Cloud Hosting

Data analytics is a powerful tool for procurement professionals to unlock value in their data but it's far from a one size fits all.

By understanding the different types, and their relevance to procurement, leaders and professionals can make informed decisions that lead to more optimised processes and better outcomes.

Data analytics can be categorised into four groups: descriptive, diagnostic, predictive and prescriptive. Descriptive and diagnostic analytics are typically more basic, while predictive and prescriptive categories are referred to as advanced because they use more sophisticated methods and uncover deeper insights.

The four categories of data analytics explained

Where does machine-learning fit in?

While there can be an overlap between advanced data analytics (ADA) and machine-learning (ML), the distinction lies in their specific use cases, the amount and complexity of the data utilised, the sophistication required and the level of human involvement versus automation involved.

Both ADA and ML can uncover insights and help make informed decisions around procurement strategy and operations by targeting processes such as demand forecasting, inventory management, and spend analysis. Some cases, involving less structured and more complex data, require cutting edge ML. For example, if a procurement team wants to analyse large volumes of supplier feedback, customer reviews, or legal contracts to identify patterns, sentiment, or risky clauses, this would require state of the art natural language processing algorithms.

ADA and ML models can overlap, but ML algorithms typically require a higher level of mathematical and statistical knowledge compared to advanced data analytics. ML can range from simple linear and logistic regression models to more complex models like decision trees, random forests and neural networks.

ADA can involve a human carefully creating a model, which is then tested for validity. In ML, a human helps train a model to understand how well it can adapt and predict new data, given business constraints. But after that, the model can theoretically re-train and re-learn from new datasets on its own, making it more autonomous and dynamic.

Its also important to stress part of the confusion between ADA and ML is related to not distinguishing between models and processes when referring to these terms. An ADA process might be obtaining insights, for instance understanding the characteristics of suspicious financial transactions based on historical data, whereas an ML process would be continuous monitoring, eg. real-time prediction of suspicious financial transactions based on historical data.

In other words, even if ADA and ML might be using the exact same mathematical model, the ML process can include the ADA process in a way that automates and optimises the tasks the ADA performs.

So where do you start when implementing procurement analytics?

Identifying the low-hanging fruit is essential, and businesses should focus on projects that provide a direct connection to value, impact multiple areas of the business, and make it easy to envision the potential of ADA and ML.

Such swift, high-ROI, holistic procurement analytics projects are feasible when expertise in data science, research, and forensic accounting are combined.

Dr Kyriakos Christodoulides is the director of Novel Intelligence.

Continued here:
How data analytics and machine learning can transform your ... - Supply Management