The bullwhip effect on inventory: a perspective on information quality

ABSTRACT

In recent years, there has been a vast increase in the quantity of information shared across supply chain. We investigate how the timeliness and accuracy of information quality affect the value of information sharing. We use the inventory bullwhip effect (BWE) to measure the value of information sharing in a two-level supply chain consisting of one retailer and one manufacturer. The retailer faces a price-sensitive demand and the price is an AR (1) process. Our study shows if customer demand and retailer’s immediate order are delayed, using retailer’s historical order quantity to forecast can decrease manufacturer’s BWE. If information errors happen, during delivery and utilization, information sharing is not always valuable for the manufacturer. Sometimes, no information sharing can decrease much more of BWE. If information errors occur when the retailer collects demand information, value of information sharing is more significant than when there are no information errors.

KEYWORDS:

• Information sharing
• information quality
• bullwhip effect
• inventory

JEL CLASSIFICATION:

• C22
• C53

Acknowledgements

The research presented in this article was supported by the National Natural Science Foundation Project of China (71390335, 71390333 and 71572145).

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ http://www.pbinsight.com/files/resource-library/resource-files/dirty-data-not-my-problem-wp.pdf.

² http://www.cio.com/article/2439601/supply-chain-management/nike-rebounds–how–and-why–nike-recovered-from-its-supply-chain-disaster.html.

³ The model here can be extended to the case in which there is one manufacturer and more than one retailer, or multilevel supply chain consisting of retailer, manufacturer, supplier and so on.

⁴ In this article, the goods sold in the market are usual commodities. The demand increases when the price decreases and vice versa. Thus, $b>0$. We do not consider Giffen Goods here.

⁵ In real business, market price information is always public. Thus, we assume that the manufacturer can get ${\eta }_{t-1}$ and record the price information in period $t-1$ without incurring much cost from the market. Thus, we think price information is known to the manufacturer while it is making decision at the beginning of period $t$.

⁶ Because of $q_t=-b\rho {\mathrm{\Psi }}_{L_1}\left(p_{t-1}-p_{t-2}\right)+d_{t-1}$, if manufacturer knows $q_t$, $p_{t-1}$ and $p_{t-2}$, it can calculate the true value of $d_{t-1}$. Consequently, the value of obtaining information about the actual demand from the retailer will be reduced. Furthermore, since it is complicated to conduct a theoretical analysis of the value of information sharing when the manufacturer utilizes historical price information and order quantities to estimate the actual demand, so we shall limit the scope of this article by assuming that the manufacturer would not estimate the true value of demand feature.

⁷ These results have been reported in our previous work (Wang et al. 2016). We just restate them here in order to compare the situations with information quality problems.

⁸ In Equation (5), the coefficient of ${\epsilon }_{t-j-1}$ is ${\rho }^{j+1}{\mathrm{\Psi }}_{L_2}$. However, in Equation (8), the coefficient of ${\epsilon }_{t-1}$ is $\rho {\mathrm{\Psi }}_{L_2}$. ${\rho }^{j+1}{\mathrm{\Psi }}_{L_2}$ is smaller than $\rho {\mathrm{\Psi }}_{L_2}$. So, we say ${\epsilon }_{t-j-1}$ has less influence on $Q_t$ than ${\epsilon }_{t-1}$.

⁹ For example, $q_{t-5}$ is older than $q_{t-1}$.

¹⁰ Because we assume the price information is public and easy to be obtained, therefore, it will be correct while retailer collecting market information.

¹¹ When $b^2{\delta }^2>\left(1-{\rho }^2\right){\beta }^2{\omega }^2$.

¹² $k={\mathrm{\Phi }}^{-1}\left(p/\left(p+h\right)\right)$ and ${\mathrm{\Phi }}^{-1}\left[\cdot \right]$is the inverse of the cumulative distribution function (CDF) for the standard normal distribution, $\varphi \left(\cdot \right)$ is the probability density function (PDF) of the standard normal distribution. $k$ is a constant that has been set to meet a desired service level. $h$ and $p$ represent the holding cost and penalty cost of retailer.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China: [Grant Numbers 71390335, 71390333 and 71572145].