Sage Journals: Discover world-class research

Abstract

The advancement of Internet of Things (IoT) technology has enabled IoT device manufacturers to collect consumer usage data (IoT data) whenever their devices are in use. Our paper examines the following setting: A manufacturer collects IoT data and decides whether to share it with a retailer, while consumers remain concerned about their privacy; the retailer, in turn, can leverage the shared data for cross-selling by investing in data-mining efforts that transform raw data into actionable insights. This aspect of data mining differentiates our study from traditional research on information sharing. Beyond selling through the existing retail channel, the manufacturer also has the option to establish a direct channel, thereby encroaching on the retailer’s market. Our analysis reveals several key insights. First, when the manufacturer both encroaches and shares IoT data, we observe a counterintuitive positive effect of the channel substitution rate: As the substitution rate increases, both the manufacturer and the retailer may see higher profits. Second, while the manufacturer always chooses to encroach when data sharing is absent, its motivation to do so weakens when it shares IoT data. Finally, we find that an increase in the value of IoT data can unexpectedly lead to a decline in the retailer’s profit.

Keywords

Internet of Things Technology Channel Encroachment Data Sharing Data Mining Game Theory

1. Introduction

1.1. Study of Current Practice and Context

With widespread e-commerce technology, many manufacturers not only sell their products on a retail channel but can also establish their own direct channels. For example, Apple sells its products on Amazon and its own website; Huawei sells its products on JD.com and through its own established online market. The phenomenon of a manufacturer establishing a direct channel to sell its product in addition to selling on an existing retail channel is called encroachment. A manufacturer can use encroachment to increase the selling quantity, but encroachment intensifies the competition faced by the retail channel and affects the profits of both the manufacturer and the retailer (Huang et al., 2018).

In practice, there are instances of manufacturers opting to distribute their products exclusively through retailers rather than direct channels. Two notable examples illustrate this approach:

Lennox Home Comfort Systems: These products are available for purchase from retail outlets like Costco.¹ However, Lennox’s official website² serves only as an information resource without offering direct sales to consumers.

Zigbang Smart Door Locks: This manufacturer’s official website provides product details and retailer information, but does not offer direct sales to consumers. Instead, Zigbang’s products are sold through authorized retailers and distributors.³

These examples demonstrate how some manufacturers choose to leverage existing retail networks for product distribution, focusing their own efforts on product development and brand management rather than direct-to-consumer sales.

The previously mentioned Lennox home comfort systems and Zigbang smart door locks are two examples of Internet of Things (IoT) devices. Such devices are internet-enabled and use sensors to collect usage data. Overall enterprise IoT spending grew 22.4% in 2021 to $158 billion, and the IoT market size is expected to grow at a compound annual growth rate of 22.0% to $525 billion from 2022 until 2027 (Wegner, 2022). In addition, the number of IoT devices worldwide is forecast to triple from 9.7 billion in 2020 to more than 29 billion IoT devices in 2030 (Statistica Inc, 2022). IoT has become one of the most important technologies of the 21st century and created high market value. With the vast amount of data that IoT devices can capture, it becomes important to examine the impact of IoT data on supply chains.

As a disruptive technology, IoT technology has changed how supply chain members interact. In a traditional supply chain, a manufacturer sells products to a retailer, and the retailer sells the products to customers. In this scenario, the manufacturer’s role becomes minimal (if not non-existent, once the product is sold to the retailer). IoT technology enables manufacturers to gather valuable customer usage data from connected products and possibly share it with retailers, as shown in Samsung’s and Lennox’s privacy policies (Lennox, 2021; Samsung, 2025). To reduce privacy concern, a manufacturer can modify the raw data before sharing to hide sensitive information (Menon et al., 2022); the manufacturer and the retailer can also use federated learning technology (Di et al., 2025; Qi et al., 2025) to improve the retailer’s machine learning model without transferring raw data to the retailer.

By strategically sharing this IoT data, manufacturers empower retailers to gain deeper insights into customer behavior, facilitating personalized product and service recommendations—a practice known as cross-selling. This data-driven approach leads to more precise recommendations and opportunities to introduce customers to additional products. In the case of Lennox’s home comfort systems, a smart Wi-Fi thermostat paired with a furnace can transmit usage data to Lennox’s servers. This data, if shared with retailers, can significantly enhance targeted marketing efforts. For example, if the data shows frequent furnace operation, it may indicate inadequate home insulation. Retailers could then recommend insulation upgrades or energy-efficient windows to improve energy efficiency. High energy consumption readings might prompt retailers to suggest eco-friendly alternatives such as solar panels. Furthermore, if the data reveals high humidity levels, retailers could propose dehumidifiers to enhance home comfort. Another illustration is Samsung’s FamilyHub smart refrigerator,⁴ which employs artificial intelligence (AI)-powered features such as AI Vision Inside to identify and monitor items within the fridge. This technology enables retailers to offer more targeted product recommendations based on shared data. For example, if a consumer frequently purchases yogurt, the retailer might suggest exploring new brands or flavors.

In some ways, IoT data is similar to cookie data and consumer purchasing history data since they can all potentially be used to understand consumer behavior. However, there are several major differences. First, the original data collectors are different. Typically, when a customer browses a retailer’s website, the retailer can collect data via cookies without consumers even buying products. Similarly, the retailer also has consumers’ purchasing history when consumers buy products from the retailer. However, it is the manufacturer that collects IoT data via IoT devices. Then, the manufacturer will decide whether to share the IoT data jointly with its other decisions. This data-sharing decision is the main component of our model. The second difference is that IoT data is generated from consumers who have purchased IoT devices. To gather more IoT data, a retailer will be motivated to sell more IoT devices since the amount of IoT data is directly proportional to the number of IoT devices sold. In contrast, cookie data can be generated before a purchase occurs, and purchasing history data are generated from a variety of products. In other words, cookie and purchasing history do not have such a strong dependence on the sale of a particular product. Our model captures the direct relationship between IoT data value and the sale of IoT devices.

1.2. Research Questions and Key Findings

In the IoT product market, some manufacturers, like Lennox and Zigbang, operate without direct channels, while others, such as Huawei and Apple, do. This contrast highlights the ongoing debate over whether manufacturers should encroach by establishing direct channels. Previous research has shown that the manufacturer’s encroachment decision can be affected by various types of information sharing. However, how data sharing enabled by IoT products affects a manufacturer’s encroachment decision is not clear. So, we raise our first research question: How will the manufacturer’s encroachment decision interact with the IoT data-sharing? Our research reveals that the answer to this question is nuanced and complex. The retailer’s strategy for data mining can significantly impact the manufacturer’s decision to encroach. Specifically, we’ve identified certain scenarios where the retailer can strategically modify its data mining efforts in a manner that reduces the attractiveness or feasibility of encroachment for the manufacturer, effectively deterring such competitive moves. The result is due to the nature of co-opetition between the manufacturer and the retailer: They compete between the retail and the direct channels and cooperate within the retail channel through IoT data sharing. Such cooperation due to IoT data sharing leads to an equilibrium where the manufacturer stops channel encroachment; without IoT data sharing, the manufacturer always chooses to encroach.

After investigating the impact of IoT data sharing on a manufacturer’s encroachment decision, we then study the impact of IoT data value on the manufacturer’s decisions and profits. This leads to the second research question: How does IoT data value affect supply chain members’ decisions and profits? Our paper shows that an increase in IoT data value has a positive effect on the manufacturer, but it may hurt the retailer. The reason is that the manufacturer’s strategy may change from no-encroachment to encroachment to exploit an increase in data mining efficiency.

When a manufacturer has already established its direct channel, it is not clear how competition between the two channels affects the manufacturer’s and retailer’s profits. Using channel substitutability as a proxy for the intensity of competition, we pose our third research question: How does channel substitutability affect the profits of supply chain members? We show that when the retailer can exert data mining effort, the impact of channel substitutability is not very straightforward. An increase in the substitution rate can have either a positive or negative impact on the profits of the manufacturer and/or the retailer, depending on the cost of data mining effort. For example, due to co-optition, when the retailer is highly effective in data mining, a higher substitution rate leads to more cooperation and higher profits for both players.

Compared with traditional supply chain models, the new features we incorporate into our model are that (1) The manufacturer collects IoT data after the IoT devices are sold; (2) the retailer must exert effort to mine the IoT data to generate cross-selling; (3) customers have privacy concerns when data is shared; and (4) the manufacturer makes a joint decision on channel encroachment and IoT data sharing. Our work has both theoretical and practical contributions. We contribute theoretically in two ways. First, despite the importance of IoT technology to the economy and the rich literature on channel encroachment, no paper has investigated the impact of IoT data sharing on channel encroachment. Our work fills this gap. Second, our work illuminates the sale of high-tech products in the IoT environment. We show how manufacturers and retailers modify their selling decisions to leverage IoT data sharing. Practically, our results can help policymakers who are concerned with improving the overall supply chain profit. They should make comparison shopping and channel switching easier for customers.

