Channel Integration Through a Wireless Applet and an E-Commerce Platform

Abstract

We study an online–merge–offline (OMO) system that integrates a retailer’s online and offline channels through an e-commerce platform and a wireless applet. The customers in the online channel are generated through paid advertising in an e-commerce platform, while the offline channel is a regional retail chain. The OMO system omnichannelizes sole-channel customers from either channel by converting them into omnichannel ones with the wireless applet and then providing them online and offline options at each touch point along the shopping journey. The prices in the OMO system across both channels are uniform. To validate the effectiveness of this new omnichannel system, we construct a legacy system that maintains separate online and offline channels with independent customer populations. Using the legacy system as a benchmark, we assume the OMO system has arbitrary omnichannelization rates of the customers flowing into the two channels. We analyse the perfect OMO system which has all the customers omnichannelized, and show its advantage over the legacy system. We then numerically find that if the omnichannelization rates in the OMO system are general then it is most efficient when products are either highly digital or highly nondigital.

1. Introduction

Recently, e-commerce giants Amazon and Alibaba were both aware that their online sales growth had become stagnant and turned to brick-and-mortar retailing for further expansion. In 2017, Amazon acquired Whole Foods [1]. In China, Alibaba implemented a large-scale omnichannel retailing strategy, new retail. The founder of Alibaba, Jack Ma, coined the term new retail to refer to the integration of online, offline, logistical, and data-based approaches across the retail value chain [2]. Despite the continuous effort of these Internet giants and their offline counterparts, online retail sales was still about 21% of the total retail sales in the US [3] and 27% in China in 2022 [4], dated back to the days during the pandemic, which drove more customers online.

Alibaba’s first attempt in new retail was its acquisition of Intime, a multibillion dollar department chain selling apparel and cosmetics, in 2017. Having 29 department stores and 20 malls across the country, Intime also has an e-commerce arm through Alibaba’s B2C subsidiary, Tmall. Tmall, together with its C2C sibling, Taobao, constitutes an e-commerce model that is different from Amazon or JD.com. Tmall is an e-commerce platform that houses millions of sellers who have to pay the platform advertising fees to generate customers; therefore, online users must become Tmall’s customers before becoming a customer of a specific retail store on the platform. Intime’s regional retail stores and e-commerce store in Tmall are merely independent in organization, marketing, pricing, and logistics. Their customer populations can also be largely different for the following reasons: (1) there is still a substantial number of customers who do not shop online, and even for those online shoppers, they may only buy specific products; (2) a regional retail chain like Intime only has a limited population of potential customers in the proximity of a retail store, while Tmall has potential customers from the entire Internet. For these simple reasons, the difference in the populations of Intime and Tmall is estimated at the magnitude of one hundred times. In other words, an online customer shopping in Intime’s online store housed in Tmall is unlikely a customer in the neighbour of Intime’s physical store, and a customer of Intime’s physical store has extremely low chance to be brought to its online store in Tmall. Alibaba’s acquisition of Intime aims at solving this disconnection through customer omnichannelization, which converts an online-only or offline-only customer into a customer who can shop both online and offline.

Omnichannelizing sole-channel customers at Intime is achieved via a wireless applet, Miaojie. This applet acts as the main portal of Intime’s offine stores to engage with consumers directly and digitize its operations. Under any circumstance, a local customer around the Intime store is invited to register with the Miaojie applet and becomes an omnichannel user, who can make online orders from the local store or pay a visit to the offline store. On the other hand, Tmall attracts online customers via advertising and directs them into online stores of retailers, who need to pay Tmall according to the customer traffic. In the case of Intime, if a product is popular in its online store, it might also be presented in Intime’s offline store, which is signaled in the online store. Hence, customers may switch to Intime’s offline stores by reaching Miaojie as this is the unique Internet portal of Intime’s offline stores. In such a function, Miaojie plays the role of webrooming for those offline stores who in return serve as the warehouses for the online channel [5]. In addition, Intime advocated to have the same price for a product sold in both online and offline, which can induce some online customers to shop offline. The effectiveness of Intime’s omnichannel initiative was proven through membership and sales data. Within 12 months after Intime digitized its own membership and integrated it with Alibaba’s membership, Miaojie had approximately 3 million registered members [6]. In 2020, it was reported to have accumulated approximately 20 million digitized members. In the same year, half of Intime’s top 20 cities in terms of sales indeed had no Intime physical stores, but their customers could order from stores in other cities.

