User Privacy Awareness, Incentive and Data Supply Chain Pricing Strategy

Abstract

In recent years, the collection, mining, and utilization of data have become a new profit growth point for enterprises, and these events have also accelerated the pace of enterprises to collect users’ data. However, the relevance of personal data privacy and the frequent occurrence of data leakage events have increased users’ privacy awareness. The purpose of our study is to enhance the effective flow of data while protecting users’ data privacy. The data supply chain consists of the end user, data provider, and service provider, and involves the flow of the value-added process of data. Our study focuses on the pricing strategy of data products considering data incentive and data protection levels. We propose three models—centralized pricing, decentralized pricing, and revenue-sharing pricing—and solve them, and then we analyze the impact of users’ privacy awareness on data incentives, protection, and pricing of data products in the three models. We also analyze which pricing method works best for participants.

1. Introduction

In 2017, The Economist magazine declared that “the world’s most valuable resource is no longer oil, but data”. In addition, according to the European Conference Report, the scale of data generated globally will increase from 33ZB in 2018 to 175ZB by 2025, a five-fold increase in just seven years [1]. Many platforms provide users with free services in exchange for their activity data [2,3] and sell the collected data to data demanders [4]; data demanders use advanced information technology to develop data products or services and sell them to end users to guide their daily business decisions [5]. Data commercialization based on supply chains has become a new business model. Today, personal data are regarded as a new asset for enterprises, which can create added value for enterprises and users, and stimulate enterprises’ determination to collect more data [6,7]. However, due to the relevance of data privacy and the frequent occurrence of corporate data breaches, users’ privacy awareness is also gradually increasing [8], which makes data collection more and more difficult and also brings new challenges to the development of data products.

Currently, many platforms set privacy policies on the use of users’ data, and these can alleviate users’ concerns about data privacy disclosure; meanwhile, data incentives can encourage users to share more personal data because data incentives can offer some kinds of economic reward or free service [9]. Under the dual effects of data protection and data incentives, it can stimulate the vitality of the data supply chain and improve the quality of data products. Therefore, our study focuses on data incentives, data protection, and data pricing strategies based on the data supply chain, which mainly involves three stages: the data collection stage when the Data Provider (DP) offers data incentives and data protection to the user; the data transaction stage when the DP sells the collected data to the Service Provider (SP); the data products transaction stage when the SP purchases data, develops data products, and sells data products to End Users (EU).

With the development of the data market, more and more scholars pay attention to data pricing [3,10]; however, they only focus on the data collection process between users and platforms or the data transaction process between data platforms and data demanders, and they lack data pricing research based on the supply chain perspective. They also point out that it is of great significance to give full play to the role of the data factor market and promote the efficient circulation of data. The data supply chain involves the entire data flow process from data generation, collection, and transaction to the final development of data products. There have been numerous reports on research based on traditional supply chain coordination mechanisms, including centralized pricing, decentralized pricing, revenue sharing, bargaining, unbiased samples [11], auction pricing [12,13], etc. These methods provide important theoretical support for our research. However, data pricing is affected by various factors such as data scale, data privacy, and development costs of data products, etc. There are obvious differences between data supply chain pricing and traditional product supply chain pricing [14]. To the best of our knowledge, there is still no data supply chain research that considered data privacy, data protection, data incentives, etc., and there is no data pricing research based on the supply chain perspective. Therefore, research on pricing mechanisms based on the total data supply chain is the first motivation of our paper.

The mining and utilization of the value of personal data have become a new profit growth point for enterprises [15,16]. However, with the enhancement of users’ awareness of data privacy and data value, users’ willingness to share data with enterprises is reduced [17]. With the rapid development of the data economy, many countries begin to pay attention to the protection of data privacy (it refers to the anonymization level of data or data protection commitments), and relevant laws have been enacted, such as the Personal Information Protection Law, the General Data Protection Regulation (GDPR) [18], etc. Many scholars have conducted research on data protection from the perspectives of legislation, policy, and technology updates [19], and these studies confirmed the positive impact of data protection on personal data-sharing behaviors. For instance, Schomakers’s research points out that anonymization is the most important factor that influences users’ decisions to share data [20]. As mentioned above, in the context of the data supply chain, there is little research on mathematical models considering users’ data protection. That is our second research motivation.

In recent years, more and more enterprises have realized the huge potential value of personal data and also users’ privacy awareness when sharing data [21]. Users with higher privacy awareness restrict their data-sharing behavior and reduce the amount of data sharing [22]. In order to obtain more user data, data incentives are the most commonly methods, including providing free services, coupons, discounts, and so on. Taking incentive measures to compensate for the loss of users’ privacy can reduce the cost of users’ privacy to a certain extent, thus stimulating users’ willingness to share data. The incentive level of the data platforms directly affects the scale of data shared by users, which in turn affects data transactions, data product development, and sales. Therefore, the design of data incentive mechanisms is particularly important.