The rest of this study is organized as follows. We review the related literature in Section 2. In Section 3, we build a game-theoretical model with an upstream manufacturer, a downstream retailer, and a continuum of customers. In Section 4, we identify the equilibria in the no-encroachment and encroachment cases and further characterize the subgame perfect equilibria. In Section 5, we check the robustness of the results by considering several extensions. We conclude the paper in Section 6.

2. Literature Review

Our paper relates to the following two streams of literature: Data sharing and monetization, and channel encroachment.

2.1. Data Sharing and Monetization

We further classify the literature on data sharing and monetizing into three categories. The first category addresses privacy and trust issues in data sharing. Li and Qin (2017) examine how to resolve the patient privacy concern in healthcare data sharing. They propose a systematic approach to anonymize medical text records and show their approach’s effectiveness with an experimental study using real-world medical documents. Ghoshal et al. (2020) investigate an overlooked problem: Differences in customer characteristics across regions lead to sensitive information. They propose an ensemble approach and show that it can hide sensitive information with negligible negative impact. Wang et al. (2021) study how to overcome the distrust, privacy concern, data misuse, and asymmetric valuation of shared data problems faced in data sharing. They show that blockchain can be used to overcome the issue by designing and implementing a blockchain-enabled data-sharing marketplace for a stylized supply chain. Li et al. (2023) show the risk of re-identification of panelists in marketing research data. To decrease the risk, they develop a graph-based minimum movement based on the $k$ -anonymity approach.

The second category focuses on the impact of data sharing. Croson and Donohue (2003) investigate how point of sale (POS) data sharing affects order decisions in the supply chain, finding that POS data sharing reduces the bullwhip effect in a stable demand setting. Ghoshal et al. (2020) investigate whether a non-personalizing firm shares its data with a personalizing firm when the latter makes product recommendations and the former does not. They find that the non-personalizing firm is always willing to share its data, while the personalizing firm is willing to use the shared data only when the learning rate influence dominates the profile influence. Choe et al. (2024) study the impact of a data-rich firm’s unilaterally sharing its customer data with a data-poor competitor. They find that the firm’s data sharing can soften the price competition.

The third category focuses on monetizing data. Bimpikis et al. (2019) analyzes how a monopolistic seller selling information to competing downstream buyers can determine the optimal strategy, considering the nature and intensity of competition. Mehta et al. (2021) examine a data seller’s pricing mechanism and obtain the price-quantity schedule with a good worst-case guarantee compared with the optimal mechanism.

Our paper differs from previous literature because it is the first study investigating a manufacturer’s IoT data sharing decision. IoT data is not collected by the seller but by the manufacturer directly. The amount of IoT data collected is endogenous and depends on the consumer demand in our setting. In addition, this paper considers the impact of IoT data sharing on a manufacturer’s channel encroachment decisions, which has not been studied before.

2.2. Channel Encroachment and Information Sharing

Our paper is also closely related to channel encroachment literature, in which a manufacturer (or supplier) decides whether to encroach. Since an early channel encroachment work by Arya et al. (2007) that shows a retailer can benefit from supplier encroachment, many papers have discussed what factors affect the supplier’s encroachment decision, such as nonlinear pricing (Li et al., 2015), quality differentiation (Ha et al., 2016), strategic inventory withholding (Guan et al., 2019), and retailer’s service effort (Ha et al., 2022). In addition, the retailer may welcome the manufacturer’s encroachment for the spillover effect of the manufacturer’s cost-reduction investment (Yoon, 2016).

Prior research has shown that a manufacturer’s decision to establish a direct channel can be shaped by different forms of information sharing, such as demand information sharing (Chen and Özer, 2019; Li et al., 2021; Liu et al., 2021; Shamir and Shin, 2018), inventory information sharing (Zhang, 2006), production yield information sharing (Choi et al., 2008; Gao et al., 2014), client type information sharing (Zhao et al., 2014), and production efficiency information sharing (Chen and Deng, 2015). Building on this stream of work, our paper examines a new form of information sharing—IoT data sharing— and its implications for channel encroachment. The potential link between information sharing and encroachment can be illustrated by the case of Lennox. As mentioned earlier, the company does not operate a direct sales channel but instead distributes its IoT products through third-party dealers. According to its online privacy statement, Lennox collects usage data from its IoT products and may share these data with third-party dealers. Such sharing could plausibly influence Lennox’s decision to refrain from encroachment. There is one key difference between traditional information and IoT data. Traditional information, such as demand or inventory information, can be utilized directly, but IoT data is more granular and can produce value only after it has been mined using various data mining algorithms. A key distinction between traditional information and IoT data is that the former (e.g., demand or inventory information) can typically be used directly, whereas IoT data is highly granular and only becomes valuable once processed through data mining algorithms.

Our research focuses on how IoT data sharing affects the manufacturer’s encroachment decision. To utilize IoT data, the retailer needs to exert data-mining effort. Also, we endogenize the amount of IoT data by assuming it is proportional to the quantity of IoT products sold in our model. This aspect of IoT data sharing has not been studied before. We obtain interesting results in this study. For example, analysis shows that IoT data sharing may alleviate the manufacturer’s incentive to encroach.

3. Model

We consider a multi-stage model with an upstream manufacturer and a downstream retailer. Table 1 summarizes the notation used in this paper.

Table 1.
Summary of notation

Notation Description

Parameters

$d$ Potential market size of IoT product

$c$ Marginal production cost of IoT product

$μ$ Consumer privacy cost of data sharing

$b$ Channel substitution rate, $b \in (0, 1)$

$k$ Data mining development cost coefficient

$σ$ Marginal data value of cross-selling from retail channel

Intermediate variables

$q_{m}$ Product selling quantity from direct channel

$q_{r}$ Product selling quantity from retail channel

$π_{m}$ Manufacturer’s profit

$π_{r}$ Retailer’s profit

Decision variables

$e$ Retailer’s data-mining effort

$p_{m}$ Price of the product from direct channel

$p_{r}$ Price of the product from retail channel

$w$ Wholesale price for the product

$J$ The joint decision of data sharing and utilization by the manufacturer and the retailer, $J \in {0, 1}$

Superscripts

$e k$ This superscript contains two indices. For the first index $e \in {N, E}$ , $N$ indicates no encroachment and $E$ indicates encroachment; for the second index $k \in {N, S}$ , $N$ indicates no data sharing, and $S$ indicates data sharing

Notation	Description
Parameters
$d$	Potential market size of IoT product
$c$	Marginal production cost of IoT product
$μ$	Consumer privacy cost of data sharing
$b$	Channel substitution rate, $b \in (0, 1)$
$k$	Data mining development cost coefficient
$σ$	Marginal data value of cross-selling from retail channel
Intermediate variables
$q_{m}$	Product selling quantity from direct channel
$q_{r}$	Product selling quantity from retail channel
$π_{m}$	Manufacturer’s profit
$π_{r}$	Retailer’s profit
Decision variables
$e$	Retailer’s data-mining effort
$p_{m}$	Price of the product from direct channel
$p_{r}$	Price of the product from retail channel
$w$	Wholesale price for the product
$J$	The joint decision of data sharing and utilization by the manufacturer and the retailer, $J \in {0, 1}$
Superscripts
$e k$	This superscript contains two indices. For the first index $e \in {N, E}$ , $N$ indicates no encroachment and $E$ indicates encroachment; for the second index $k \in {N, S}$ , $N$ indicates no data sharing, and $S$ indicates data sharing

3.1. The Manufacturer

The manufacturer produces an IoT product that gathers consumer usage data, such as physical exercise data, through sensors on the product. Applications can use the data to provide customers with various value-added services by utilizing the manufacturer’s data analytic tools through an application programing interface. For example, Huawei’s health kit provides a data platform for device developers to store and share activity and health data (Huawei, 2022). Also, IoT device manufacturers such as Samsung and Xiaomi have similar cloud storage to gather and store data (Samsung, 2022; Xiaomi, 2020). After the manufacturer collects the customers’ usage data via IoT devices, it can share the data with the retailer for free. In Appendix D.1, we discuss the case where the manufacturer sells IoT data to the retailer, but we find that the two cases are equivalent. In this paper, we use $J$ to indicate whether the manufacturer shares data and the retailer utilizes it. When the manufacturer shares data and the retailer utilizes it, $J = 1$ . Otherwise, if the manufacturer decides to share the data and the retailer rejects utilizing the data, or the manufacturer decides not to share the data, then $J = 0$ .

In the case of $J = 1$ , consumers face a privacy cost if they become aware of data sharing. Consumers can discover data sharing through several channels. Retailers’ personalized recommendations based on IoT data, such as furnace usage or refrigerator stock levels, can indicate data sharing. Consumers can also directly ask manufacturers about their data sharing practices or check company privacy policies, such as Lennox’s, which allows California residents to request “a description of whether we have disclosed their Personal Information to third parties $\dots$ .” Furthermore, regulations such as the General Data Protection Regulation (European Parliament and Council of the European Union, 2016) and the Personal Information Protection Law of the People’s Republic of China (China’s National People’s Congress, 2021) require manufacturers to disclose data sharing practices, increasing transparency and consumer awareness.

