EPM vs BI: Spot the differences

EPM vs BI: Spot the differences

Do you know the difference between a butcher and a surgeon?

Of course, even if both cut meat, the stakes are obviously not the same. 

Now, same question for BI and EPM. You have a doubt ? Then this article is for you, and I hope that at the end of this article you will not confuse them anymore.

A bit of history

BI was born with the birth of databases and the limitations that computer scientists had in using these volumes of data and making them intelligible. Indeed, relational databases made it possible to store increasingly large volumes of data, but these were complex to analyse and restore (mainly because of the technical limitations of the infrastructure – CPU, RAM and I/O). BI started to create toolboxes (for interfaces, statistical processing, …) and then reshaped the way information was stored to make it more easily and quickly accessible.

To a technical problem, the computer scientists responded to a technical solution.

Then other problems appeared understanding of certain business rules, interdependence with IT making certain processes inert, lack of sharing and adoption of practices by other departments within the company, qualitative data processing.

Indeed, finance, which must, by definition, “manage the risks of the company”, was quickly interested in the new tools developed by BI. The more information we have about an activity, the less uncertainty there is, the lower the risk. However, BI has had difficulty in moving away from IT technicality to adapt to the needs of finance.

A technical language, maintained by computer specialists who only gave back information at the end of the chain, was replaced by processes and applications for finance teams, giving access to a financial toolbox and allowing better collaboration between stakeholders (the main lever for reducing operational risk): the EPM was born.

To come back to the initial image, the width of the blade did not allow a clean and efficient work. This led to a diversification of the cutting tools, and some techniques, to move from a large work, to a more precise and less risky work.

... and semantic

According to Wikipedia, business intelligence BI is IT for decision-makers. It refers to the means, tools and methods that make it possible to collect, consolidate, model and restore a company’s data, whether tangible or intangible, in order to provide decision support and enable a decision-maker to have an overview of the activity being processed.

Two remarquable points on the definition:

  • The absence of processes (or man-machine interactions) in the definition is a first point. For BI, the machine replaces man, and becomes passive (except for computer scientists). Users are data consumers.
  • BI assumes exhaustive data at the beginning of the chain, no interaction exists between the beginning of the chain and the end, to validate, complete or control the consistency of the data,

Now EPM according to Gartner: EPM is the process of monitoring performance across the entire enterprise with the goal of improving business performance. The terms tools and IT have disappeared. We now talk about processes. This process is actually based on BI technology, but has been redesigned for the finance function. Moreover, quality is driven by the desire for monitoring and improvement. The aim is no longer the exploitation of company data for decision-making, but the improvement of performance.

Different stakes

BI, as seen above, aims to exploit the company’s data. The speed and quantity of information, overrides interactions, and qualitative information.

EPM is based on three pillars: automation, collaboration and auditability. Because automation is the prerogative of machines, BI is a very good (technical) support for this process. Auditability issues are not necessarily integrated, especially when it comes to making information related to the human or machine process intelligible. Finally the collaboration aspect does not exist. BI responsibilities are not shared outside IT departments, and users have a passive role as consumers of data.

Technologies not so similars

The BI uses datamarts and then restores via OLAP technology (cubes). The cubes make it possible to pre-digest information in order to restore it more quickly. Thus aggregations are native on all axes, as opposed to relational databases that use complex and expensive indexing systems to group and restore each subtotal.

BI, with OLAP recipes, is thus a very good tool for EPM except that (non exhaustive list):

  • Intra-group reconciliations and controls, a process allowing to control the declarations, and which is based, ideally, on transactions that have to be reconciled one by one: in this case the transactions impose a relational database and aggregation is useless,
  • non-reciprocal elimination mechanism: the OLAP works correctly in the case of pyramidal aggregation of data, one starts from a single base (a fine level) and goes up to the top, however some EPM processes are not pyramidal, rely on several bases (several stacks, usually 2, a declarer and a receiver), or the aggregation system no longer works naturally, and the rule of course (of the calculations) no longer follows the computer logic,
  • some technologies (such as in-memory) compete with OLAP operations, which remain limited to disk access capacities, where an EPM base can hold (in-memory) by prioritizing / limiting cached years,

On these three objections, hybrid systems are increasingly equipping EPM tools, reinforcing the difference between these two decision concepts.

EPM and BI may have common goal, but the way to reach them is different. BI gave interesting technologies to EPM, but EPM had to leverage others capacities around controls of information, and collaboration, which may be complexe or impossible with a BI focus.

Still not convinced: let me propose a very good artisan for your appendix removal 🙂