At present, some research on user data incentives mainly focuses on algorithm design based on incentive mechanisms [23,24] and bargaining between users and data demanders. Today, data incentives and data protection are common ways to obtain user data (data protection can reduce users’ privacy concerns, and data incentives can compensate for users’ privacy losses). To the best of our knowledge, there are no mechanism-design studies that consider both data incentives and data protection. When data incentives and data protection coexist, users face a trade-off between privacy costs and economic benefits [25], which has become a new challenge for the data supply chain. Therefore, a study on the data supply chain pricing strategy considering data protection and data incentives is our third research motivation.

To sum up, our study focuses on the pricing strategy of data products under the consideration of data protection and incentives. The main contributions of our paper are as follows:

(1) With the data supply chain as the background, this study analyzes the complete supply chain process from data generation, collection, and transaction to data products development and transaction. Data are the main flow factor.

(2) In the process of data collection, taking into account users’ privacy awareness, the data providers take two positive actions: data incentives and data protection.

(3) Based on the three pricing models, the optimal data incentives level, protection level, and data products pricing are solved, and the impact of users’ privacy awareness, data product development technology level, and other parameters are analyzed.

(4) Simulation analysis verifies the validity and accuracy of the model. We focus on the comparative analysis of the decentralized pricing model and the revenue-sharing model, study the impact of users’ privacy risk awareness on data incentives, protection, and data product pricing, and analyze the optimal decisions of all participants in different market environments, providing a scientific basis for enterprise data collection and data exchange.

2. Theoretical Background

2.1. Data Supply Chain

Based on the two-sided platform, data is regarded as an economic commodity, and data sharing activities are regarded as an economic transaction [26]. Users and data demanders complete data transmission through an intermediate data platform. Shota Ichihashi points out that the data market is composed of consumers, data intermediaries, and downstream firms. Data intermediaries collect personal data from consumers and sell this information to downstream firms [4]. Bataineh proposes that the data market mainly consists of three parties: data providers, data brokers, and data consumers [26]. Among them, data providers refers to individuals, enterprises, or governments, which are mainly responsible for producing, storing, and integrating data, which is the starting point of the data supply chain. Data brokers play the role of intermediaries, and they are a supply and demand negotiation platform for data providers and data consumers, providing a place for data transactions between the two. Data consumers are organizations whose business often requires huge amounts of data with particular specifications. The data supply chain refers to the supply chain including the Data Provider (DP), Service Provider (SP), and End User (EU), in which data as the key element involve the value-added process flow from generation, collection, and processing data to finally becoming some kind of data products or data services for end users.

In our article, a DP is similar to a data broker, providing data incentives and data protection to encourage users to share more data, and then selling the collected raw data to an SP. The SP is the data consumer, mainly engaged in three businesses: purchase of raw data from the DP at a certain price, processing of data using information technology, development of data products, and selling data products to the EU. The EU is the demander of data products: it can be a government or an enterprise. Scholars use empirical research to measure the value of user data and propose a combined auction data pricing method [27]. Taking a game model analysis based on bargaining, the pricing of data is affected by factors such as the bargaining power of the supply and demand sides [28]. Therefore, we also study the pricing of data supply chains based on game theory model.

2.2. Data Privacy, Privacy Awareness and Data Protection

Data privacy, also known as online privacy or information privacy, is the inability of individuals to fully control how their data are shared online, and the personal privacy involved is managed jointly by the individual and other users on the platform [29]; it indicates that consumers are worried about how information in their personal files is shared with organizations and across organizations [30]. In recent years, the frequent occurrence of data leakage incidents has drawn the attention of consumers and the state to the protection of personal data privacy [31], and users’ data privacy awareness has gradually increased. Although some platforms have set up privacy statement policies, the privacy notices are usually too long and the content is difficult to understand, leading users to unconsciously agree to the platform’s privacy policy, which poses some risk of customer data privacy being exposed. Most of the scholars’ research on users’ data privacy protection focuses on three aspects: (1) algorithm research to strengthen users’ data privacy protection, mainly including adding noise, blockchain technology, K-means algorithm, etc. [32,33,34,35]; (2) research on data protection measures taken by users themselves, such as paying to protect privacy, choosing to refuse data sharing, erasing data traces, etc. [9,36,37,38,39,40,41]; (3) data protection research, such as improving legislation, establishing data ownership, etc. [42,43]. Correspondingly, Choi’s theoretical model on privacy emphasizes that data collection requires the consent of consumers and that consumers are fully aware of the consequences of such consent, emphasizing the close connection between data protection and users’ awareness of privacy [8]. These methods have strengthened the protection of users’ data privacy to a large extent, but there is a lack of research on data protection based on mathematical models.

Our study argues that data protection has a positive impact on users’ data-sharing behavior, but a too high level of data protection means higher data protection costs and a lack of personalized customer data, thus reducing the value of collected data. Therefore, the data supply chain is faced with how to determine the optimal data protection level.

