Spend Classification is a topic that's coming into sharper focus with our clients as they look to better understand and analyze their spend and make savings across the organization. Oracle provide a Spend Classification complementary component for Oracle BI Applications to assist with this analysis and over the next series of posts we will look at the reasons for considering such a solution, how to implement it and how to actually make use of it.
Products such as Oracle Purchasing allow the buyer to select categorization when entering PO lines, however there's some issues here. Firstly, it's not normally really "sense checked" at this point, so as long as a valid selection from whatever categorization setup has been implemented is selected then all seems well. When the product being procured is "computer mouse" and the buyer enters "Hardware.Mouse" to categorize it, then that's a good start. However, they could just enter something generic like "Misc.Misc" or perhaps (even worse?) something misleading such as "Environmental Services.Mouse". This categorization has normally been setup by the Finance department but what if you actually wanted to analyze spend in a different way, not aligned to this categorization but a different taxonomy, also what if you had a number of ERP systems in the organization that had completely separate categorization taxonomies, then it's getting increasingly difficult to categorize spend.
What Oracle provide is a "rules engine" solution that can analyse loaded spend data and compare to a central knowledge base in order to assign a classification to the spend data that will allow the organization to slice and dice it how they like.
The following is the outline of the process :
· Create a training data for spend classification. Consider the "training data" as a cleansed dataset ( and a spreadsheet template is provided ) of all the categories that you desire and a number of examples real data with the fields which need to be assessed in order to make that judgement. Spend classification will find patterns in the "training data set" to describe the spend for each category and the more relevant information that it has the greater the degree of accuracy.
· Create a Taxonomy. Basically, this is hierarchical "Hardware.Mouse" breakdown of all the categorisations to be used.
· Create a Knowledge Base based on the taxonomy and the training data. This knowledge base will be used as the master repository against which to compare any batches of spend data that need to be analyzed. If people can buy computer mice from different suppliers, have different descriptions, etc then the more complete the knowledge base, the more accurate the classification process will be.
· Perform the classification of a real spend dataset. This utilises the knowledge base as a "rules engine" to assign categorisations to each element of spend.
· Check, Amend, Approve.
This sounds straightforward, but there's a lot of background detail to this process; for example one can employ a standard knowledge base using APIs or an advanced model that utilizes Oracle Data Mining techniques.
I was initially a bit cautious when performing the installation of the product as the current version is "7.9.6" which is a nod to the older version of BI Applications. Fortunately, it is indeed certified with the 18.104.22.168.1 releases of BI Applications - and there's a script specifically to change some of the product views to align them to the new data structures provided with the 11.1.1.X versions of BI Applications.
Once the installation is complete, the application is accessed in a rather unique way that makes access quite seamless for users as some of the Spend Classification dashboards are actually containers for embedded applications as can be seen below.
The tool actually does integrate with iProcurement so that when any off contract spend is performed, then the buyer can be provided with an assisted pop-up that suggests appropriate categorisation to at least help out in that respect. This requires the profile "POR: Enable Category Classification on Non Catalog Request page" to be set.
The next planned blog entries for Spend Classification will look at the key documents that Oracle provide for this product, notes on the implementation and how to actually use it to perform analysis. Stay Tuned!