While digitizing customers is a prerequisite for customer omnichannelization, product digitization is also required to achieve this goal, that is, to visualize the product through digital mediums such as photos or videos posted on websites. The magnitude that a customer can grasp the product attributes through digital mediums versus personal inspection is termed product digitization. Customer omnichannelization and product digitization are the two caveats in the OMO strategy.

In an online–merge–offline (OMO) approach, Alibaba and Intime have pioneered a new omnichannel strategy. As the OMO system offers two channel choices to omnichannelized customers, it seems to be superior to the traditional legacy system that maintains separate online and offline channels. Nevertheless, Intime adopts uniform pricing across its online and offline channels to counter intracompany price competition, which seems a sub-optimal rule. This raises the question whether the gain from omnichannelization can offset the loss from uniform pricing. Since omnichannelizing sole-channel customers is a new concept first raised and analyzed in this paper, we build a legacy system and an OMO system to demonstrate its effectiveness. The online and offline channels of the legacy system operate independently, wherein the two customer streams are fully independent. The OMO system adopts the two channels of the legacy system and sets a common product price across them. The process of omnichannelization allows some or all customers from the two channels to purchase from both channels. In this paper, we first model the legacy and the omnichannel systems to depict the omnichannelization process and then address the following research questions:

(1)

How does the OMO system influence market coverage with respect to the legacy system?

(2)

Does the OMO system outperform the legacy system, and if so, when?

(3)

How does product digitization impact the value of ommichannelization?

We model the two systems starting with constructing customer utility functions based on their hassle costs to shop online and offline, as well as the level of product digitization. We first find that if all the customers in the system are omnichannelized, it expands sales opportunities and generates higher profit than the legacy system, and in most cases, it generates higher sales. We then numerically solve the OMO system with the general customer omnichannelization rates, finding that the total profit and total sales of the OMO system increases with the omnichannelization rates, while this system is most efficient when products are either highly digital or highly nondigital. However, the OMO system may perform worse than the legacy system if the omnichannelization rates are low. It indicates that high omnichannelization rates and choosing appropriate products are necessary for the OMO system to be effective. The last point well explains the new retail implementation between Intime and Alibaba, wherein popular products in the online stores were chosen to display in the physical stores and the other way round, but not every product at Intime was chosen to sell both online and offline.

This paper is structured as follows. Section 2 provides a literature review. Section 3 models the legacy system, while Section 4 models the OMO system. Section 5 analyses the OMO model under special conditions, and Section 6 numerically solves the general OMO model. Section 7 concludes the paper and outlines some limitations in the model settings.

2. Literature Review

In the past two decades, the retail business has evolved from multichannel retailing to omnichannel retailing [7], and voluminous empirical and theoretical research has appeared in this area. In this section, we outline the main progress and position our work in the literature.

Multichannel retailing is a natural execution when Internet and mobile technology are applied in the retail business. A list of value propositions for e-commerce versus traditional physical retail is proposed in [8]. Chiang et al. (2003) [9] modelled the competition between a manufacturer’s direct online channel and its retailer, and they proved the value of adding an online channel. Brynjolfsson and Smith (2000) [10] empirically examined the competition between an online store and a traditional physical store and provide support for these theoretical insights. Chintagunta et al. (2012) [11] quantified online and offline transaction costs that affect households’ channel choices regarding grocery purchasing. They provide monetary metrics for several types of transaction costs, including travel time and transportation cost, in-store shopping time, item-picking cost, and inconvenience cost. Cavallo (2017) [12] compared online and offline prices at a number of retailers, finding a common price 72% of the time. Wiesel et al. (2011) [13] analysed customer funnel progression when customers choose between an online channel and an offline channel in the context of cross-channel migration. The legacy system framed in this paper is a simple multichannel system, which serves as a benchmark for the focal omnichannnel system.