2.3. Data Awareness and Data Incentive

Today, personal data is seen as new oil or currency in the digital world; at the same time, the advancement of information technology has reduced the cost of data storage and data processing, and also promoted the emergence of platforms that use users’ data to make profits through targeted advertising, personalized recommendations, and price discrimination [25,44,45,46,47]. However, with the enhancement of users’ privacy awareness, it is more difficult to obtain users’ data for free. Enterprises often use discounts and subsidies to make up for the loss of users’ privacy, alleviate users’ privacy concerns, and obtain more data shared by customers. Zhang studies the incentive design under the mid-stem model and proposes a REAP (REconciling Aggregation accuracy and individual Privacy) mechanism to coordinate the aggregation accuracy of the data-collection platform and the data privacy of individuals, but he pays more attention to algorithms [23]. Guan et al. conducted a conjoint analysis and found that financial incentives do affect individuals’ preferences for websites with different privacy policies [48]. Meanwhile, Smith’s empirical research shows that platforms can facilitate users’ data sharing through financial incentives [49]. Li et al. proposed a simple linear incentive model, which requires the corresponding incentive degree to be determined according to the amount of data shared by users [50]. Our paper follows the design of data incentive mechanisms related to the amount of data—that is, the more data users share, the more incentives they receive.

We use

Table 1. The publications related to data privacy and data incentive problems.

Theme	References	Similarities	Differences
Data privacy	Smith et al. [49]	Data pricing; Data value	Theoretical analysis We use a mathematical model.
Data compensation	Kummer et al. [25]	Data incentive; Data protection	Emphasis on platform use and data collection Similar to the first stage of our paper.
	Montes et al. [47]	Personal information	Endogenous privacy and price discrimination We develop data product.
Data collection	Johnson et al. [44]	Data value; privacy concerns	Targeted advertising We are dealing with the data.
	Rafieian, O. and H. Yoganarasimhan [45]	Data value; Consumer privacy; Use of data
	Zuiderveen Borgesius, F. and J. Poort [46]		Price discrimination We develop data product and then sale.
Privacy preferences	Guan et al. [48]	Data collection Data quality	Conjoint analysis Empirical research
Data incentive	Zhang et al. [23]	Data privacy Data incentive	Computer algorithm
	Li et al. [50]	Data incentive and data scale	Data collection stage Our research covers the whole stage from data generation to transaction.

to describe the similarities and differences obtained from the existing literature.

To sum up, data incentives can make up for the loss of users’ privacy to a certain extent, and data protection can reduce users’ privacy awareness. The combined action of these two factors can affect the scale of data shared by users. The pricing of service providers’ data products is affected too. Therefore, our research on data protection, data incentives, and data products pricing is of great significance to improving the value of the data supply chain.

2.4. Pricing Model

In centralized pricing, enterprises at each node of the supply chain pursue the maximization of the total profit of the supply chain system based on cooperation. In decentralized pricing, enterprises at each node of the supply chain try to maximize their own profits, without considering the profits of the entire supply chain. It is a pricing method based on a non-cooperative game model. Revenue sharing is also a common model in supply chain pricing: entity A shares r portion of its products revenue with B to achieve a win-win situation for both parties. We also use this traditional research framework.

3. Model

3.1. Problem Description

The model assumes that the data market structure is a duopoly market consisting of only one DP and one SP and that the SP will purchase all the raw data collected by the DP (see Figure 1). The whole transaction process is as follows: first, the SP puts forward the demand for data protection level $\beta$ and the relationship between the wholesale data price $w$ and the data protection level; then the DP decides the data incentive level $\alpha$ to compensate for the loss of users’ privacy and encourages users to share data under the condition of ensuring that the SP demand is met, and, finally, obtains user data with data size of $n$; then the DP sells user data to the SP at the wholesale price $w$; the SP reprocesses the purchased data at a technical level $h$, develops a data product of quality $q$, and sells it to the EU at the price $p$. Please the summary of notations at

Table 2. Summary of notations.

Symbol	Descriptions
$k_1$	User’s privacy risk awareness
$k_2$	The price factor for SP to buy a unit of data from DP
$k_3$	The cost coefficient of data protection
$c$	The cost of data product development
$w$	Wholesale price for SP to buy a unit of data from DP
$n$	Scale of raw data obtained by DP from EU
$q$	Quality of data product developed by SP
$D$	Total end-market demand for data product
$a_0$	Initial market potential for data product
$q$	Quality of data products developed by SP
$r_0$	Critical point of revenue-sharing rate
${\pi }_1$	Profit function of SP
${\pi }_2$	Profit function of DP
${\pi }_3$	Consumer Surplus
${\pi }_c$	The total profit of the data supply chain under centralized pricing model
Decision variables
$\alpha$	Data incentive level
$\beta$	Data protection level
$p$	Data product pricing
$r$	Proportion of revenue sharing between SP and DP

.

For the DP and SP, it is necessary to invest a significant amount of funds in the construction of IT infrastructure at an early stage, which is regarded as sunk costs in our study. Because data products are similar to information products, we do not consider the replication cost of data products. The special cost structure of data and data products makes data supply chain pricing model different from traditional product supply chains.

3.2. Transaction Process of Data Supply Chain

3.2.1. Data Privacy, Privacy Awareness, and Data Collection

We consider the pricing of data products under different data protection and data incentive levels, assuming that users agree to the privacy policy of the platform when they accept the services provided by the platform. When the data are processed at the same technical development level, the higher the data scale, the higher the quality of the data products.

