An Oracle of Spend Classification or Just Another Spend Tool? (Part 1)

Earlier in the month, I noted in a post that Oracle unveiled a new spend classification capability at Open World. In announcing the features of the solution, Oracle clearly took a few cards from Zycus, Spend Radar, Ariba, Emptoris, BravoSolution and others offering competitive classification capabilities, noting that the module was designed to help "procurement departments categorize spend into a target taxonomy". It accomplishes this by relying on "Oracle Data Mining for machine learning and other predictive techniques to automatically categorize spend." But how does the solution stack up to other approaches in the market? Last week, I had the time to look at a relatively quick demonstration of the application and my initial impression was quite favorable.

The solution, which lists for only $40K per seat, enables users to classify spend on their own, without the help of armies of offshore resources (or onshore consultants). When considering this pricing, it is important to note that a typical organization will only need one or a handful of seats because the licensing is based around power-users who will be the ones doing the classifying. Still, before you get too excited about the price point, remember that it does not include other Oracle products (including their Spend Analysis capabilities) nor back-end software and hardware necessary to deploy the capabilities. Still, relative to other providers in the market, the price point is certainly aggressive for organizations that think they can get away with only a handful of seats.

Oracle Spend Classification lets companies classify spend using multiple taxonomies across disparate categories and internal coding (e.g., corporate, business unit, regional, facility). It relies on a formal ETL process to pull data in from both Oracle and non-Oracle systems (Excel imports are also an option) across all transaction types. Once in the system Oracle Spend Classification lets users classify spend using three different classification modes. These include an automated approach, which essentially puts the system on autopilot, classifying spend it deems above a user-definable accuracy threshold.

It also includes an assisted model, which permits end-users to provide feedback into the classification process to provide increased matching confidence (and in theory, also allowing a greater percentage of overall spend classification). Last, Oracle also enables users to classify spend on the fly during a one-time buying situation at the point of requisitioning if a particular good or service has never been recorded in the data set in the past. I saw this feature in action and theoretically, it's a nifty feature for those companies running an all Oracle spend shop.

Oracle suggested to me that their system includes an intelligence-based learning approach that "learns from your spend data, improves its capabilities over-time with user feedback, and continually enriches the system and overall accuracy". At the time of the announcement, I had more than one person come to me and suggest that Silver Creek systems, an Oracle partner, might have been behind the classification engine. But I confirmed they were not by calling Silver Creek directly. It is, indeed, an Oracle engine that the company claims has patents-pending on.

How does Oracle stack up to the competition? It's hard to say because I only had a short time to see the application (my fault, not Oracle's, mind you) and there are no references for the product yet. But I'll share my quick thoughts on how they stack up in my next post capturing the latest from Oracle. Check back later this week if you're curious to learn more about this products and others Oracle discussed at Open World.

- Jason Busch