In the existing retail channel, the manufacturer sells an IoT product to the retailer through a wholesale contract, and the retailer, in turn, sells the product to customers. In addition, the manufacturer can establish a direct channel and sell the product directly to customers if doing so is profitable.

3.2. The Retailer

In the data-sharing case where the manufacturer shares data and the retailer utilizes it, the retailer can invest in data mining and benefit from cross-selling. That is, by utilizing the data, the retailer can more accurately recommend related products and services and gain higher revenue from the retail channel. Since the retailer does not have the contact information of customers in the direct channel, it cannot cross-sell other products to these customers. Then, the manufacturer may share the consumers’ IoT usage data in the retail channel, but not in the direct channel. Consequently, consumers in the retail channel will have privacy concerns about data being shared, while their counterparts in the direct channel will not, and the demand in the direct channel (see Equation (4)) is not affected by the privacy concern parameter $μ$ .

3.3. Model Without Encroachment

When the manufacturer does not encroach (i.e., it does not set up its own channel), it only sells the product through the existing retail channel using a wholesale contract. The inverse demand function of the retail channel is

p_{r} = d - J μ - q_{r} .

(1)

The parameter

d

is the market potential for the IoT product. Parameter

μ

is the market potential decrease due to consumer privacy concerns caused by the manufacturer’s data sharing. In Appendix D.2, we extend the result to the case that the manufacturer can determine the amount of data to share. We find that the manufacturer will either share all data or share no data. We further explore a scenario in Appendix D.7 where only a subset of consumers choose to share their data.

Let $w$ be the wholesale price the manufacturer charges the retailer. The manufacturer’s and the retailer’s profit functions are given by

π_{m} = (w - c) q_{r}

(2)

and

\begin{aligned} π_{r} = (p_{r} - w) q_{r} + J (e σ q_{r} - k e^{2}) . \end{aligned}

(3)

In Equation (2), the parameter

c

is the marginal cost of producing an IoT product. In Equation (3), the term

e σ q_{r}

represents the additional revenue from cross-selling using IoT data. It is proportional to the amount of consumer data (represented by the number of IoT devices sold). The parameter

σ

is the cross-selling coefficient from the retail channel when the manufacturer shares data. The decision variable

e

represents the retailer’s data-mining effort, and

k e^{2}

is the cost of the data-mining effort. We use a linear-quadratic framework to capture the idea that the cross-selling value cannot be infinite. Such a framework allows us to analytically obtain an interior solution of data-mining effort

e

and has also been used to model the impact of IT investment on product quality and production cost (Demirhan et al., 2007; Xin and Choudhary, 2019). We use the multiplicative form

e σ q_{r}

to represent the value of cross-selling, which implies that when there is more data

q_{r}

, the retailer benefits more from cross-selling. Additionally, if the retailer invests more effort in data mining to understand consumer preferences, it can gain more from cross-selling. However, there is a decreasing marginal return for data-mining effort. The reason is that it becomes more difficult to get additional value out of the same dataset when the retailer has spent enough effort. When

J = 1

, the retailer undertakes the data-mining effort and benefits from cross-selling; when

J = 0

, the retailer does not utilize the data for cross-selling and only earns revenue from selling the product.

3.4. Model with Encroachment

In the case of manufacturer encroachment, the manufacturer establishes its own direct channel and sells through both direct and retail channels. Following the previous works that consider competition between a manufacturer and a retailer (Guo et al., 2014; Ha et al., 2022), we assume the manufacturer and the retailer engage in price competition in the end market for customers. In Appendix D.3, we extend our model to the quantity competition case, and the main results continue to hold. Let $q_{m}$ and $q_{r}$ denote the product selling quantities in the direct and retail channels, respectively. For simplicity, we present the inverse demand functions in the direct and retail channels as follows (for completeness, we show the demand functions in Equations (E.45) and (E.46) in the appendix).

p_{m} = d - q_{m} - b q_{r}

(4)

and

\begin{aligned} p_{r} = d - J μ - q_{r} - b q_{m} . \end{aligned}

(5)

where the parameter

b

(

b < 1

) is the substitution rate between the two channels. It also represents the price competition between these two channels. Intuitively, when

b

increases, there is higher substitution between these two channels, leading to more intensive price competition. In Appendix C, we show that this inverse demand function can be derived from the utility function.

The manufacturer’s profit function is given by

π_{m} = q_{m} (p_{m} - c) + (w - c) q_{r},

(6)

which consists of two parts: Profit from the direct channel, that is,

(p_{m} - c) q_{m}

, and profit from the retail channel, that is,

(w - c) q_{r}

. We do not consider the manufacturer’s cross-selling value in the manufacturer’s profit function. Unlike retailers that sell thousands of products, manufacturers do not sell other products (at most they only sell a few types of products), so the manufacturer’s cross-selling value is negligible compared with the retailer’s. The retailer’s profit function is

π_{r} = q_{r} (p_{r} - w) + J (e σ q_{r} - k e^{2}),

(7)

which consists of three parts, profit from selling the product

q_{r} (p_{r} - w)

, profit from cross-selling by exploiting consumer usage data

e σ q_{r}

, and the cost of data-mining effort

k e^{2}

. Note that in our model, we assume the retailer can not exploit the data from consumers in the direct channel when the manufacturer only shares IoT device usage data. The reason is that the retailer does not know the contact information of customers in the direct channel and can not cross-sell by recommending other products to these customers.

3.5. Time Sequence

The time sequence of the game is depicted in Figure 1. In Stage 1, the manufacturer first decides whether to encroach and then decides whether to share IoT data with the retailer. Since both decisions are determined by the manufacturer, the simultaneous decisions are equivalent to sequential decisions. For simplicity, we present these decisions sequentially and let the manufacturer decide whether to encroach and then whether to share data. In Stage 2, the retailer decides whether to utilize the IoT data. If it utilizes the data, it decides how much effort to exert in data mining. The data-mining effort is zero if it does not utilize the data. In Stage 3, the manufacturer sets the wholesale price $w$ . In Stage 4, the retailer decides $p_{r}$ in the retail channel, and the manufacturer simultaneously decides $p_{m}$ in the direct channel if the manufacturer encroaches. Note that the data-mining effort is decided before the wholesale price. Once the manufacturer chooses to share data, the retailer must plan and acquire data mining resources in advance. This is a reasonable model assumption since the wholesale price is easy to adjust, while data mining resources are difficult to change.

Figure 1.

Time sequence of the game.

4. Analysis

To study the manufacturer’s decision of whether to encroach and share data, we first investigate the data-sharing decision with and without channel encroachment in Subsections 4.1 and 4.2, respectively. Then, we obtain the subgame perfect equilibria in Subsection 4.3 and contrast data sharing with information sharing in Subection 4.4.

4.1. Data-sharing Decision Without Encroachment

We solve the equilibrium in the case without encroachment, that is, $N N$ and $N S$ cases, with backward induction. In the $N N$ case, there is no data sharing and therefore no data-mining effort. We show the equilibrium of the $N N$ case in Appendix E.1 (all proofs are available in the E-Companion). In this case, the manufacturer’s profit is always greater than the retailer’s profit, and the wholesale price is always less than the retail price.

We then proceed to the IoT data-sharing without encroachment case, that is, the $N S$ case. In this case, the retailer decides the data-mining effort, and then the wholesale and retail prices are determined sequentially. We summarize the result in Appendix E.2. In the $N S$ case, we find that the retail price could be smaller than the wholesale price, that is, $p_{r}^{N S} < w^{N S}$ , in certain situations. This result differs from that in the $N N$ case, where the retail price is always greater than the wholesale price. The reason is that the retailer could lower its retail price to less than the wholesale price, thereby increasing the selling quantity and gaining high cross-selling value. This result is consistent with pricing strategies observed in practice. For example, Tmall Genie is an IoT product sold by Alibaba at a retail price lower than its wholesale price (Liangshuo Technology, 2023).

Through comparative statics, we can show how IoT data value affects the firms’ profits and decisions in the $N S$ case in the following proposition. IoT data value is defined to increase when either the cost coefficient of data-mining effort $k$ decreases, the privacy concern $μ$ decreases, or the cross-selling value $σ$ increases.

Proposition 1
(Impact of IoT data value in $N S$ case) In the $N S$ case, when the IoT data value increases ( $k$ or $μ$ decreases, or $σ$ increases),
the manufacturer’s and the retailer’s profits ( $π_{r}^{N S}$ and $π_{m}^{N S}$ ), the wholesale price ( $w^{N S}$ ), and the data-mining effort ( $e^{N S}$ ) all increase,

the retail price could increase or decrease. Specifically, in the region where the data-mining effort is costly, the retail price increases as privacy concern $μ$ decreases, that is, $\frac{\partial p_{r}^{N S}}{\partial μ} < 0$ if $k > \frac{σ^{2}}{12}$ .

