Contextual factors such as product margin and problem framing trigger cognitive biases that lead decision-makers in operations to deviate systematically from optimal behaviour. This paper investigates the impacts of experience (as learning-by-doing) and training (as instruction-based learning) as different sources for adjustment in newsvendor behaviour and compares their effects on ordering performance in a within-subject experiment with 152 participants across two sessions. Measuring performance as deviation from the optimal newsvendor quantity, we find that theoretical instruction produces a significant fall in the mean of absolute deviation of 29%. This effect is approximately 4.8 times larger than the cumulative improvement attributable to within-task repetition alone. Both within-task and between-task experiential learning channels attenuate to near-zero after the lecture intervention, indicating that the two mechanisms target overlapping corrigible components of ordering bias.The training effect is heterogeneous across different product margins (high vs low) and problem framing (cost vs gain) with the largest reduction in high-margin and gain framing, where pre-training deviations were greatest. Training selectively neutralises framing-amplified distortions rather than eliminating deeper anchoring tendencies; indeed the pull-to-centre bias persists post-training in all conditions. The results suggest that instruction produces faster, larger gains, but persistent performance plateau remains. These findings have direct implications for the design of training interventions in supply chain and operations management practice.

Keywords

Newsvendor model;  Behavioural Operations Management; Learning-by-doing; framing; instruction-based learning