On the value of information sharing in the presence of information errors

Highlights

• Information errors have a huge impact on the value of information sharing.

• Different information errors have different impacts on the manufacturer's decision.

• Both information sharing strategies may be valueless with transmission error.

• Only demand-only information sharing may be valueless with source error.

Abstract

Errors may occur in the information received by a manufacturer from the retailer. There are two types of information errors: transmission error that occurs in the transmission of information from the retailer to the manufacturer, and source error that occurs in the data observed by the retailer. This paper studies the value of information sharing in the presence of either or both types of information errors. In particular, when information is shared, the manufacturer may use both the shared demand information and the retailer's order quantity to make decisions, or it may rely solely on the shared demand information and disregard the retailer's order quantity when forecasting demand for the future. We analyze the values of information sharing for both settings, and characterize the lowest forecast error information sharing strategy for the manufacturer. Our results suggest that transmission error and source error have significantly different impacts on the value of information sharing and the manufacturer's optimal strategy.

Introduction

Information Sharing is valuable in supply chain management. It can make a 2% improvement in on-shelf availability, 10–15% reduction in inventory and a 10% increase in forecast accuracy (ECR Europe, 2011). However, according to a survey in 2017, 28% of companies are not using data shared by retailers, and one of the key reasons is that “the quality of the data is insufficient” (Askuity, 2017). When there is asymmetry between information available to different members of a supply chain, reliable information sharing becomes critical for the supply chain to operate effectively (Spiliotopoulou et al., 2016). Data quality is one important issue in the digital economy era (Liu et al., 2018, 2019). Poor data can have a huge negative influence on operations management. A survey of 200 IT decision-makers and influencers discovered that 94% of respondents believe that business value is lost because of poor data quality — 65% of respondents believe that 10–49% of business value can be lost due to poor data quality, while 29% of respondents said 50% or more of business value can be lost (Lehmann et al., 2016). Harvard Business Review presents $3.1 trillion as IBM's estimate of the yearly cost of poor quality data, in the US alone, in 2016 (Redman, 2016). These costs, however, are not solely financial; businesses can see a loss of reputation, missed opportunities and higher-risk decision-making as a result of low confidence in data (Forbes, 2017).

In the supply chain environment, two types of errors can occur in data received by a manufacturer from its retailers. First, errors may occur when data is transmitted from a retailer to a manufacturer. According to a study, retailers who are sharing data are most likely to be doing it via manual processes such as spreadsheets or flat files (CGT and RIS News, 2014). Such approach of information sharing can be “error-prone” and “time-consuming” (Brown, 2017). While some supply chain partners use Electronic Data Interchange (EDI) for information sharing, data errors can still happen when data are transferred from one system to another (Liaw & Cova, 2006). A study in 2018 shows that a significant majority of respondents - 82% - reported that they are currently experiencing issues with their EDI data and/or retailer portals (Askuity, 2018). After new products are introduced and the old ones are discontinued. This does not always get reflected in the system leading to a discrepancy between retailer and vendor records (EKN, 2016). This type of error affects the quality of data observed by the manufacturer but not the retailer, and we call it the Transmission Error. Errors may also occur in the data before it is transmitted from the retailer to the manufacturer (Informatica, 2006). Unrecorded sales transactions and duplication of sales transactions are two common issues with POS data of retail business (Singleton, 2016). In addition, inaccurate data capture will lead to a lack of trusted data which often holds back effective collaboration between retailers and manufacturers (McGovern et al., 2016). This type of error affects the quality of data observed by both the retailer and the manufacturer. We call this type of error the Source error.

