Throughout this past year I have stumbled across a lot of articles, blog posts, podcasts, and other content directed as all of the big buzzwords in analytics including, but not limited to, predictive analytics, machine learning, and artificial intelligence. For instance, Pierre Mitchel over at Spend Matters wrote a fascinating blog series at the beginning of the year entitled Artificial Intelligence in Contract Management and more recently Philip Ideson posted an interesting podcast entitled The Application and End Game of Machine Learning w/ David Hearn. Being fascinated by data, I devour this content. It’s always interesting to see the application areas that are, or might be in the future, explored by some combination of data engineers, data analysts, and data scientists. A nice snapshot of potential machine learning use cases within the domain of procurement was created by SAP Ariba and can be found over at slideshare.

Alternatively, as a grounded pragmatist, I can’t help but also notice the many doomsday articles that have proliferated over the past few years. A few choice statistics:


  •  At the beginning of 2017, Gartner projected that 60% of big data projects would fail. (1)
  • Bruce Rogers, Forbes’ Chief Insights Officer, claims 84% of companies fail at digital transformation. (2)
  • A survey discussed in a report by Pricewaterhouse Coopers (PwC) and Iron Mountain found that of those companies who responded, 43% “obtain little tangible benefit from their information” while 23% “derive no benefit whatsoever”. (3)
Clearly the pragmatist's perspective doesn't align with that of the data enthusiasts. How does one unify these two perspectives? How can we be certain that Nick Heinzmann's claims How AI Will Help Procurement Advance Analytics Beyond Basic Spend Analysis are not simply the those of an idealist? To be blunt, we can't be certain of anything except this: 

Growth as a company inherently requires risk. Risk cannot be eliminated, but it can be effectively controlled and mitigated through proper planning and execution. 
I'd like to emphasize the above words: planning and execution. As an example, consider the process of adopting a spend management tool. The tool will not simply improve your organization over night. Rather, as is the case with all new technologies, the new system will most assuredly have idiosyncracies that create new work flow problems that must be addressed. For instance, let us assume that the new platform is based off of some machine learning mechanism. Then if it starts to identify a tendency in the data that is wrong, likely through what is known as overfitting, then the system will probably reinforce this protocol until it becomes common practice. To be more precise for the analytically inclined, a  supervised machine learning system learns what to do from past data. If the past data is wrong then doing the wrong thing will be reinforced. Without occasional data reviews this error would be propagated and never fixed.

In summary, technology is a blessing and a curse. Using it effectively requires blood, sweat, and tears. It requires concerted effort and a strategic vision. Most effective data analytics workflows look something like this:

  • Data Engineering
  • Exploratory Analysis
  • Model Building & Validation
  • Production
  • Ongoing Maintenace

Yes, adopting procurement analytics platforms and processes will likely improve the efficiency and effectiveness of a practice at large. However, they only represent a small piece, or pieces, of the overall workflow. In many cases they will likely not completely eliminate roles within the organization. Rather, they will shift the nature of workloads. As a result, before heading down the path of building an advanced analytics practice, one should understand what the workflow for an advanced analytics practice looks like. By understanding a technology agnostic process, one will best be positioned to identify potential initiatives and prioritize them based upon likely ROI. What is considered to be a great risk ostensibly changes depending on the utility curve of the individuals involved. Only you, as a manager, can determine what level of risk is considered to be a reasonable one. 



Share To:

James Patounas

Post A Comment:

0 comments so far,add yours