How OR can help data science move from insight to impact

As organisations push towards faster, more automated decision-making, the gap between prediction and action is becoming harder to ignore. In a recent webinar for the OR Society, dunnhumby research data scientist specialist David Hoyle argued that operational research is increasingly becoming the missing link in modern analytics teams, particularly in industries where businesses must make rapid commercial decisions under complex constraints.

Speaking about the evolution of data science, Hoyle suggested that the next phase of analytics will depend not only on predicting outcomes, but on selecting the best course of action from multiple competing possibilities. While machine learning and forecasting models remain central to commercial analytics, he argued that optimisation and decision-making frameworks are becoming equally important as businesses move towards real-time, closed-loop systems.

The challenge is particularly visible in retail, where demand modelling plays a major role in pricing, promotions and stock management. Retailers routinely use demand models to estimate how sales may respond to factors such as price changes, competitor activity, promotional campaigns and seasonal variation. However, forecasting demand alone does not determine the best commercial decision.

A price reduction may increase sales volume but reduce margin. Promotions may drive short-term uplift while also shifting demand away from other products. Regional differences, customer sensitivity, marketing activity and external trends can all interact in ways that create difficult trade-offs. In practice, businesses need systems capable not only of forecasting outcomes, but of balancing objectives and constraints to determine the most effective action.

This is where operational research becomes increasingly relevant. Hoyle argued that organisations can no longer afford to treat prediction and optimisation as separate disciplines, particularly as automated systems become more deeply embedded in pricing, supply chain management and customer targeting.

The discussion reflects a broader shift taking place across commercial analytics. Data scientists are no longer only developing models that describe customer behaviour or forecast demand. Increasingly, they are contributing to systems that directly influence operational and strategic decisions in real time. That shift raises questions not just about model accuracy, but about objectives, trade-offs, constraints and how analytical outputs are ultimately used.

For operational researchers, the message is a familiar one. Prediction alone rarely solves complex business problems. The ability to make robust decisions under uncertainty, balance competing priorities and optimise outcomes across interconnected systems has always been central to OR practice.

As businesses continue investing in automated decision environments, the relationship between data science and operational research looks set to become even closer. Rather than replacing one another, the two disciplines increasingly appear to be converging around a shared challenge: turning data into effective decisions.