From Proposition 1, we can see that both the manufacturer’s and the retailer’s profits increase with the cross-selling value coefficient $σ$ . However, what is interesting is that the ratio of the manufacturer’s profit over the retailer’s profit increases with $σ$ , as we can show from (E.12) and (E.13). This means that, as cross-selling becomes more valuable to the retailer, the manufacturer’s share of gain from cross-selling increases while the retailer’s share decreases. Such a counterintuitive result can be explained as follows. As $σ$ increases, each consumer becomes more valuable to the retailer for cross-selling. Then, the retailer lowers the retail price to attract more consumers for cross-selling. At the same time, it also invests more in data mining. However, the manufacturer can exploit this situation by increasing the wholesale price without bearing the cost of data mining. As a result of this free-riding, the manufacturer gains a larger share of the “pie,” the total channel profit. This result, due to data mining effort in our model, contrasts with that in a traditional information-sharing setting where each player’s relative share of the channel profit does not change with information value (Ha et al., 2022; Huang et al., 2018).⁵

Also, from Proposition 1, we find that the retailer could drop the retail price $p_{r}^{N S}$ to attract more consumers as consumer data becomes more valuable. However, the impact of privacy concerns on the retail price is not straightforward: In the region where the data-mining effort is costly ( $k$ is large), the retailer actually increases the retail price when consumer privacy concern decreases (which should mean that cross-selling by using consumer data becomes more attractive), contrary to what one might have expected. This can be explained as follows. The privacy concern has two effects on the retail price: direct and indirect. When consumers become more concerned about privacy due to data sharing by the manufacturer, their utility decreases, putting downward pressure on the retail price. That is, the privacy concern has a negative direct effect on the retail price: Increased privacy concern tends to cause the retail price to decrease. At the same time, as fewer consumers buy the product due to the negative direct effect, the cross-selling value decreases. Then, the data-mining effort decreases, further decreasing the cross-selling value. Subsequently, the cross-selling revenue becomes less important to the retailer. As a result, the retailer does not have to decrease its retail price to attract more consumers. That is, the privacy concern has a positive indirect effect on the retail price: An increase in privacy concern tends to cause the retail price to increase. Our result shows that the net effect of privacy concern on retail price depends on the cost parameter of data-mining effort $k$ . When the data-mining effort is costly, the initial data-mining effort is small, implying that the retailer can not reduce the data-mining effort by much. Consequently, the negative direct effect dominates the positive indirect effect, and the retail price decreases with the privacy concern.

Next, we determine whether the manufacturer shares data at equilibrium in the no-encroachment case. Since the retailer can reject utilizing the IoT data shared by the manufacturer, we need to consider both the manufacturer’s and the retailer’s incentives. We summarize the findings in the following lemma.
Lemma 1
(Data-sharing decision in the no-encroachment case) When the manufacturer does not encroach, the manufacturer is willing to share data and the retailer is willing to utilize shared data when the privacy concern is small. Specifically,
when $μ < μ_{r 1}$ , $π_{m}^{N N} < π_{m}^{N S}$ and $π_{r}^{N N} < π_{r}^{N S}$ ;

when $μ_{r 1} < μ < μ_{m 1}$ , $π_{m}^{N N} < π_{m}^{N S}$ and $π_{r}^{N N} > π_{r}^{N S}$ ;

when $μ > μ_{m 1}$ , $π_{m}^{N N} > π_{m}^{N S}$ and $π_{r}^{N N} > π_{r}^{N S}$ .
In addition, $μ_{r 1} < μ_{m 1}$ and $μ_{r 1}$ is a decreasing function of $k$ . The expressions for $μ_{m 1}$ and $μ_{r 1}$ are given in Appendix E.4.

If the marginal cost from consumer privacy concerns is small, both the manufacturer and the retailer can benefit from data sharing; then, data sharing can be achieved. When the marginal cost from consumer privacy concerns is large, neither the manufacturer nor the retailer can benefit from data sharing, so there is no data sharing in this case. When the marginal cost from consumer privacy concerns is moderate, the manufacturer can benefit from data sharing, but the retailer is worse off for utilizing that data. Since the manufacturer does not share data when the retailer does not utilize the data, there is no data sharing in this case either. The intuition is that the retailer incurs the cost of data-mining effort, so even when both the manufacturer and the retailer benefit from cross-selling, the retailer’s cost of data-mining effort dominates the benefit from cross-selling, and the retailer does not want to utilize the data for cross-selling. This result shows that the manufacturer’s and the retailer’s incentives are not aligned regarding data sharing.

Next, we will explore the manufacturer’s IoT data-sharing decision in the encroachment case.
4.2. Data-sharing Decision with Encroachment

We first show how the value of consumer data affects the manufacturer’s and the retailer’s profits and decisions in the $E S$ case (The equilibria of both $E N$ and $E S$ cases are provided in Appendix E.5 and E.6). The result is presented in Proposition 2.

Proposition 2
(Impact of IoT data value in $E S$ case) In the $E S$ case, when the value of consumer data increases (either $k$ or $μ$ decreases, or $σ$ increases),
the manufacturer’s and the retailer’s profits ( $π_{m}^{E S}$ and $π_{r}^{E S}$ ), the wholesale price ( $w^{E S}$ ), the data-mining effort ( $e^{E S}$ ), and the retail price in the direct channel ( $p_{m}^{E S}$ ) all increase.

the retail price in the retail channel ( $p_{r}^{E S}$ ) could increase or decrease. Specifically, it decreases with privacy concerns in regions where data-mining effort is not so costly. That is, when $k > \frac{(b^{2} + 2)^{2} σ^{2}}{(1 - b^{2}) (b^{2} + 6) (b^{2} + 8)}$ , $\frac{\partial p_{r}^{N S}}{\partial μ} < 0$ .

In the $E S$ case, in contrast to the $N S$ case, the manufacturer needs to determine the retail price in the direct channel $p_{m}^{E S}$ . Proposition 2 shows that $p_{m}^{E S}$ increases with IoT data value. The intuition is that when the IoT data value increases, cross-selling becomes more important to the retailer. Then, the manufacturer can indirectly benefit from cross-selling by charging a higher wholesale price. The manufacturer can also increase the retail price $p_{m}^{E S}$ in the direct channel so that more products will be sold in the retail channel. As a result, there is less price competition between the two channels. The other results in the $E S$ case presented in Proposition 2 resemble those in the $N S$ case shown in Proposition 1.

Using Lemma E.4, we can compare retail prices in the retail and the direct channels to get the following result.
Lemma 2
When consumer privacy and data-mining effort costs are small, the retail price in the retail channel is less than that in the direct channel. That is, when $μ < μ_{0}$ and $k < k_{0}$ , we have $p_{r}^{E S} < p_{m}^{E S}$ . The expressions for $μ_{0}$ and $k_{0}$ are given in Appendix E.8.

From Lemma 2, we find that the retail price $p_{r}^{E S}$ in the retail channel is less than $p_{m}^{E S}$ in the direct channel when $μ$ and $k$ are small. In this case, consumer privacy concerns are small, and the initial amount of data-mining effort is large due to the low-cost coefficient $k$ . Then, the retailer has an incentive to greatly lower its retail price and gain from high cross-selling value, causing the retail price in the retail channel to be smaller than that in the direct channel.

We then show how the substitution rate $b$ affects the firms’ profits. We summarize the results of $E N$ and $E S$ cases as follows.
Proposition 3
In the EN case, the profits of both the manufacturer and the retailer decrease with the substitution rate. In the ES case, although both firms’ profits still decrease with the substitution rate when data mining is costly, their profits increase with the substitution rate when the cost of data-mining effort is small. That is, in the $E N$ case, $\frac{\partial π_{r}^{E N}}{\partial b} < 0$ and $\frac{\partial π_{m}^{E N}}{\partial b} < 0$ . In the $E S$ case, $\frac{\partial π_{r}^{E S}}{\partial b} > 0$ for $k < k_{b 2}$ and $\frac{\partial π_{r}^{E S}}{\partial b} < 0$ for $k > k_{b 2}$ ; $\frac{\partial π_{m}^{E S}}{\partial b} > 0$ for $k < k_{b 1}$ and $\frac{\partial π_{m}^{E S}}{\partial b} < 0$ for $k > k_{b 1}$ . In addition, $k_{b 2} < k_{b 1}$ . The expressions for $k_{b 1}$ and $k_{b 2}$ are given in Appendix E.9.