A number of papers empirically examine how an online store can drive sales to an existing offline store [14] or how an offline store can drive sales to an online store [15,16,17]. Dzyabura et al. (2019) [18] identified discrepancies when products are evaluated online vs. offline, showing that e-commerce is suitable for digital goods while nondigital goods are better suited for physical stores. These papers consider multichannel retailing with or without cross-channel customer migration, and online and offline channels do not directly interact. For a firm with online and offline channels, it is suggested that the two channels have independent prices if there is no customer migration but uniform prices if there is low customer migration [19]. The uniform or independent prices in the online and offline channels are interpreted as the evidence of different levels of digital and nondigital attributes of the product [20]. Pricing is the foremost decision in both multichannel and omnichannel systems in this paper too. Moreover, our systems are unique in that the online customers are endogenized through advertising, which is typical in an e-commerce platform.

Channel integration, also termed omnichannel retailing, is a natural extension of the multichannel retailing strategy. Evidence was provided that such integration leads to a competitive advantage over channel cannibalization [21]. In the context of online-first omnichannel operations, retailers have tested showrooming, which allows customers to experience products in physical stores while making purchases online [22,23]. However, refs. [24,25] both empirically verified that showrooming expands market coverage and promotes products in the long run. Buy-online/pickup-in-store (BOPS) is another online-to-offline strategy in which a customer completes a purchase online and picks up the product in a store [26,27]. Ref. [28] found that customers may choose research-online-purchase-in-store (ROPS, aka. webrooming), which reduces online transactions but increases store sales and traffic. Motivated by the same practice at Intime as introduced in Section 1, ref. [5] investigated the benefits of physical stores as fulfilment centres for an online channel in the competitive environment.

Webrooming became prevalent because customers can search the product information with wireless technologies without the constraints of time and location. A series of empirical papers has examined the concept of webrooming from customers’ perspectives. It is believed that webrooming makes customers feel confident and informed before purchasing [29,30]. In [31], it is suggested that customers’ anticipated utility for purchasing online or offline can be predicted based on some antecedents of webrooming.

With substantial understanding about BOPS, showrooming and webrooming, it is unclear to what extent an industry or a company should develop omnichannel retailing. While between-retailer across-channel omnichannel integration is emerging [32], empirical and modelling studies on omnichannel retailing are mainly on the within-retailer across-channel effects of the customers’ purchase journeys on the firms’ aggregate outcomes such as sales and profit. A customer journey is generally divided into three stages: search, purchase, and post-purchase [21,33], while customers at each touch point can be modulated by firms [34] with technologies. Based on 50 empirical papers, Timouni et al. (2022) [35] concluded the impact of various omnichannel strategies, including adding clicks to bricks, adding bricks to clicks, adding mobile channel, cross-channel integration, on the online and offline channels, as well as the retailer as a whole. In their study, customers are segmented and can take different journeys, while retailers can be single channel, multichannel, and omnichannel.

In another meta study, Neslin (2022) [36] proposed the concept of the omnichannel continuum to characterize channel integration extensiveness. Combining customer journey with channel choice, he suggested four omnichannel strategies, among which “unconnected” and “complete” are on the two extremes, and in between are “horizontal” and “vertical”. Integration can be achieved horizontally by integrating channels at a decision stage or vertically by integrating channels as customers move along the stages of their purchase journey. In Neslin’s classification, BOPS and showrooming are vertical strategies because customers are encouraged to search fproduction information in one channel but purchase the product in another. Ref. [37] presented a horizontal strategy at the post-purchase stage, whereby retailers provide customers cross-channel delivery services.

We identify that the omnichannel strategy between Intime and Tmall is a complete integration strategy, since the integrated system provides customers online and offline options at each stage along the purchase journey. These problem characteristics automatically distinguish this research from the existing literature in omnichannel integration. As pointed out in [38], omnichannel operations have revived retailer stores, for which we have provided a practical case to support their observations. In the retail sector, omnichannels and digital transformation are becoming an industrial norm [39,40] and the incident of COVID-19 has magnified this trend. While omnichannel retailing is still developing [41], we present a new initiative wherein omnichannelizing customers is the key instrument in constructing this system.