The amount of data shared by users is affected by both the level of data protection and the level of incentive. Many scholars have described the relationship between users’ privacy awareness, data protection, and data volume [2,20,51]. As described by Karimi Adl, et al. [52], we also assume there is a linear relationship between the amount of data shared by users and the level of data incentives and data protection. When the incentive level $\alpha$ given to users is higher, users are willing to share more data; when the level of data protection $\beta$ increases, the privacy loss perceived by users decreases, and the corresponding data sharing scale increases. The function is expressed as follows:

$$n=\alpha -k_1(1-\beta )$$

(1)

where $n$ indicates the scale of raw data that DP can collect, which is proportional to the incentive level $\alpha$ and protection level $\beta$; $\alpha$ indicates the unit data incentive given by DP to users, and when users share the unit of data, they can obtain economic compensation with value $\alpha$; $\beta$ indicates the level of data protection required by SP, which is mainly realized through DP’s investment in information technology and related to the privacy policy. The higher the data protection level, the smaller the probability of users’ privacy exposure or the loss; $k_1$ indicates the users’ awareness level of privacy risks, $k_1(1-\beta )$ representing the privacy loss brought about by users’ data sharing at the protection $\beta$.

Therefore, the consumer surplus is:

$${\pi }_3=\alpha n-k_1(1-\beta )$$

(2)

where $\alpha n$ represents the total incentive obtained by users, and $k_1(1-\beta )$ represents the privacy loss caused by users’ sharing the unit of data when the data protection level is $\beta$. It can be seen that consumer surplus is jointly affected by data incentive $\alpha$ and data protection level $\beta$.

3.2.2. Data Transaction Process

The DP sets the corresponding data incentive level after accepting the data protection requirements of the SP. Meanwhile, users decide whether to share data based on the level of data incentives $\alpha$ and data protection $\beta$. Then the DP sells the collected data to the service provider at the wholesale price $w$, thereby obtaining the corresponding revenue. Supported by Nget, et al. [53], we assumed that the wholesale price of data is related to the level of data protection, $w=k_2(1-\beta )$, that is, the higher the protection level of the DP for personal data, the lower the price the SP is willing to pay, as the data contain less customer personalization or preference information. Meanwhile, the corresponding data collection costs need to be paid in the data collection process, mainly including data incentive cost and data protection cost. The larger the scale of user data, the higher the corresponding collection cost. The higher the level of data protection, the higher the data protection costs. Therefore, the profit function of the DP is as follows:

$${\pi }_2=wn-\alpha n-k_3{\beta }^2$$

(3)

where $w$ indicates the wholesale price of a unit of data purchased by SP, which is affected by the level of data protection $\beta$; $\alpha n$ represents the total cost of data incentive when the scale of data collected by DP is $n$, $k_3$ represents the cost coefficient of data protection, and $k_3{\beta }^2$ indicates the total investment cost of data protection technology.

3.2.3. Data Product Supply Process

We assume that the SP only develops one data product, and the EU is willing to pay for high-quality products. Meanwhile, the data product demand is affected by product price and quality, and [15,54] and others have used similar expressions.

The demand function of the end market for data products is:

$$D=a_0-p+hn$$

(4)

where $a_0$ indicates the initial market potential for the data product. The market demand function shows that the total demand for data product. $D$ is positively correlated with the data product quality $q$, and negatively correlated with the price of the data product $p$. We assume that the quality of data products has a linear relationship with the scale of data and data product development technology, that is, $q=hn$. We propose that the relationship between data product development costs $c$ and data scale $n$ is nonlinear, given the development technology $h$, that is, $c=h{n}^2,h>0$. Therefore, the profit function of SP is as follows:

$$\begin{array}{l}{\pi }_1=pD-wn-h{n}^2\\ =p(a_0-p+hn)-k_2(1-\beta )n-h{n}^2\end{array}$$

(5)

where $p$ is the subscription fee for data product, $D$ is the total demand for data product in the end market,$w$ represents the wholesale price of data, and $h{n}^2$ is the development cost of data product. SP determines the data product subscription fee to maximize its own profit.

4. Pricing Mechanism of Data Supply Chain

In this section, we discuss three different pricing mechanisms—centralized pricing, decentralized pricing, and revenue sharing—and build different data pricing models to coordinate and optimize decision-making by stakeholders in the data supply chain. Meanwhile, we also provide some propositions that analyze the impact of users’ privacy risk awareness on data incentives, protection, and the best pricing method for each participant.

4.1. Analysis of Centralized Pricing Model

4.1.1. Data Product Pricing in Centralized Pricing Model

The centralized pricing model of a data product is a two-stage game. In the first stage, the SP has a better understanding of the EU’s demand for data product quality, and the SP proposes the required level of users’ privacy protection $\beta$ and the corresponding wholesale price of data $w$, and the DP makes decisions on the data incentive level $\alpha$. In the second stage, the SP decides on the data product subscription fee $p$ and then obtains the data product demand according to the demand function (4). Therefore, the total supply chain profit can be expressed as:

$${\pi }_c={\pi }_1+{\pi }_2=pD-h{n}^2-\alpha n-k_3{\beta }^2$$

(6)

According to the transaction process, we use backward induction: first, solve $p$ and then solve $\beta$ and $\alpha$, finally, obtain the optimal data incentive, data protection, and data production pricing.

Theorem 1.

