Resilient Supply Chain Podcast: Why Better Forecasts Are Not Enough for Resilience
Supply chain forecasting is becoming more sophisticated, but better predictions do not necessarily produce better operational decisions. In the latest episode of the Resilient Supply Chain Podcast, host Tom Raftery speaks with Benji Price, Co-founder and Chief Product Officer at Wild Ducks, about the gap between forecasting accuracy and operational responsiveness. Their discussion examines how fragmented data, misaligned incentives and rigid planning systems contribute to excess inventory, stockouts and delayed decisions. For supply chain leaders, the central question is how to combine automation with human judgement to improve decision-making under uncertainty. The full episode is available at www.resilientsupplychainpodcast.com
When Better Predictions Arrive Too Late
Supply chain organisations have invested heavily in forecasting technologies, including increasingly sophisticated AI models. Yet improved predictive accuracy offers limited operational value if businesses cannot respond before conditions change. Price challenges the assumption that decision quality can be measured independently of decision speed. “The wrong decision made faster is probably better than the right decision made too slow,” he argues, qualifying the statement by emphasising that decisions must remain within acceptable margins of error.
The distinction has significant implications for supply chain management. Forecasting systems may identify emerging demand or supply constraints, but fragmented processes, slow approvals and disconnected systems can prevent organisations from acting on that information. Resilience therefore requires more than accurate predictions. It depends on shortening the time between identifying a change and implementing an appropriate response.
The Hidden Incentives Behind Excess Inventory
The conversation also examines why inventory levels can remain unnecessarily high even when planning systems recommend otherwise. Price describes a hypothetical planner responsible for ordering 100 units of a product. Following an unexpected demand spike, the planner faces criticism from both a customer and management. To avoid repeating the experience, the planner subsequently orders 200 units, despite the system recommending 100. That additional inventory may remain in place for years.
When similar behaviour occurs across hundreds of planners, individually understandable decisions can accumulate into substantial excess inventory. The underlying problem is organisational rather than purely technical. Planners face immediate consequences for stockouts, while inventory holding costs may be less visible or poorly reflected in performance incentives. For executives, this raises questions about whether planning metrics encourage decisions that benefit the wider business or simply minimise individual exposure to criticism.
Automation Cannot Replace Missing Knowledge
Automation offers a response to another fundamental constraint: scale. According to Price, major supply chains can involve millions or even billions of individual SKU-location combinations, making comprehensive manual optimisation unrealistic. However, automated systems depend on the information available to them. “These systems are only as good as what they can see.”
Critical operational knowledge frequently exists outside formal databases, including supplier relationships, customer priorities and exceptions known only to experienced employees. Price advocates automated routine decisions while retaining human oversight where financial exposure, uncertainty or missing information makes intervention necessary. This introduces an important governance requirement: organisations must define which decisions can be delegated to software, when exceptions should be escalated and who remains accountable for the consequences.
Forecasting Requires Better Inputs, Not Just Better Models
The discussion challenges another increasingly common assumption: that newer AI forecasting models necessarily outperform established statistical techniques. Price argues that matching forecasting methods to the characteristics of the data is often more important than adopting the latest technology. For intermittent demand, established approaches such as Croston's method may remain appropriate.
He also emphasises the importance of leading indicators. In agricultural equipment supply chains, for example, rainfall, drought conditions and crop economics can influence future demand. Incorporating these signals may improve forecasts compared with relying exclusively on historical sales. The broader implication is that forecasting quality depends on relevant information, appropriate modelling techniques and the organisation's capacity to respond.
From Forecast Accuracy to Operational Accountability
The conversation ultimately positions resilience as a question of decision architecture rather than technology adoption alone. Supply chain leaders must connect forecasting, inventory economics, operational constraints and human expertise while establishing clear accountability for automated decisions. Systems must also accommodate industry-specific requirements, including regulatory restrictions and physical operating limits.
The strategic opportunity lies in making decisions faster without sacrificing appropriate controls. Organisations that combine better information with clear escalation rules and adaptable planning processes will be better positioned to respond to disruption, control inventory costs and protect customer service.
