Realizing the promise of predictive procurement
The promise of predictive procurement is one that has been unfulfilled so far. Has AI in Procurement flattered to deceive? It certainly has not delivered the kind of results experts predicted. Millions of dollars in savings due to deal-value and strike-point predictions? Un-check. Smart vendor selection based on predictive vendor-performance? Un-check again. Clustering algorithms to provide obvious insights into procurement deals that can be otherwise done in an Excel-spreadsheet. Unfortunately, check. So here are 3 reasons Data Scientists should be actively recruited — that’s right — should be.
1. Data is the new oil and we need refineries — fast!
I come from the world of Oil & Gas, hence this is my favorite analogy — Data is the new Oil! This analogy is true in so many dimensions. For example — not all crude oil is equal. There is sweet crude oil that is easy to extract high-value gasoline with low sulfur and there is the “dirty” crude or sour crude oil that has high sulfur and tough to extract gasoline from. Regardless of the quality, the effort needed to extract high-value gas still needs to be put in. Residuals/Fuel oils require little effort to extract and are often the least value product, while gasoline requires a lot more effort and investment for extraction.
Each stage of the fractional distillation column is analogous to the different levels of data processing and analytics maturity in any organization. Manual data capture and isolated analysis typically using spreadsheets are the “least-effort, least-value” products in this “Data refinery.” Many procurement organizations operate at a higher level, with more automated data capture, which in turn feeds a multitude of reports and dashboards. This is more of “Descriptive Analytics,” often presenting results from past data — past performance, past deals summary. At this stage Dashboarding and Reporting Tools are used. “If we only knew this was coming before, rather than afterward,” is a common refrain heard when operating in this stage.
The most advanced stage comes into the realm of “Prescriptive Analytics.” By this stage, the data has been cleansed so that predictions can be done on the data. Typically, Data Scientists are involved at this stage, leveraging the full power of AI and ML. The transition from just BI to AI is by no means a simple one — and this is precisely the stage where Data Scientists can be critical. Cleaning up of data and aggregating information at the right levels often involves changes to the very systems that are used for data gathering. Data Scientists know the end state very well and so are critical in making these changes. Data Scientists are also expected to “get their hands dirty” by actually cleaning up the data to make it conducive for predictive analytics.
2. Invest sooner, reap rewards sooner
The Gartner Hype Cycle © Gartner, Inc. for AI gives a good picture of how companies are trying to adopt and demystify AI.
From the Hype Cycle, it is clear that there are only a few AI technologies that are less than 2 years from reaching their “plateau of productivity.” Robotic process automation and Speech recognition fall in this bucket and we can certainly see these in action in many procurement organizations with many manual procurement tasks now already automated. Most AI technologies fall into the 2–5 years bucket — of these, the most relevant for procurement being Machine Learning, Deep learning, Chatbots and Auto-ML.
For the strategically inclined leader in Procurement, investing in these technologies now would definitely reap rewards in the medium term (2–5 years) and that means — hire Data Scientists to start working on these applications now! Of course, there will be a “trough of disillusionment”, but as you can see that’s just the way to the “plateau of productivity.”
The more relevant Hype Cycle is that for procurement and sourcing released by Gartner in 2018, which shows that AI for Procurement and Smart Contracts are in the 5–10 year turnover range. These are currently in the Innovation Trigger stage.
3. Promote a Statistical way of thinking
Now, this may sound not that important, but I believe a statistical way of thinking is key for any organization that aims to be Data-centric. Data Scientists are, at heart, statisticians building statistical models. By definition, they have to deal with uncertainties and probabilities. They have to understand what the sources of variations and biases are as part of building any statistical model. The great statistician/mathematician Abraham Wald’s work on the analysis of British bombers that survived attacks during World War II, is a great example of statistical way of thinking.
Chris Fonnesbeck’s superb talk on the statistical way of thinking for Data Scientists highlights some of the pitfalls associated with relying on non-representational datasets for coming to conclusions and ultimately driving decisions. Good Data Scientists are constantly aware of these pitfalls when gathering data to build models. This in turn, forces changes upstream — in the systems that help generate and gather data. In case of procurement, these are procurement managers and specialists. In order for this to happen completely, there has to be a good integration of Data Scientists and Procurement specialists in one team — i.e. both of them need to understand each other’s thought process. In other words, Data Science teams in procurement cannot operate in isolated silos.
In conclusion…
Procurement is an organization that is very ripe for disruption. In many companies this org is a mix of old-world wizards of negotiation, who rely on a lot of tribal knowledge, as well as a new crop of data-savvy procurement specialists. I constantly dream of the day when 100 % of the decisions in procurement are 100 % data driven — and Data Scientists lead the way in transforming this function into a lean mean negotiating machine!
