A Critical Note on Information Distortion in a Supply Chain The Bullwhip Effect

TL;DR

This paper critically examines the assumptions behind the bullwhip effect caused by order batching in supply chains, revealing that previous models may overstate the impact and proposing a more accurate variance-based analysis.

Contribution

It identifies and corrects two improper assumptions in LPW's model, providing a refined analysis that challenges the established link between order batching and the bullwhip effect.

Findings

Order batching does not necessarily cause the bullwhip effect under certain assumptions.

Correct modeling of demand aggregation alters the perceived impact of batching.

Variance decomposition reveals nuanced effects of demand correlation on supply chain fluctuations.

Abstract

In the seminal paper Information Distortion in a Supply Chain: The Bullwhip Effect (Lee, et al. 1997, hereafter referred to as LPW), order batching is regarded as one of the four sources of the bullwhip effect. LPW proved that, in all cases (random ordering, balanced ordering, and correlated ordering), order batching will surely lead to bullwhip effects. However, we identify two improper assumptions in LPW. First, the batched order Zt is de facto the moving summation of previous demands, including overlapping demands. In fact, the batched order should be modeled as periodic summation of previous demands. Second, in the random ordering case, the number of retailers n is modeled as a binomial variable which is identically distributed for a randomly chosen period t in a review cycle. In fact, n should follow a sequential hypergeometric distribution. To address the two issues, we decompose a demand sequence using law of the total variance, exploring variance interplay between batched and non-batched demands in the positively correlated case. We find that even under the contrived iid assumption in LPW, order batching does not necessarily lead to the bullwhip effect.