Machine Learning & AI: Powerful new tools for procurement professionals
Welcome to the future
According to the Big Data Dilemma Inquiry, a shocking 90% of data currently available in the world was generated in the last few years. Organisations leveraging this data are on average 10% more productive than those who don’t, but most companies estimate they are analysing only 12% of this data.
Spend data is no exception, even though complete visibility of an organisations spend is the bedrock of procurement excellence.
Data is difficult
So why aren’t organisations leveraging the data they produce?
Performing a spend analysis - the process of collecting, cleansing and classifying data into meaningful supply market categories that can be transformed into insight - is time and labour intensive.
It requires both market expertise, and a unique understanding of how your business functions. It may require collating and synthesising data from multiple disparate sources such as ERPs, finance systems and BI data warehouses. Done properly, this requires a significant investment of time from often under-resourced procurement departments.
Making the process repeatable is also a significant challenge. Without a well-established data structure, that feeds results into easily digestible form, such as a spend analytics platform, outputs are often stuck in a procurement department silo, leaving thousands of savings never to be implemented.
But the real, bottom line, value that procurement departments produce comes from the insight that is gained from this process, rather than the process itself.
Enhancing Decision Making
This is where a ‘Machine Learning’ and ‘AI’ (Artificial Intelligence) toolkit come in.
Classification and Natural Language Processing algorithms trained on large sets of historic spend data can partner with the procurement professional to automate spend analysis, freeing up their time to engage in real value-add activities such as strategic sourcing.
Building an effective data processing pipeline, combined with a robust spend analytics platform, can provide visibility of spend across an organisation that unlocks new levers for cost reduction.
For example, international companies can leverage visibility of spend across subsidiaries to take advantage of opportunities for consolidation of volumes or Global Sourcing. Refreshed regularly, this can lead to improved insights that are maintained over the long term.
Managed effectively, this process provides an organisation with the tools to identify and predict trends in spend, increase spend under management, and can partner with other initiatives such as eProcurement to massively increase productivity and improve the results of the procurement department.
There are limitations
Neural Networks, the powerhouse behind the AI revolution and the key to increasing machine classification accuracy are incredibly data hungry. In order to provide a meaningful output to a classification problem, they require large volumes of pre-categorised ‘training’ data.
Additionally, whilst these machines are ‘smart’ in producing a specified result given a large volume of inputs, they are simultaneously ‘dumb’, in that they cannot synthesize information into actionable output in the same way a procurement expert can. For now, these tools also cannot develop higher level strategy that takes into account the specific needs of the business as well as expert knowledge of the supply market.
There is also the consideration of accuracy. No predictive machine learning algorithm is 100 % accurate. Any algorithm producing a 100% accuracy score is likely ‘over fitting’ to the training set. Some manual classification, with contextual knowledge of the business and supply market, will always be required to drive categorisation accuracy from 90% up to 100%.
Pulling it together
Effective procurement has always been a data driven process, and advances in Machine Learning and AI will vastly increase the speed and accuracy of data processing. Organisations that can take advantage of this technology to enhance spend visibility will be able drive significant savings with leaner procurement teams. The key to unlocking this value is developing a robust data processing strategy, as well as training procurement professionals to understand the advantages and limitations of predictive models. The combination of machine and expert knowledge is the best way to leverage these outputs for effective procurement strategy development.
Points to consider
- Leveraging data for spend visibility is key for effective spend management.
- Machine Learning and AI, when coupled with a robust spend analytics platform, are fantastic new tools that can be utilised to increase the productivity of a procurement department.
- A realistic understanding of the advantages and limitations of predictive modelling is essential for effectively harnessing the technology to improve procurement strategy.