transaction level.
- Emailed reports that provide business users with static reports on a regular basis.
- Implemented with Pentaho
- Data stores in a form of star-schema.
- Data marts are built driven by application and data consolidation.
- Star schema is multi-layer by nature with various forms of summary.
- Dimensions are shared across data marts.
- Data gets updated from multiple sources through ETL.
- ETL is implemented to extract data from multiple sources, to consolidate data, to cleanse data, to standardize data, to transform data before data is made into the target schema ready for uses. It is also equipped with data quality auditing functionality.
- ETL is implemented using Informatica.
- Database platform is MySQL with load balance through replication.
Although we are still in the process of building and enhancing BI at Intela, we have already started to appreciate the presence of BI at Intela. We have seen the following benefits:
- We have a much better understanding of our data. This has allowed us to identify the data issues and gradually fix the data to provide more accurate information to our users. .
- We have built a data warehouse that allows us to view, analyze, and report data and data relations because we are able to integrate revenue, cost, net/gross registration information that come from different data sources.
- We have gained better data quality in reporting thanks for data cleansing, standardization and transformation through ETL.
- We have made data available to business users much sooner than ever. The automated process in BI has greatly decreased the time it takes to get useful information to the business users immensely. .
- We have saved incredible labor cost to handle requirements for ad hoc reports.
- All of the above show a good ROI, considering our investment in BI development.
Post by: Ningsheng Liu, Database Manager & Architect