There exists a unique solution to the centralized pricing model. The optimal decision for the DP is:

$${\alpha }^{\ast }=\frac{a_{0}h(k_1^2-2k_3)-2k_1{k}_3(4h-h^2+2)}{2\left(k_1^2+h^2{k}_3-4k_3-4h{k}_3\right)}$$

(7)

The optimal decisions for the SP are:

$${\beta }^{\ast }=\frac{\left(2k_1-a_{0}h\right)k_1}{2\left(k_1^2+h^2{k}_3-4k_3-4h{k}_3\right)}$$

(8)

$$p^{\ast }=\frac{a_0{k}_1^2+2h{k}_1{k}_3-4a_0{k}_3(1+h)}{2\left(k_1^2+h^2{k}_3-4k_3-4h{k}_3\right)}$$

(9)

Total supply chain profit of centralized pricing model is:

$${\pi }_c=\frac{a_0{}^2{k}_1^2-4k_3\left(a_0{}^2\left(1+h\right)-a_{0}h{k}_1+k_1^2\right)}{4\left(k_1^2+k_3(h^2-4h-4)\right)}$$

(10)

The consumer surplus is:

$$\begin{array}{l}{\pi }_3=\alpha n-k_1(1-\beta )\\ =\frac{k_3\left(2k_1-a_{0}h\right)\left(a_{0}h{k}_1^2-2k_3\left(a_{0}h+k_1\left(2+4h-h^2\right)\right)\right)}{2{\left(k_1^2+\left(h\left(h-4\right)-4\right)k_3\right)}^2}-k_1\left(1+\left(\frac{k_1\left(a_{0}h-2k_1\right)}{2\left(k_3\left(4h+4-h^2\right)+k_1^2\right)}\right)\right)\end{array}$$

(11)

See Appendix A for Proof of Theorem 1.

Proposition 1.

When users’ privacy risk awareness $k_1$ is satisfied $k_1{}^2>z$ with users’ privacy risk awareness $k_1$ increasing, the data incentive level $\alpha$ also increases. Otherwise, it is the opposite, where $z=k_3(2k_1{}^2-1)(4h+4-h^2)$.

See Appendix B for proof. The data incentive level $\alpha$ gradually increases with the increase in users’ privacy risk awareness $k_1$ and gradually decreases after $k_1$ reaching the highest point.

Proposition 2.

The subscription fee for data product $p$ is inversely proportional to the users’ privacy risk awareness $k_1$ and the cost coefficient of data protection $k_3$.

See Appendix B for proof. Proposition 2 indicates that a higher users’ privacy risk awareness level or a higher cost coefficient of data protection will always negatively affect the subscription fee for data product. Its impact on the subscription fee for the data product depends on a number of factors, including privacy policy, competitive environment, and so on. We use a numerical experiment in Section 5 to demonstrate the relationship between the users’ privacy risk awareness and the subscription fee for the data product.

4.1.2. Total Supply Chain Profit of Centralized Pricing Model

Proposition 3.

When the users’ privacy risk awareness  $k_1$ increases, it negatively affects the total profit of the data supply chain  ${\pi }_c$. If the technical level of data product development  $h$ increases, the total profit of the data supply chain also increases.

See Appendix B for proof. When $k_1$ increases, it leads to a decrease in the total scale of data in the supply chain and a reduction in the quality of the data product. It negatively affects the total profit of the data supply chain. When $h$ increases, the cost of the data product development increases, which leads to the increase in sales revenue of data products, and, finally, the total profit of the data supply chain increases.

4.2. Analysis of Decentralized Pricing Model

4.2.1. Data Product Pricing in Decentralized Pricing Model

In the decentralized data supply chain, we build a Stackelberg game model for data pricing. In the first stage, the SP as a leader, proposes data protection level $\beta$ and the data wholesale price it is willing to pay $w$; the DP decides the incentive level $\alpha$. In the second stage, the SP purchases data with a scale of $n$ and develops a data product of high quality and sells it to the EU at price $p$ to maximize its own profit. Therefore, the profit function of the SP can be expressed as:

$$\max {\pi }_1=pD-h{n}^2-wn$$

(12)

$$s.t\arg \underset{\alpha ,\beta }{\max }\{{\pi }_2=wn-\alpha n-k_3{\beta }^2\}$$

(13)

According to the transaction process among participants, we obtain the optimal data incentive, data protection, and data product pricing in the decentralized pricing model.

Theorem 2.

