Effective spend analysis is not possible without precise data classification. Without classification, everything is cluttered and unorganized. Many of the large enterprises Sievo works with have many ERP’s, each having its own structure, material numbers, accounts and a lot more. Let’s take the example of a laptop. In a larger organization, this one item can be called differently depending on how it has been sourced in tens of different ways, it can be Dell X260, Power Laptop, or perhaps Ordenador in Spanish.
Unifying heterogeneous spend data into clearly defined categories makes them easier to address and manage across the whole organization. Even with the example of the pen, knowing exactly how many pens are bought by different departments and subsidiaries of a company and at what price gives procurement the upper hand in negotiations with new vendors. So, in essence, classification is about harmonizing all purchasing transactions to a single taxonomy enabling customers to gain visibility to the global spending in order to make better sourcing decisions.
Once you’ve properly classified your data, results are presented in intuitive reports and dashboards right away. The harmonized data can then show potential savings and opportunities.
The answer to accurate classification is not a black box
Complex data classification solutions are not always simple to implement, but they can be improved over time. What stands out with Sievo’s data classification tool is that it’s not a black box—it is transparent. Customers can see how information has been classified and make exceptions to the classifications where they feel necessary. For example, if one category is classified wrongly, it can be changed effectively as needed. Sievo empowers users to add data categorization based on their own business logic and they can easily classify data through intuitive add-ons and different levels of categorization. They can slice and dice data to smaller managable junks on any attribute. Hierarchical classification decisions enable them to manage and pinpoint exceptions both at scale and in detail.
The visibility for the customers makes the whole process collaborative as the customers can freely participate and make exceptions if necessary. Once classified, the information goes standard across the organization and users can see based on what data characteristics the classification has been done.
The key to success is collaborative classification
According to Spend Matters, Sievo’s collaborative classification engine is one of the most distinct spend classification platforms in the market today. Not only does it break down spend into categories and sub-categories, the classification can also be assigned to customer category experts. This results in expert system implementation where the knowledge of the expert is capitalized in ensuring deeper level classification within a category or sub-category and resulting to higher level of accuracy.
For organizations with billions of dollars of spend, this level of collaboration is revolutionary. No one person can be expected to know the full organization’s spend inside-out, but individual category managers can contribute to a more accurate level of classification, and the end result can be ground-breaking. The best practice in spend classification suggests at least 90% accuracy to identify spend opportunities, however reaching 95% or better accuracy level is crucial for savings tracking and management. As the further breakdown occurs for a deeper level reporting and tracking management, Sievo’s collaborative classification approach allows spend to be mapped down to the last dimension 100% error free.
The future is human + machine collaboration
Even when category experts can improve classification close to the 100% error free level, there is still the question of prioritization. How many hours of an expert’s time is worth spending improving the classification of less strategic, long tail spend?
The answer to this comes from a relatively new field of computer science called machine learning.
At Sievo, we’re developing an approach of human + machine collaboration, where the advanced knowledge of procurement experts is complemented by our own proprietary machine learning solutions. In practice, we continuously train computers to learn and adapt the classification techniques performed by humans. The more data computers see classified, the more they learn. This approach is currently tested with a limited number of clients, but we believe in the future we can reach even greater levels of precision on data classification by using the best skills of both humans and computers – the next stage in a truly collaborative classification experience for procurement organizations.
Proceed to Part 3: Spend Analysis
Back to Part 1: Data Extraction