From Proposition 3, in the $E N$ case ( $J = 0$ ), we can see that both firms’ profits decrease when the substitution rate $b$ increases. We can expect this from common intuition: Higher substitution leads to more intensive competition and hurts both parties. However, the results in the $E S$ case ( $J = 1$ ) are not so straightforward. From Proposition 3 and Figure 2, both the manufacturer’s and the retailer’s profits decrease with $b$ in Region III where $k$ is large, as expected. In Region I, where $k$ is small enough, that is, $k < k_{b 2}$ , we find that both the manufacturer’s and the retailer’s profits increase with $b$ . This somewhat counter-intuitive result can be explained as follows: As Proposition 2 demonstrates, data-mining effort increases with the reduction in the cost of data mining. This provides both the manufacturer and the retailer the potential to gain from higher cross-selling value if the relationship between these two players shifts away from competition toward cooperation. Cooperation occurs when, in response to an increase in the substitution rate, the retailer decreases the retail price and increases the data-mining effort to enhance the cross-selling value (see Table E2 in Appendix E.9). With the increased marginal benefit of cross-selling, the manufacturer is then motivated to increase the price in the direct channel to drive consumer demand toward the retail channel. The retailer can benefit from the increased cross-selling value, while the manufacturer can also indirectly gain by setting a higher wholesale price. With low k, the retailer’s cross-selling value will be high, allowing the manufacturer to raise the wholesale price. Then the manufacturer’s profit gain in the retail channel dominates the loss in the direct channel. So both firms’ profits increase. Consequently, the channel profit (the sum of the manufacturer’s and the retailer’s profits) also increases. In Region II, $k$ is moderate, and the data-mining effort is more costly than in Region I. When the substitution rate increases, even though the manufacturer still raises the price in the direct channel, the retailer will not benefit much from cross-selling. In contrast, the manufacturer can still gain from a higher wholesale price. Ultimately, the retailer is worse off, while the manufacturer is better off when $b$ increases.

Figure 2.
Impacts of $b$ on profits ( $σ = 3$ , $c = 1$ , $d = 3$ , $k = 1$ , $μ = 1$ ).

To summarize, in the $E S$ case, our study discovers a novel, positive effect of the substitution rate: A higher substitution rate can benefit the manufacturer and the retailer simultaneously (see Region $A$ in Figure 2). This surprising result differs from the common intuition that a higher substitution rate hurts both firms’ profits, as we have seen in the $E N$ case. The proposition demonstrates that IoT data sharing can foster greater cooperation between manufacturers and retailers. When the IoT data is shared, and the data-mining effort is not costly, An increase in the substitution rate aligns both parties’ interests.

In the previous works on manufacturer encroachment (Huang et al., 2018; Niu et al., 2019; Yoon, 2016), the substitution rate always negatively affects the manufacturer’s and retailer’s profits. However, our result shows that the manufacturer and the retailer could benefit from a higher substitution rate when the manufacturer encroaches and shares IoT data. This result has an important practical implication. Policymakers, such as those involved in industrial alliances, concerned with improving overall supply chain profit, should offer incentives to increase the substitution rate by making it easier for customers to switch between channels. For example, policymakers could promote information transparency, allowing customers to compare the two channels more easily before deciding which one to use for their purchase.

Comparing the manufacturer’s and the retailer’s profits in the encroachment cases with data sharing ( $E S$ ) and without data sharing ( $E N$ ), we can obtain the manufacturer’s and the retailer’s optimal decisions regarding data sharing, as stated in the following lemma.
Lemma 3
(Data-sharing decision in encroachment case) When the manufacturer encroaches, the manufacturer is willing to share data and the retailer is willing to utilize the shared data when the privacy concern is small. Specifically,
when $μ < μ_{r 2}$ , we have $π_{m}^{E N} < π_{m}^{E S}$ and $π_{r}^{E N} < π_{r}^{E S}$ ;

when $μ_{r 2} < μ < μ_{m 2}$ , we have $π_{m}^{E N} < π_{m}^{E S}$ and $π_{r}^{E N} > π_{r}^{E S}$ ;

when $μ > μ_{m 2}$ , we have $π_{m}^{E N} > π_{m}^{E S}$ and $π_{r}^{E N} > π_{r}^{E S}$ .
In addition, $μ_{r 2} < μ_{m 2}$ and $μ_{r 2}$ is a decreasing function of $k$ . The expressions for $μ_{m 2}$ and $μ_{r 2}$ are given in Appendix E.10.

Similar to the $N S$ case (shown in Lemma 1), there are two privacy concern threshold values in the $E S$ case: $μ_{m 2}$ and $μ_{r 2}$ . This means the incentives of the manufacturer and the retailer may be misaligned. When the privacy concern is moderate, that is, $μ_{r 2} < μ < μ_{m 2}$ , the manufacturer benefits from data sharing while the retailer is worse off for utilizing data. The reason is that even though both firms benefit from cross-selling, the retailer could be worse off if it uses data for cross-selling after taking into account the cost of the data-mining effort.

With regard to the channel profit, there is a threshold value $μ_{a}$ between $μ_{r 2}$ and $μ_{m 2}$ such that when $μ$ is greater than this threshold value, the channel profit decreases for data sharing; otherwise, it increases. In the region where $μ_{r 2} < μ < μ_{a}$ , the whole channel would be better off with data sharing, but the data is not shared. To improve supply chain efficiency, a contract with a transfer payment can be devised to make both parties better off.
4.3. Subgame Perfect Equilibria

This subsection shows the subgame perfect equilibria by considering the manufacturer’s encroachment decision. We first compare the manufacturer’s profits with and without encroachment by using Lemma 3 and obtain the following lemma. This lemma aids in determining the subgame perfect equilibria outlined in Proposition 4.

Lemma 4
(Manufacturer’s profit comparison) When consumer privacy is large ( $μ > μ_{r 2}$ ), the manufacturer does not share data if it encroaches. Also, its profit in $E N$ case $π_{m}^{E N}$ is
always greater than that in $N N$ case.

greater than that in $N S$ case when the privacy concern is large, that is, $π_{m}^{E N} > π_{m}^{N S}$ if $μ > μ_{3}$ ; $π_{m}^{E N} < π_{m}^{N S}$ otherwise.
When consumer privacy is small ( $μ < μ_{r 2}$ ), the manufacturer shares data if it encroaches. Also, its profit in $E S$ case $π_{m}^{E S}$ is
always greater than that in $N N$ case.

greater than that in $N S$ case when the effort cost coefficient is large enough or small enough; and otherwise, $π_{m}^{E S}$ is smaller than that in $N S$ case. That is, $π_{m}^{E S} > π_{m}^{N S}$ if $k < k_{1}$ or $k > k_{2}$ ; $π_{m}^{E S} < π_{m}^{N S}$ otherwise.
The expression for $μ_{r 2}$ is given in Appendix E.10, and those for $μ_{3}$ , $k_{1}$ , and $k_{2}$ in E.11.

We first discuss the case that $μ$ is large, that is, $μ > μ_{r 2}$ . In this case, the manufacturer prefers the $E N$ strategy to the $E S$ strategy when it encroaches. By comparing $N N$ and $E N$ cases, we find that the manufacturer’s profit in $E N$ is always greater than that in the $N N$ case. This is because the manufacturer will be better off by encroaching. Otherwise, it can set an infinitely large price in the direct channel and earn the same profit as in the $N N$ case. Next, we compare the manufacturer’s profits $π_{m}^{E N}$ and $π_{m}^{N S}$ . The manufacturer’s profit $π_{m}^{N S}$ decreases with privacy concern $μ$ while its profit $π_{m}^{E N}$ does not depend on $μ$ . Thus, there is a threshold value of privacy concern $μ_{3}$ : When $μ < μ_{3}$ , the manufacturer is better off by sharing data; otherwise, the manufacturer does not share data and gains higher profit in the $E N$ case.

In the case that $μ$ is small ( $μ < μ_{r 2}$ ), since the manufacturer prefers the $E S$ case to the $E N$ case and the $E N$ case to the $N N$ case, the manufacturer will always choose the $E S$ case over the $N N$ case. The result is not straightforward when comparing the $E S$ and $N S$ cases. When $k$ is large enough ( $k > k_{2}$ ), the retailer will not exert much effort in data mining, and the cross-selling value will be small enough. Then, by encroaching, the manufacturer gains additional revenue from the direct channel, while its loss in the retail channel is not much compared with the gain in the direct channel. Therefore, the manufacturer prefers $E S$ to $N S$ . However, as $k$ decreases, the data-mining effort will increase, making cross-selling more important and indirectly benefiting the manufacturer. When $k$ is lower than the threshold $k_{2}$ , it is not worthwhile for the manufacturer to encroach and potentially reduce cross-selling value, so the manufacturer prefers the $N S$ case. The implication is that IoT data sharing can stop the manufacturer from encroaching when the data mining cost is moderate ( $k_{1} < k < k_{2}$ ), compared with the no data-sharing case in which the manufacturer always encroaches.