There exists a unique solution to the decentralized pricing model. The optimal decision for the DP is:

$${\alpha }^{\ast }=\frac{a_{0}h\left(k_1+k_2\right)}{h\left(h-4\right)\left(k_1-k_2\right)+8k_2}$$

(14)

The optimal decisions for the SP are:

$${\beta }^{\ast }=1-\frac{2a_{0}h}{h\left(h-4\right)\left(k_1-k_2\right)+8k_2}$$

(15)

$$p^{\ast }=\frac{2a_0(2k_2+h(k_2-k_1))}{h\left(h-4\right)\left(k_1-k_2\right)+8k_2}$$

(16)

The profits of SP and DP are:

$${\pi }_1=\frac{h^2{a}_0^2{\left(k_1-k_2\right)}^2-k_3{\left(h\left(h-4\right)\left(k_1-k_2\right)+8k_2-2h{a}_0\right)}^2}{{\left(h\left(h-4\right)\left(k_1-k_2\right)+8k_2\right)}^2}$$

(17)

$${\pi }_2=\frac{a_0^2\left(h\left(k_2-k_1\right)+2k_2\right)}{h\left(h-4\right)\left(k_1-k_2\right)+8k_2}$$

(18)

The consumer surplus is:

$${\pi }_3=\frac{a_{0}h\left(a_{0}h{k}_2^2-h{k}_1^2\left(2h+a_0-8\right)-2k_1{k}_2\left(4h+8-h^2\right)\right)}{{\left(h\left(h-4\right)\left(k_1-k_2\right)+8k_2\right)}^2}$$

(19)

See Appendix A for proof of Theorem 2.

Proposition 4.

The data protection and data product subscription fee  $p$ are inversely proportional to the users’ privacy risk awareness  $k_1$.

See Appendix B for proof. When the users’ privacy risk awareness $k_1$ increases, the SP decreases the data protection level $\beta$ and the data product subscription fee $p$. The cost of data protection will decrease as a result of the decrease of $\beta$ and $p$.

4.2.2. Analysis of DP, SP Profit of Decentralized Pricing Model

Proposition 5.

When the users’ privacy risk awareness  $k_1$ increases, the profit of DP will decrease. The profit of SP increases if the level of data product development  $h$ increases.

See Appendix B for proof. As the users’ privacy risk awareness $k_1$ increases, it negatively affects the profit of the DP. With the increase in the level of data product development $h$, the quality of the data product increases, and the profit of the SP increases.

4.2.3. Consumer Surplus in Decentralized Pricing Model

Proposition 6.

When the users’ privacy risk awareness  $k_1$ increases, the consumer surplus,  ${\pi }_3$, will decrease.

See Appendix B for proof. With the increase of $k_1$, the cost of users’ privacy increases, and the consumer surplus decreases, it indicates that higher users’ privacy risk awareness negatively affects the consumer surplus.

4.3. Analysis of Revenue Sharing Pricing Model

4.3.1. Data Product Pricing in the Revenue Sharing Model

Revenue sharing is an effective supply chain pricing mechanism. In this section, we establish a revenue-sharing model of data pricing to coordinate the data supply chain, where the most important issue is to determine the revenue-sharing ratio between the DP and SP. Suppose that the revenue-sharing rate between DP and SP is $r$, namely, the SP returns the revenue obtained by selling data products $r$ to the DP, $r\in (0,1)$. In the first stage of the Stackelberg game, the SP proposes the required data protection level $\beta$ and the data wholesale price $w$ it is willing to pay; the DP decides the incentive level $\alpha$. In the second stage, SP processes the data of the purchased scale $n$ and develops a data product with quality $q$ and sells it to the EU at the price $p$. Therefore, the profit of the SP is as follows:

$${\pi }_1=(1-r)pD-h{n}^2$$

(20)

The profit of DP is:

$${\pi }_2=rpD-\alpha n-k_3{\beta }^2$$

(21)

Then the model can be rewritten as:

$$\max {\pi }_1=(1-r)pD-h{n}^2$$

(22)

$$s.t\arg \underset{\alpha ,\beta }{\max }\{{\pi }_2=rpD-\alpha n-k_3{\beta }^2\}$$

(23)

According to the transaction process among participants, we first resolve the expression of the optimal data incentive, data protection, and data product pricing with respect to $r$. Then, we determine the optimal revenue-sharing ratio $r^{\ast }$ through the DP and SP negotiation. Finally, we resolve the optimal decision variables.

Theorem 3.

There exists a unique solution to the centralized pricing model. The optimal decision for the DP is:

$${\alpha }^{\ast }=\frac{a_{0}h{k}_1^2-2k_3\left(a_{0}h+2k_1+4h{k}_1-h^2{k}_1\right)}{2\left(k_1^2+h^2{k}_3-4h{k}_3-4k_3\right)}$$

(24)

The optimal decisions for the SP are:

$${\beta }^{\ast }=\frac{k_1\left(2k_1-a_{0}h\right)}{2\left(k_1^2+h^2{k}_3-4h{k}_3-4k_3\right)}$$

(25)

$$p^{\ast }=\frac{a_0{k}_1^2+2k_3\left(h{k}_1-2a_0-2a_{0}h\right)}{2\left(k_1^2+h^2{k}_3-4h{k}_3-4k_3\right)}$$

(26)

The optimal revenue sharing ratio is:

$$r^{\ast }=\frac{4a_0{k}_3+8k_1{k}_3-a_0{k}_1^2-2h{k}_1{k}_3}{4a_0{k}_3+4a_{0}h{k}_3-a_0{k}_1^2-2h{k}_1{k}_3}$$

(27)

The profit of SP and DP are:

$${\pi }_1=\frac{a_0{k}_3\left(2k_1-a_{0}h\right)}{k_1^2+k_3\left(h^2-4h-4\right)}$$

