When we talk about enterprise resource planning (ERP), we associate the acronym with the totally integrated application suites delivered by the likes of SAP, Oracle, Lawson, PeopleSoft etc. A fundamental tenet of an ERP system is that it contains a ‘single source of the truth’ (ie: transaction data) held in a highly efficient transactional database, usually built on a row-based relational database technology, for access and use by all modules within the suite. The goal of an ERP solution is to support the operational activity, streamlining the exchange of information between the many functional business units involved in supplying either products or services to the customer. So, typically, the core application suite is configured or customised to fit an enterprise’s unique character and often requires niche or specialised applications to be bolted on and integrated if not at the process level, certainly at the database level, avoiding, where possible, duplication of data. At this point, the management focus is on reporting and accounting for what has happened, stating what the current status is, and identifying how well the business is performing against the business plan. Enter the enterprise data warehouse (EDW), not just a repository of transactional data, but a ‘storehouse’ encompassing transformed, extended and derived data sets used in online analytical processing (OLAP). For instance, derived data may include evaluation of some comparative measure for each transaction, such as a cost to deliver, assessment of quality or categorisation of consumers. The EDW enables both reporting and basic analysis of the transactions to provide the variety of dashboards, key performance indicators, balanced scorecards and the like relied upon by end-users and management to maintain control and support operational decisions. The characteristics of an EDW are:
Not surprisingly, ERP vendors have responded by expanding their data repositories to encompass the EDW functionality and to include analytic tools that provide easier ad hoc query and end-user reporting capability. Furthermore, through acquiring business intelligence (BI) tool specialists, ERP vendors have integrated BI capability and are planning to claim this space as well. Oracle’s acquisition of Hyperion in 2007, then SAP’s acquisition of Business Objects and IBM’s acquisition of Cognos in 2008, all go towards illustrating this point. Nonetheless, BI is not a predicated path of processing but a reiterative process of discovery, and today’s advanced analytics involve techniques from data mining to pattern recognition and statistical pattern learning. The goal of BI is to identify factors and relationships that would have otherwise gone un-noticed, and for the management and end-user to then formulate strategies, policies and procedures to enable exploitation of the knowledge provided using their own savvy. What’s more, the "art of BI" to recognise the value of data sets external to the EDW and find ways of introducing, experimenting and incorporating the public and market indices into the models generated by the BI engines, has resulted in database structures that do not sit well with the normal EDW design. So, is it reasonable to expect the ERP vendor to provide out-of-the-box BI tools? As observed by Phillip Russom, Senior Manager of TDWI Research at The Data Warehousing Institute* : "Given the unpredictable nature, quickly evolving data, and demanding workload of advanced analytics, users are choosing to offload analytic data from an EDW to a secondary platform called an analytic database. Any database management system (DBMS) can manage an analytic database. But in response to demand this decade, software vendors have produced new DBMSs that are purpose-built for analytics or DW. Many new analytic DBMSs are now available, based on appliances, columnar data stores, MapReduce and open source. Some of these are available through ‘clouds’ or ‘software-as-a-service’ licenses." At the end of the day, a business expert is still needed to take the outcomes of a BI analysis exercise and apply a real-world interpretation. However, the accuracy, refinement and validity of the findings are more likely to provide solid material for business decisions where the data has been sourced from an ERP than from a multitude of disparate databases.
- All-encompassing and scalable (all the data for all the business units is held in one repository
- Clear definition of data interpretation and relationships (no ambiguity)
- Compliant levels of quality control for accuracy, timeliness and completeness that can be applied across all contributors.