More interestingly, in the case that $μ < μ_{r 2}$ , the manufacturer encroaches again when $k$ is small enough ( $k < k_{1}$ ). To explain this result, we compare data-mining efforts in the $E S$ and $N S$ cases, resulting in the following lemma.
Lemma 5
(Data-mining effort comparison between $E S$ and $N S$ cases) When the cost of data-mining effort is small, the retailer’s data-mining effort in the $E S$ case is higher than in the $N S$ case; otherwise, the retailer’s data-mining effort in the $E S$ case is smaller. That is, when $k < k_{5}$ , $e^{E S} > e^{N S}$ ; when $k > k_{5}$ , $e^{E S} < e^{N S}$ . The expression for $k_{5}$ is given in Appendix E.12.

According to Lemma 5, when $k$ is large, the data-mining effort in the $N S$ case is higher than in the $E S$ case. The intuition is that the manufacturer’s encroachment decreases the selling quantity in the retail channel, so the retailer has less incentive to exert effort in data mining since the effort is costly; that is, the data-mining effort decreases for encroachment. However, in the case of small $k$ , when the manufacturer encroaches, it will attract some consumers away from the retail channel. Due to the low cost of data-mining effort, the retailer increases the data-mining effort and lowers the product price to compete with the direct channel, thereby mitigating demand loss.

The result in Lemma 5 can explain why the manufacturer encroaches when $k$ is small enough. According to this lemma, by encroaching, the manufacturer causes the retailer to exert more effort in data mining. Then, even if cross-selling value decreases due to encroachment, it will not decrease too much. As a result, the manufacturer will be better off overall by encroaching.

By combining the results in Lemmas 1, 3, and 4, we have the following proposition.
Proposition 4
(Subgame perfect equilibria) In equilibrium, the manufacturer
encroaches but does not share data, that is, the manufacturer chooses the $E N$ strategy (Region A in Figure 3) when consumer privacy concern is large. That is, —
$μ > max {μ_{r 1}, μ_{r 2}}$ ,
—
$μ_{r 2} < μ < μ_{r 1}$ and $μ > μ_{3}$ .

does not encroach but shares data, that is, the manufacturer chooses the $N S$ strategy (Region B in Figure 3) when the privacy concern is medium, or the privacy concern is small and the effort cost coefficient is medium. That is, —
$μ_{r 2} < μ < μ_{r 1}$ and $μ < μ_{3}$ ,
—
$μ < min {μ_{r 1}, μ_{r 2}}$ and $k_{1} < k < k_{2}$ .

encroaches and shares data otherwise, that is, the manufacturer chooses the $E S$ strategy (Region C in Figure 3).
The expression for $μ_{r 1}$ is given in Appendix E.4, for $μ_{r 2}$ in Appendix E.10, and those for $μ_{3}$ , $k_{1}$ , and $k_{2}$ in E.11.
Figure 3.
Subgame perfect equilibria when $b = 1 / 20$ , $c = 1$ , $d = 3$ , $σ = 3$ .

For ease of understanding, we also plot a figure (Figure 3) that shows the regions corresponding to the equilibria in Proposition 4. Proposition 4 shows that the $N S$ equilibrium arises in two regions, $B_{1}$ and $B_{2}$ , where the manufacturer chooses not to encroach. In Region $B_{1}$ , privacy concerns are moderate and data mining cost is high, limiting the retailer’s benefit from data sharing. If the manufacturer were to encroach in this case, the retailer would avoid using IoT data due to privacy costs, intensifying channel competition and leaving both parties worse off. As the data mining cost $k$ decreases, the retailer’s gain from cross-selling rises. Even with manufacturer encroachment, the retailer still benefits from using IoT data—this corresponds to Region $C_{1}$ . When $k$ decreases further, the cross-selling value becomes sufficiently high such that if the manufacturer refrains from encroachment, the resulting increase in indirect cross-selling benefits outweighs the loss of direct channel revenue. This gives rise to Region $B_{2}$ . Finally, when $k$ becomes very small ( $k < k_{1}$ ), manufacturer encroachment reduces retail demand, but the retailer can offset this effect by increasing its data mining effort (per Lemma 5). As a result, the cross-selling benefit remains substantial. In this case, the manufacturer gains revenue from the direct channel without suffering significant losses in the retail channel—corresponding to Region $C_{2}$ .

Past literature has shown that the incentive of a manufacturer’s encroachment could be reduced due to certain factors such as nonlinear pricing (Li et al., 2015), quality differentiation (Ha et al., 2016), strategic inventory withholding (Guan et al., 2019), and retailer’s service effort (Ha et al., 2022). Our study contributes to this body of literature by considering the impact of IoT data sharing on encroachment. Proposition 4 shows that the manufacturer could stop encroaching in the presence of IoT data sharing if we also factor in the data mining effort. The managerial implication is as follows. A manufacturer should be cautious when contemplating encroachment since encroachment can reduce the benefit of IoT data sharing. It might need to stop encroaching with the arrival of IoT. In Appendix E.13.1, we show that for small $μ$ , the manufacturer’s and the retailer’s profits increase with the substitution rate in the subgame perfect equilibrium $E S$ when $k < k_{b 2}$ . Using Figure 3 (in the two-dimensional space of data mining cost $k$ and privacy cost $μ$ ), we can show how the change of IoT data value, represented by data-mining effort cost, affects equilibria and profits in Proposition 5.
Proposition 5
(Impact of data-mining effort cost on equilibria and profits) When the data-mining effort cost coefficient $k$ increases, the equilibria and profits change in the following ways.
The equilibrium first changes from $E S$ to $N S$ when crossing the line $k = k_{1}$ . The manufacturer’s profit does not change, but the retailer’s profit increases discontinuously (increasing with a jump).

Then it changes from $N S$ to $E S$ when crossing $k = k_{2}$ . The manufacturer’s profit remains unchanged, but the retailer’s profit decreases discontinuously.

Finally, it changes in two scenarios. —
In the first scenario, the privacy cost is small. It changes from $E S$ to $E N$ directly when crossing the line $μ = μ_{r 2}$ . The manufacturer’s profit decreases discontinuously (which we refer to as decreasing with a drop), while the retailer’s profit does not change.
—
In the second scenario, the privacy cost is not so small. The equilibrium changes in the following ways. *
First it changes from $E S$ to $N S$ when crossing the line $μ = μ_{r 2}$ . The manufacturer’s profit decreases (with a drop), and the retailer’s profit increases (with a jump).
*
Then it changes from $N S$ to $E N$ when crossing the line $μ = μ_{r 1}$ or $μ = μ_{3}$ . When it crosses the line $μ = μ_{r 1}$ , the manufacturer’s and the retailer’s profits decrease (with a drop). When it crosses the line $μ = μ_{3}$ , the manufacturer’s profit does not change, and the retailer’s profit decreases (with a drop).

Proposition 5 shows how the change in data-mining effort cost affects equilibria and firms’ profits. To help readers visually see such impacts and gain more insights, we illustrate the results in Figure 4. From Proposition 5 and Figure 4, we can see that the manufacturer’s profit decreases monotonically with $k$ . One interesting case is that the manufacturer’s profit decreases with a drop when the equilibrium changes from the $E S$ to $N S$ by crossing the line $μ = μ_{r_{2}}$ . This $N S$ equilibrium differs from the manufacturer’s preferred strategy $E S$ . This result can be understood as follows. When data sharing occurs, channel encroachment by the manufacturer diminishes the retailer’s incentive to invest in data mining efforts, as the associated marginal benefit declines. As a result, the retailer’s effort level is lower in the presence of encroachment than in its absence. Therefore, it is beneficial for the manufacturer to refrain from encroaching, as this encourages the retailer to invest more in data mining, ultimately minimizing profit losses for the manufacturer. The manufacturer’s profit also decreases with a drop when the equilibrium changes from the $N S$ to $E N$ . This is because the data mining cost is too high, and it is not worthwhile for the manufacturer to share the data. In this case, the manufacturer chooses to encroach instead.

Figure 4.
Impacts of $k$ ( $b = 1 / 20$ , $c = 1$ , $d = 3$ , $σ = 3$ , $μ = 0.9$ ). (a) Manufacturer’s profit; (b) Retailer’s profit.

However, the retailer’s profit does not decrease monotonically with $k$ . In particular, the retailer’s profit increases when the equilibrium strategy moves from $E S$ to $N S$ by crossing the line $k = k_{1}$ or $μ = μ_{r 2}$ (see Figure 4(b)). This counter-intuitive result is due to the following mechanism related to the data mining cost $k$ . When the cost of data-mining effort increases and crosses the line $k = k_{1}$ or $μ = μ_{r 2}$ , the retailer will decrease the data-mining effort while the manufacturer stops encroaching and tries to benefit more from cross-selling instead. Then, with the disappearance of channel competition, the retailer benefits despite the increased cost of data-mining effort. From the above discussion, we can see that cross-selling due to data-mining efforts plays an important role in this counter-intuitive result. Since the profits of the manufacturer and the retailer can change in different directions, the channel profit (i.e., the sum of both profits), can either increase or decrease when the equilibrium strategy moves from $E S$ to $N S$ . That is, an increase in data-mining effort cost can make the channel worse off or better off when the supplier stops encroaching but still shares IoT data.

