Written by Matt Holzapfel
“The overall goal was to centralize the planning process to have one single procurement department.”
I was taken aback when I heard Dave Girouard (founder, CEO at Upstart; former President of Google Enterprise) say this in reference to a consulting project he worked on at Arthur Andersen (more than 30 years ago) in the late ‘80s for a global, multi-billion dollar manufacturer. This is the exact same problem that I encountered in 2008 when working in procurement at Dell, and one of the common challenges we’re solving at Tamr for our customers today.
Which leads to the obvious question: If this has been such a big problem for 30+ years, why has it taken so long to solve?
Simply put, Enterprise Resource Planning (ERP) systems were not designed with information silos in mind. These tools were designed to be universally adopted throughout the enterprise. Organizations operating with multiple ERPs and systems of record are unable to reap the full benefits that ERPs were designed to deliver. Specifically, gaining a centralized view of business activity has required a herculean effort when information flow is decentralized. Until now.
Recently developed products, such as Tamr, use machine learning-based approaches to automate the time intensive process of matching and categorizing data from across different sources to create a consistent, global source of insight. This provides sourcing teams with the information needed to operate as a centralized unit, regardless of their actual structure. As we’ve seen with our customers at Tamr, this view can help procurement drive full percentage points off their spend and reshape their supply-base to align with strategic shifts in the business.
The Source of Information Silos
Businesses’ need for a common and trustworthy source of information has remained consistent over time. This is evident in the evolution of the modern ERP system, which began as a system for tracking Bill of Materials information for manufacturers. As technology evolved, additional features were developed to automate more business processes and integrate more business activities, all with the end goal of having an enterprise-wide system for planning and managing business activity.
Today’s ERP systems have succeeded in delivering the features needed to serve functional groups throughout the organization, but the bottleneck to gaining a centralized view of activity has shifted. Many ERP systems were implemented in silos – optimized around the needs of individual business units and facilities, not around a common platform. As a result, each of these silos has a central view of their activity, but rolling up this information to gain a global view and coordinate disparate functional groups, such as procurement, is a significant challenge.
Breaking Down Silos
The solution to this problem 30 years ago was to hire outside consultants, such as Dave’s team at Arthur Andersen, to develop custom software for integrating these systems together. This was both expensive and time consuming. As an example, gaining a centralized view of all POs required millions of dollars and months of development time alone. Fully centralizing processes into one system could easily take multiple years and tens or hundreds of millions of dollars to accomplish.
Some companies still solve this problem in a similar manner as Dave’s client — by hiring consultants and contractors to develop a custom integration across systems. Others leave the problem unsolved, and hire analysts to manually integrate data across sources to gain a global view of their procurement. A large reason for this is because there hasn’t been an elegant solution to this problem until recent products, such as Tamr, came to market.
Specifically, a sustainable solution to this problem requires a machine learning-based approach that learns from user input. The machine learning aspect is critical because simply taking a rules-based approach to procurement information will quickly fail once the items being purchased change or a new buyer is hired who uses different naming conventions. Further, user input is needed to ensure that the machine is learning from experts in the business who understand how data should be integrated and categorized.
The end result is clean, trustworthy information available at the global level, which can be used to drive critical strategic sourcing decisions and enable procurement to centrally organize. Our customers who take this approach are realizing millions of dollars in cost savings in a matter of weeks.
There is finally reason to believe that the decades old problem of having the information and technology needed to centralize procurement is being solved for good.
To learn more about Tamr’s sourcing analytics solution and schedule a demo, click here.