(28)

$${\pi }_2=\frac{a_0{}^2{k}_1^2-4k_3\left(a_0{}^2+a_0{k}_1\left(2-h\right)+k_1^2\right)}{4\left(k_1^2+k_3\left(h^2-4h-4\right)\right)}$$

(29)

The consumer surplus for end users is:

$${\pi }_3=\frac{\left(2k_1-a_{0}h\right)((k_1^4+h^2{k}_1^2{k}_3-4h{k}_1^2{k}_3-4k_1^2{k}_3)+\left(a_{0}h{k}_1^2{k}_3-2a_{0}h{k}_3{}^2+4k_3{}^2{k}_1+8k_3{}^{2}h{k}_1-2k_3{}^2{h}^2{k}_1\right))}{2{\left(k_1^2+h^2{k}_3-4h{k}_3-4k_3\right)}^2}-k_1$$

See Appendix A for proof of Theorem 3.

4.3.2. Profit Analysis of DP and SP in the Revenue-Sharing Model

Proposition 7.

When the level of users’ privacy risk awareness  $k_1$ increases, the profit of DP,  ${\pi }_2$ will increase, while the profit of SP,  ${\pi }_1$ will decrease.

See Appendix B for proof. With the increase of $k_1$, the scale of raw data provided by the DP increases, with more revenue from data dealing, thus the profit of the DP increases. Nevertheless, if $k_1$ increases, thus increasing the difficulty of data collection and raising the total cost, the profit of the SP decreases.

5. Numerical Experiments

In this section, a numerical study is carried out to further analyze the data pricing mechanisms of the data supply chain, and to gain insights from different data product pricing models. The relevant parameters are set as follows: $k_2=0.6,k_3=0.3,h=0.5,a_0=2$. Considering that the centralized pricing strategy has few application scenarios in real life, this section focuses on relevant decisions under decentralized pricing and revenue-sharing pricing models. Considering that the centralized pricing mechanism is rarely used in supply-chain practice scenarios, we mainly discuss the decisions of decentralized pricing and revenue sharing and their impacts on stakeholders.

The numerical simulation in our paper is carried out in the Windows 10 system and Wolfram Mathematica 12.1 software. See Appendix C for detailed code.

5.1. Analysis of Total Profit of Data Supply Chain

Firstly, we analyze the total profit of the data supply chain under the centralized pricing and decentralized pricing models. As can be seen from Figure 2, with the increase in users’ privacy risk awareness, the total profit of the data supply chain gradually decreases. This is because, given data protection and data incentives, the increase in users’ privacy risk awareness leads to a reduction in the total scale of data, and then leads to a reduction in the scale of data flowing in the data supply chain. At the same time, the reduction in the data scale leads to a decrease in the quality of data products, and, finally, the total profit of the data supply chain decreases. Meanwhile, no matter what the users’ privacy risk awareness level is, the total profit of the centralized-pricing data supply chain is always higher than that of the decentralized pricing, that is, adopting the centralized pricing mechanism can realize the coordination of the data supply chain, and, finally, achieve the purpose of maximizing the profit of the data supply chain.

5.2. Analysis of Data Incentive

As shown in Figure 3, in the revenue-sharing pricing model, with the increase in users’ privacy risk awareness level, the data incentives first increases and then decreases; in the decentralized pricing model, with the increase in users’ privacy risk awareness, the level of data incentives increases. The incentive level of revenue sharing before point M (shown in Figure 3) is higher than that under decentralized pricing, and after point M, it is the opposite.

Given the level of data protection, with the increase in users’ privacy risk awareness, the privacy cost caused by data sharing also increases, resulting in a decrease in the amount of users’ shared data. Therefore, in order to obtain more data, certain data incentives are set by the DP. With the increase in users’ privacy risk level, the incentive level also increases. Meanwhile, in the revenue-sharing model, it reaches the highest value at point A (shown in Figure 3). After point A, the incentive level decreases with the increase in users’ privacy risk awareness because the benefits of data incentives cannot cover the cost of data incentives.

Before point M (shown in Figure 3), the incentive level under revenue sharing is higher than that under decentralized pricing. After point M, if the users’ privacy risk awareness continues to increase, the DP’s pursuit of maximizing its own profits under decentralized pricing increases the level of data incentives; meanwhile, under data sharing, higher data incentives bring higher incentive costs and contribute less to profits. Therefore, the level of data incentives under decentralized pricing is higher than under revenue-sharing pricing.

5.3. Analysis of Data Protection

As shown in Figure 4, as the level of users’ privacy risk awareness increases, the data protection level of the decentralized pricing and revenue sharing shows opposite trends.

Given the data incentives, the level of data protection tends to decrease with the increase in personal privacy risk awareness. Because under decentralized pricing, all participants seek to maximize their own profits, and lower levels of protection mean that the scale of data is lower, which also reduces the cost of data product development, which is more beneficial to the SP. In the revenue-sharing pricing model, with the increase in users’ privacy risk awareness level, the level of data protection also increases because both parties have reached an agreement on the sharing ratio—that is, the overall revenue is considered when making relevant decisions. At this time, a higher level of users’ perception of privacy risk indicates that users are more concerned about personal data privacy. Under a given data incentive level, in order to obtain more user data, the corresponding level of data protection is increased to reduce users’ privacy concerns.