To summarize, we find that the retailer and the supply chain can be worse off when IoT data value increases (either data mining cost $k$ or privacy cost $μ$ decreases or cross-selling value $σ$ increases). The intuition is that the manufacturer has higher market power than the retailer does. Therefore, although increased IoT data value could generate additional cross-selling value, the manufacturer can adjust its channel encroachment and wholesale price to gain higher profit. As a result, the retailer and the whole supply chain could be worse off due to heightened competition. This result contradicts the common intuition that higher data value benefits firms.

Past literature has shown the positive impact of IoT data in areas such as process improvement (Shi et al., 2025) , risk assessment (Ho et al., 2022), predictive printing (Song and Zhang, 2025), and preventative maintenance (Pei et al., 2023). Also, Sun and Ji (2022) find that a retailer always benefits from an increase in IoT value. However, our study reveals that as the value of IoT data increases, the retailer’s profit may decline due to the manufacturer’s first-mover advantage, which could lead to encroachment at the retailer’s expense. For a retailer, this implies the need for caution when increasing the value of IoT data. For instance, it should carefully evaluate the adoption of more cost-efficient data mining technologies for customer insights or enhancements in ad impression designs to attract customers. In some cases, such actions could trigger manufacturer encroachment, ultimately reducing the retailer’s profit. For the manufacturer, our findings indicate that to encourage the retailer to enhance its overall cross-selling effectiveness, it could provide sufficient financial incentives or share expertise to help the retailer analyze and utilize IoT data more cost-effectively.
4.4. Information Sharing Vs Data Sharing: Role of Data-mining Effort

In a traditional information-sharing setting, information such as inventory or demand information can provide value without requiring a data-mining effort. In the following, we show the key role that data-mining effort plays in generating the important results in this study. If shared data is directly usable by the retailer for cross-selling, then data-mining effort is not necessary. The inverse demand functions of direct and retail channels are the same as (4) and (5), and the manufacturer’s profit is the same as (6). The retailer profit function becomes

π_{r} = q_{r} (p_{r} - w) + J β q_{r},

(8)

where

β

is the marginal value of information. In this retailer profit function, there is no data-mining effort, unlike (7). The time sequence is similar to the main model, except there is no data-mining effort decision. We can solve the equilibrium in the encroachment and no encroachment cases with backward induction to get the following lemma.

Lemma 6

(Subgame outcomes with information sharing) When the manufacturer does not encroach, the subgame outcomes are

π_{r}^{N k} = \frac{1}{16} (d - c - J (μ - β))^{2}

(9)

and

π_{m}^{N k} = \frac{1}{8} (d - c - J (μ - β))^{2} .

(10)

When the manufacturer encroaches, the subgame outcomes are

π_{r}^{E k} = \frac{{(b^{2} + 2)}^{2} ((1 - b) (d - c) - J (μ - β))^{2}}{(1 - b^{2}) {(b^{2} + 8)}^{2}}

(11)

and

\begin{aligned} π_{m}^{E k} = \frac{(b (b^{2} + b + 4) + 12) (1 - b) (d - c)^{2} - 4 J (μ - β) (2 (1 - b) (d - c) - J (μ - β))}{4 (b + 1) (b^{2} + 8) (1 - b)} . \end{aligned}

(12)

Superscript $k \in {N, S}$ .

Building on this, we have the following proposition.

Proposition 6

(Subgame perfect equilibria with information sharing) In the subgame perfect equilibria, the manufacturer will always encroach. When the effect of privacy concern is smaller than the effect of cross-selling value, the manufacturer shares information; otherwise, the manufacturer does not share information. That is, when $μ < β$ , $π_{m}^{E N} < π_{m}^{E S}$ ; otherwise, $π_{m}^{E N} > π_{m}^{E S}$ .

This proposition differs sharply from Proposition 4. In Proposition 4, the strategy that the manufacturer shares data but does not encroach can be an equilibrium strategy. However, in Proposition 6, this is not true; the manufacturer always encroaches, even if it shares information. The difference comes from the data-mining effort. When we consider the data-mining effort, the manufacturer’s encroachment decision will affect the retailer’s data-mining effort, so the manufacturer needs to cautiously balance the benefit of encroaching and the loss in the retail channel due to the reduced retailer’s effort. However, in the case without data-mining effort, the manufacturer does not need to consider the potential negative effect of a reduction in data-mining effort, leading it to always encroach in this case. This result shows that data-mining effort, essential in turning raw data into useful information, is an important feature, and considering it distinguishes this paper from other traditional information-sharing works.

5. Extensions

We have also checked the robustness of our main model by relaxing certain assumptions. Specifically, we have considered various cases, including: (1) The manufacturer could sell the data, (2) the manufacturer shares partial data in the retail channel, (3) two firms engage in quantity competition, (4) the manufacturer incurs cost when it shares data, and (5) some proportion of consumers is unwilling to share data. The main results remain the same. Due to space constraints, the details of the extensions are provided in Appendix D.

6. Conclusion

This paper studies the role of IoT data sharing in a supply chain where a manufacturer sells through a retailer and, possibly, a direct channel. We analyze the strategic interaction between IoT data sharing and channel encroachment.

6.1. Theoretical Contributions and Implications

Our paper contributes to the literature in three aspects. First, we show that the manufacturer and the retailer can benefit from a higher substitution rate when the manufacturer encroaches and shares IoT data. This phenomenon occurs in the region where the cost of data-mining effort is small enough. In this region, when the channel substitution rate increases, the manufacturer sells less in the direct channel to reduce channel competition and sells more through a wholesale contract in the retail channel. As a result, the retailer is better off due to a higher cross-selling value, and the manufacturer’s profit also improves. Our research adds to the active debate on when competition hurts firms (Huang et al., 2018; Niu et al., 2019; Yoon, 2016) or benefits firms (Li et al., 2022). In addition, this result suggests that when retailers have a highly cost-effective way of mining IoT data, policymakers should encourage manufacturers and retailers to make their channels more transparent. Then, customers can access more information about each channel and compare products more easily, resulting in a higher product substitution rate.

Second, we are the first to show that IoT data sharing can significantly affect the manufacturer’s encroachment decision. In the cases without data sharing, the manufacturer always encroaches. However, in the data-sharing cases, the manufacturer may not encroach. One reason is that if the manufacturer encroaches, the retailer’s equilibrium data-mining effort could decrease, hurting the manufacturer’s profit. Interestingly, it is again optimal for the manufacturer to encroach when the data-mining effort is small enough, contrary to intuition. In this region, the retailer actually increases its data-mining effort, rather than reducing it, in response to the manufacturer’s encroachment. Our findings suggest that, with the advent of IoT technology, manufacturers must strike a balance between the direct advantages of encroachment and the indirect benefits of IoT data sharing. In some cases, it may be more beneficial to forgo encroachment and instead collaborate with the retailer, leveraging IoT data sharing for mutual gain. Our results provide a plausible explanation for why certain IoT manufacturers such as Lennox choose to sell their products through retailers only.

Third, we find that when the IoT data value increases, the retailer’s profit could decrease with a drop. That is, IoT data more helpful to the retailer for sales may end up hurting the retailer’s profit. The reason is due to the manufacturer’s opportunistic encroachment behavior. When IoT data value increases, the manufacturer expects the retailer to increase its data-mining effort. Then, the manufacturer can encroach without decreasing the cross-selling value by much. In other words, the manufacturer tries to gain from both channels, squeezing the retailer and leaving the retailer worse off. The underlying force that drives the above results is the existence of data-mining effort. Our research contributes to the ongoing discussion on the challenges businesses face when implementing IoT technologies and the ways to overcome these challenges (Forbes, 2023). Our findings suggest that retailers should be mindful when adopting more efficient data mining technologies, as this could inadvertently trigger manufacturer encroachment and lead to profit reductions. For manufacturers, encouraging retailers to enhance the value of IoT data can be achieved by providing financial incentives or sharing expertise to help them mine data more cost-effectively.

6.2. Future Direction

Finally, we discuss avenues for future work. First, future research could consider the impact of IoT data on a retailer’s product assortment decision. A retailer could optimize its product assortment to leverage IoT data better. Without cross-selling, two products might have the same marginal value to a retailer. However, with information available for cross-selling, these products could be of different value. Therefore, the retailer should consider the possibility of cross-selling when making assortment decisions. Second, the manufacturer may encroach through other means rather than establishing its own channel in certain cases. For example, some platforms provide an agency channel to the manufacturer and charge a commission rate. Therefore, it would be interesting to consider channel encroachment through an agency channel instead of a direct channel.