Transmission errors and source errors are both prevalent in supply chains. We have conducted a simple survey with 70 respondents including mid-level managers and senior executives from some well-known Chinese companies. These companies cover the industries of Manufacturing, Wholesale and Retail, Transportation and Inventory, Information Technology Services, etc. More than half of respondents (52.86%) come from the Manufacturing Industry. 84.29% of respondents have worked for more than 5 years. Among the 70 respondents, around 67% agree that source error is a widespread phenomenon in supply chains while 63% perceive transmission error as a widespread phenomenon in supply chains. The average magnitudes of source and transmission error estimated by the respondents are around 14.57% of demand and 15.43% of demand, while the magnitude of demand uncertainty is estimated to be only 20.71% of demand. In other words, the magnitudes of source error and transmission error are estimated to be around 70% and 75% of demand uncertainty. The goal of our study is to shed light on how information errors affect supply chain information sharing decisions. In particular, is information sharing still valuable? How do these errors affect the value of information sharing? In addition, as the information shared by a retailer “does not come with extensive documentation or a 1–800 help-line number to call with questions” (The Grocery Manufacturers Association, 2009), how can a manufacturer make better use of the shared information for making decisions?

We consider a multi-period stochastic model with a manufacturer (her) selling to a retailer (him). In each period, the retailer places an order with the manufacturer, and the manufacturer places an order with her external supplier. Both the retailer and the manufacturer adopt an order-up-to policy and use the MMSE (minimum mean-squared error) technique when making ordering decisions. The retailer forecasts his replenishment lead-time demand (i.e., market demand), and the manufacturer forecasts her replenishment lead-time “demand”.

How the manufacturer forecasts her replenishment lead-time “demand” depends on the availability of information. When the retailer does not share demand information with the manufacturer, she can only use the retailer's historical orders to forecast his future orders. When the retailer shares demand information with the manufacturer, however, the latter can decide according to two strategies: one is using both the demand information and the retailer's ordering quantity in her decision-making process (called demand-order information sharing here), the other one is using only the demand information for making her ordering decisions (called demand-only information sharing here). The models in Chen and Lee (2009); Giloni et al. (2014); Kovtun et al. (2014); Lee et al. (2000); Leng and Parlar (2009) consider the case where the manufacturer uses both demand information and retailer's order quantity for making her ordering decisions, while the models by Chen et al. (2000a); Zhang and Cheung (2011); Zhang and Zhao (2010) assume that the manufacturer only uses demand information and disregards retailer's order quantity. A question may be noted is that why the manufacturer may ignore retailer's order information when forecasting. Bullwhip Effect occurs when supply chain members process the demand input from their immediate downstream member for generating their respective forecasts. One way to avoid this multiple demand forecast updates is “making demand data at a downstream site available to the upstream site” (Lee et al. (1997). Thus, when the manufacturer is aware of the customer demand deeply and completely through information sharing, she would only use the actual customer demand to estimate the mean lead-time demand. In this paper, we consider both types of information sharing (i.e., demand-order information sharing and demand-only information sharing) to identify the lowest forecasting error strategy for the manufacturer.

Our analysis leads to several interesting insights. First, information errors affect the manufacturer's optimal information sharing strategy. Without information errors, it is always optimal for the manufacturer to use all available information in making decisions when time period is finite. In the existence of information errors, it may be optimal for the manufacturer to disregard the retailer's order quantity and use only the demand information.

Our second main result is that information errors have a huge impact on the value of information sharing. Furthermore, transmission error and source error have significantly different impacts. While transmission error occurs in the process of information transmission from the retailer to the manufacturer, it has no impact on the retailer's ordering quantity. On the other hand, source error affects the retailer's ordering quantity as it also affects the retailer's forecast of future demand. One key consequence of this difference is that the values of both information sharing strategies may be negative with transmission error, however, only the value of demand-only information sharing may be negative with source error.

Our third main result shows the difference in the way the manufacturer uses available information. In particular, when the manufacturer uses both demand information and retailer's order quantity for forecasting the retailer's future order quantities, the impact of transmission error magnifies when retailer's replenishment lead-time increases. Yet, this magnification does not happen when the manufacturer disregards the retailer's current order quantity. As a result, when the retailer's replenishment lead-time is long and the magnitude of transmission error is small to moderate, the manufacturer's optimal strategy is to consider only demand information when making decisions.

The remainder of the paper is organized as follows. Section 2 reviews the literature. Section 3 introduces the model. Section 4 presents the two types of information sharing and analyzes the value of information sharing in the absence of information errors. Section 5 studies the impact of information errors. Finally, Section 6 provides concluding remarks.