According to the Forrester survey (https://www.forrester.com/blogs/straight-from-the-source-collecting-zero-party-data-from-customers/ (accessed on 30 July 2020)), some customers intentionally and proactively share their data with the data provider, data platform, or APP for personalized experiences, which is called “zero-party data”. If the users’ privacy risk awareness level is almost zero or zero privacy [55], it is beneficial for the DP and SP to reduce the cost of data protection and data incentives.

5.4. Analysis of Data Product Price

As shown in Figure 5, with the increase in users’ privacy risk awareness, the price of data products shows a decreasing trend. Given the users’ data protection level and data incentive level, with the increase in the users’ privacy risk awareness level, the scale of data that the DP can collect gradually decreases, and the scale of data transactions between the DP and SP decreases, resulting in the reduction in the quality of data products and, ultimately, the reduction in the price of data products. The price of data products is affected by the DP incentives level.

5.5. Analysis of the Profit of SP and DP

As shown in Figure 6, the profit of the SP gradually decreases with the increase in users’ privacy risk awareness level. With the increase in users’ privacy risk awareness, the total scale of data collected by the DP decreases. Under decentralized pricing, DP’s corresponding data transaction revenue decreases, resulting in a downward trend in the final overall income. In the case of revenue sharing, a reduction in the data scale will lead to the reduction in the data product quality and its price, and the final revenue will show a decreasing trend. Meanwhile, SP’s revenue under the decentralized pricing model is higher than the revenue under the sharing pricing. It also indirectly shows that the revenue-sharing model is not always beneficial, and the coordination of the supply chain may not always be achieved. Therefore, the decentralized pricing model is more suitable for the SP.

Figure 7 demonstrates that with the increase in users’ privacy risk awareness, the DP’s profits show different trends in the decentralized pricing and revenue-sharing models. With the increase in users’ privacy risk awareness, the scale of users’ shared data decreases, and DP’s data transaction revenue decreases under decentralized pricing. In the revenue-sharing model, their profits are on the rise due to the influence of the proportion of revenue sharing. In other words, adopting the revenue-sharing pricing method is more beneficial to the DP, and allows for obtaining benefits higher than under decentralized pricing.

5.6. Analysis of Consumer Surplus

As shown in Figure 8, given the level of data incentives and data protection, consumer surplus decreases with the increase in users’ privacy risk awareness. The increase in user privacy costs results in a reduction in consumer surplus. At the same level of privacy risk awareness, consumer surplus under revenue sharing is higher than under decentralized pricing. Under the condition of decentralization, all participants seek to maximize their own profits; the protection level of users under decentralized pricing is low, while the data protection level in revenue-sharing pricing is high, so the privacy cost to users is relatively small. Thus, the consumer surplus of revenue sharing is higher than that of decentralized pricing.

6. Conclusions

Based on the value-added process of user data in the data supply chain, we have analyzed the data transaction process between participants and put forward three different data pricing models: centralized pricing, decentralized pricing, and revenue sharing. We have calculated the optimal data incentives, data protection, and data pricing level, and discussed the impact of users’ privacy risk awareness level on data incentives, data protection, and data product pricing under the three data-pricing strategies, as well as the profit of the DP, SP, and consumer surplus. The main conclusions of our study are as follows:

Firstly, the data-pricing problem is studied from the perspective of the data supply chain composed of the EU, DP, and SP; the optimal data incentives, data protection level, and data products pricing are analyzed under different pricing mechanisms. For the DP, it is more advantageous to adopt the revenue-sharing model, while for the SP, the decentralized pricing model is a better choice, which also shows that the revenue-sharing model does not always coordinate the supply chain. At the same time, our results show that in the revenue-sharing pricing mechanism, the DP should subsidize the SP to achieve a reasonable distribution of profits, so that the profit of both sides can reach the optimal state, and, finally, achieve the coordination of the data supply chain.

Secondly, we find that in the decentralized pricing and revenue-sharing models, the users’ privacy risk awareness has different effects on data protection and data incentives. The higher the users’ privacy risk awareness level, the lower the data protection level is in decentralized pricing, but there is an upward trend in the revenue-sharing pricing. Meanwhile, the higher the users’ privacy risk perception level, the higher the data incentives are under decentralized pricing, but in revenue sharing, there is a trend of rising first and then falling.

Finally, the impact of users’ privacy risk awareness on the participants in the data supply chain is studied. For the EU, the higher the level of users’ privacy risk awareness, the lower the consumer surplus is; for the SP, whether it adopts decentralized pricing or revenue sharing pricing, its profit tends to decline with the improvement of users’ privacy risk awareness; for the DP, the higher the level of users’ privacy risk awareness, the lower the profit is under decentralized pricing, and the higher the profit is under the revenue-sharing model.

The limitation of our study is that it does not consider the impact of differentiated personal privacy on data protection and data incentives, nor does it consider different amounts of data shared by different users. Our study does not evaluate where these pricing mechanisms fall within the national and regional data protection and privacy laws. Further research is needed to reconcile this economic development with the law. This is because there are likely to be cross-border data issues related to data within supply chains. These problems need to be studied in the future.