Supplemental Material

sj-pdf-1-pao-10.1177_10591478251400471 - Supplemental material for Channel Encroachment and Supply Chain Performance: The Effects of Internet of Things Data Sharing

Supplemental material, sj-pdf-1-pao-10.1177_10591478251400471 for Channel Encroachment and Supply Chain Performance: The Effects of Internet of Things Data Sharing by Can Sun, Yonghua Ji and Radha Mookerjee in Production and Operations Management

Footnotes

Acknowledgments

The authors gratefully thank the department editor Subodha Kumar, the senior editor, and two anonymous reviewers for their constructive comments that have greatly improved the paper.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received the following financial support for the research, authorship and/or publication of this article: Can Sun’s research is supported by the National Natural Science Foundation of China [Grants 72201262, 72581260225] and the Fundamental Research Funds for the Central Universities [Grant WK2040000112]. Yonghua Ji’s research is supported by University of Alberta School of Business Alex Hamilton Professorship of Business.

ORCID iDs

Can Sun

Yonghua Ji

Radha Mookerjee

Supplemental Material

Supplemental material for this article is available online (doi: ).

Notes

How to cite this article

Sun C, Ji Y and Mookerjee R (2025) Channel Encroachment and Supply Chain Performance: The Effects of Internet of Things Data Sharing. Production and Operations Management XX(X): 1–16.

References

Arya

Mittendorf

Sappington

DEM

(2007) The bright side of supplier encroachment. Marketing Science 26(5): 651–659.

Bimpikis

Crapis

Tahbaz-Salehi

(2019) Information sale and competition. Management Science 65(6): 2646–2664.

Chen

Y-J

Deng

(2015) Information sharing in a manufacturer-supplier relationship: Suppliers’ incentive and production efficiency. Production and Operations Management 24(4): 619–633.

Chen

Özer

(2019) Supply chain contracts that prevent information leakage. Management Science 65: 5619–5650.

China’s National People’s Congress (2021) Personal information protection law of the People’s Republic of China. https://www.gov.cn/xinwen/2021-08/20/content_5632486.htm (accessed December 23, 2024).

Choe

Cong

Wang

(2024) Softening competition through unilateral sharing of customer data. Management Science 70(1): 526–543.

Choi

Blocher

Gavirneni

(2008) Value of sharing production yield information in a serial supply chain. Production and Operations Management 17(6): 614–625.

Croson

Donohue

(2003) Impact of POS data sharing on supply chain management: An experimental study. Production and Operations Management 12(1): 1–11.

Demirhan

Jacob

Raghunathan

(2007) Strategic IT investments: The impact of switching cost and declining IT cost. Management Science 53(2): 208–226.

10.

Yang

, et al. (2025) PPFL: A personalized federated learning framework for heterogeneous population. INFORMS Journal on Computing 0(0): 0. DOI: 10.1287/ijoc.2023.0376.

11.

European Parliament, Council of the European Union (2016) General data protection regulation. https://gdpr-info.eu/art-7-gdpr/ (accessed December 23, 2024).

12.

Forbes (2023) Implementing IoT tech? 15 challenges to be ready for. https://www.forbes.com/councils/forbestechcouncil/2023/06/08/implementing-iot-tech-15-challenges-to-be-ready-for/ (accessed Febuary 23, 2025).

13.

Gao

Shou

(2014) Information acquisition and voluntary disclosure in an export–processing system. Production and Operations Management 23(5): 802–816.

14.

Ghoshal

Hao

Menon

, et al. (2020) Hiding sensitive information when sharing distributed transactional data. Information Systems Research 31(2): 473–490.

15.

Ghoshal

Kumar

Mookerjee

(2020) Dilemma of data sharing alliance: When do competing personalizing and non-personalizing firms share data. Production and Operations Management 29(8): 1918–1936.

16.

Guan

Gurnani

Geng

, et al. (2019) Strategic inventory and supplier encroachment. Manufacturing & Service Operations Management 21(3): 536–555.

17.

Guo

Zhang

(2014) Strategic information sharing in competing channels. Production and Operations Management 23(10): 1719–1731.

18.

Long

Nasiry

(2016) Quality in supply chain encroachment. Manufacturing & Service Operations Management 18(2): 280–298.

19.

Luo

Shang

(2022) Supplier encroachment, information sharing, and channel structure in online retail platforms. Production and Operations Management 31(3): 1235–1251.

20.

Tong

Wang

(2022) Channel structures of online retail platforms. Manufacturing & Service Operations Management 24(3): 1547–1561.

21.

Y-J (Ian)

Siyuan

, et al. (2022) Is it all about you or your driving? Designing IoT-enabled risk assessments. Production and Operations Management 31(11): 4205–4222.

22.

Huang

Guan

Chen

(2018) Retailer information sharing with supplier encroachment. Production and Operations Management 27(6): 1133–1147.

23.

Huawei (2022) About HUAWEI Health Kit. https://developer.huawei.com/consumer/en/doc/development/HMSCore-Guides/health-introduce-0000001053684429 (accessed February 11, 2024).

24.

Lennox (2021) Privacy. https://www.lennox.com/privacy (accessed December 23, 2024).

25.

Tian

Zheng

(2021) Information sharing in an online marketplace with co-opetitive sellers. Production and Operations Management 30(10): 3713–3734.

26.

Schneider

, et al. (2023) Reidentification risk in panel data: Protecting for k-anonymity. Information Systems Research 34(3): 1066–1088.

27.

Chen

Y-J

(2022) Strategic inventories under supply chain competition. Manufacturing & Service Operations Management 24(1): 77–90.

28.

X-B

Qin

(2017) Anonymizing and sharing medical text records. Information Systems Research 28(2): 332–352.

29.

Gilbert

Lai

(2015) Supplier encroachment as an enhancement or a hindrance to nonlinear pricing. Production and Operations Management 24(1): 89–109.

30.

Liangshuo Technology (2023) Why does Tmall Genie, which ranks first in China and second in the world, have to be sold at a loss? https://jiameng.baidu.com/content/detail/6079492356?from=search&rid=1.35.41.132 (accessed December 23, 2024).

31.

Liu

Zhang

(2021) Information sharing on retail platforms. Manufacturing & Service Operations Management 23(3): 606–619.

32.

Mehta

Dawande

Janakiraman

, et al. (2021) How to sell a data set? Pricing policies for data monetization. Information Systems Research 32(4): 1281–1297.

33.

Menon

Ghoshal

Sarkar

(2022) Modifying transactional databases to hide sensitive association rules. Information Systems Research 33(1): 152–178.

34.

Niu

Chen

Fang

, et al. (2019) Technology specifications and production timing in a co-opetitive supply chain. Production and Operations Management 28(8): 1990–2007.

35.

Pei

Yan

Kumar

(2023) No permanent friend or enemy: Impacts of the IIoT-based platform in the maintenance service market. Management Science 69(11): 6800–6817.

36.

Wang

(2025) Privacy protection and statistical efficiency trade-off for federated learning. INFORMS Journal on Computing 0(0): 0. DOI: 10.1287/ijoc.2024.0554.

37.

Samsung (2022) Samsung health SDK for device. https://developer.samsung.com/health/device/overview.html. (accessed February 11, 2024).

38.

Samsung (2025) Samsung privacy policy. https://www.samsung.com/us/account/privacy-policy/#pp_03. (accessed September 13 2025).

39.

Shamir

Shin

(2018) The perils of sharing information in a trade association under a strategic wholesale price. Production and Operations Management 27(11): 1978–1995.

40.

Shi

, et al. (2025) Analytics for IoT-enabled human–robot hybrid sortation: An online optimization approach. Production and Operations Management 34(11): 3364–3380.

41.

Song

J-S

Zhang

(2025) Predictive three-dimensional printing of spare parts with internet of things. Management Science 71(3): 1925–1943.

42.

Statistica Inc (2022) Number of Internet of Things (IoT) connected devices worldwide from 2019 to 2021, with forecasts from 2022 to 2030. https://www.statista.com/statistics/1183457/iot-connected-devices-worldwide/. (accessed January 11, 2024).

43.

Sun

(2022) For better or for worse: Impacts of iot technology in e-commerce channel. Production and Operations Management 31(3): 1353–1371.

44.

Wang

Zheng

Z(Eric)

Jiang

, et al. (2021) Blockchain-enabled data sharing in supply chains: Model, operationalization, and tutorial. Production and Operations Management 30(7): 1965–1985.

45.

Wegner

(2022) Global IoT market size grew 22% in 2021 – These 16 factors affect the growth trajectory to 2027. https://iot-analytics.com/iot-market-size/ (accessed February 11, 2024).

46.

Xiaomi (2020) Xiaomi health privacy policy. https://privacy.mi.com/xiaomihealth/en_US/ (accessed February 11, 2024).

47.

Xin

Choudhary

(2019) IT investment under competition: The role of implementation failure. Management Science 65(4): 1909–1925.

48.

Yoon

D-H

(2016) Supplier encroachment and investment spillovers. Production and Operations Management 25(11): 1839–1854.

49.

Zhang, . (2006) Competition, cooperation, and information sharing in a two-echelon assembly system. Manufacturing & Service Operations Management 8(3): 273–291.

50.

Zhao

Xue

Zhang

(2014) Outsourcing competition and information sharing with asymmetrically informed suppliers. Production and Operations Management 23(10): 1706–1718.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.03 MB