3. Legacy Channel Systems

In the legacy system, the retailer operates an e-commerce branch by setting up a store on a B2C platform and acquiring customers through advertisements. Online customers are drawn from the entire Internet with no temporal or geographic constraints, while orders are fulfilled via one centralized warehouse. The retailer also operates an offline channel with a regional retail chain, offering services to neighbouring customers. We establish a series of assumptions to create a benchmark. We first have the following:

Assumption 1.

The online and offline channels in the legacy system are independent in organization but they sell the same product, and the customers entering the two channels are independent, sole-channel customers.

As introduced earlier, this system can well approximate the motivating case, in which the e-commerce platform induces customers to visit the retailer’s online store who mainly faces competition from its peers in the platform than its offline counterpart. The online and offline channels each make individual choices to maximize their own profitability. Other than the motivating case, this benchmark also applies when a new product enters the market by selling in both the online and offline channels, while customers in one channel may not know the existence of the other channel [20].

As suggested in [8], we further have the following:

Assumption 2.

Each online or offline customer entering the system will make the purchase decision solely based on the economic utilities after considering other non-economic value propositions along the customer journey.

To focus on the main research questions, this benchmark system based on Assumptions 1 and 2 captures pricing, advertising, and channel structures.

We depict how potential customers flow within the two channels in Figure 1. The offline market is determined by the customers in the proximity of a retail store, so its size is assumed fixed and scaled to be one unit of mass. The online customers are generated through online advertising which is interactive and requires the direct participation of potential customers, whereby the customer amount is determined by the advertising amount paid to the platform who hosts the retailer. The current online advertising mechanisms in Tmall are sponsor search and display advertising, which are both complicated and implemented through specific technologies. In the view of [42], any advertising mechanism could be regarded as a specific technology, and the associated cost is assumed convex increasing in the amount of advertised customers. To be tractable, we assume that the cost to generate x potential online customers is $a{x}^2$, where a is termed the cost of online advertising hereafter. As explained, ($a,x$) are converted from the cost and the number of reached customers associated with a corresponding advertising technology.

We use the superscript L to denote the legacy system. The subscript i denotes the retail channel, where $i=1$ is the online channel, and $i=2$ is the offline channel. The focal product’s selling price in channel i is denoted as $p_i$, and the corresponding demand is denoted as $D_i$. It is noted that $D_1$ is decided by $p_1,x$ while $D_2$ is decided by $p_2$.

To derive the demand functions, $D_1,D_2$, we first need to construct the customer utility functions. Following [9], the product’s valuation, when purchased offline, is $\nu$, which is assumed to be heterogenous across all potential customers and is uniformly distributed between 0 and 1. The product’s digitization coefficient, $\theta$(<1), represents the discount on the product value obtained by potential customers that purchase online. High $-\theta$ products are digital goods whose value easily translates to an online setting, while low $-\theta$ products are nondigital, requiring customers’ direct touch and feeling to assess their appropriateness and value. There are also hassle costs $h_i$ associated with each channel i. Online hassle costs include searching online, waiting for delivery, and the cost and effort associated with shipping and returning. Offline hassle costs include driving to the store, waiting to purchase an item, and transporting the item back home.

To derive the demand functions, we first define the utility function, $u_i$, of a potential customer buying one product unit from channel i:

$$\begin{array}{ccc}\hfill u_1& =& \theta \nu -h_1-p_1\hfill \end{array}$$

(1)

$$\begin{array}{ccc}\hfill u_2& =& \nu -h_2-p_2\hfill \end{array}$$

(2)

The product’s digitization coefficient, $\theta$, affects the product’s value in the online channel. The hassle costs, $h_1$ and $h_2$, might be customer-specific, but without loss of generality, we assume that both $h_1$ and $h_2$ are constants between zero and one. According to [43], if the hassle costs are heterogeneous among the customers, their randomness can be shifted to the product value by redefining $\nu$ and $\theta \nu$ for the offline and online products, respectively. Equivalently, we can convolute the randomness generated by the production valuation and hassle cost associated with each customer, although this often results in an intractable form. Therefore, we have to impose simplification assumptions about the base models in (1) and (2) to conduct tractable analysis.

We will not explicitly model the costs arising from sourcing, warehousing and delivery in the online and offline channels. In our motivating case, the delivery and warehousing services in Alibaba’s e-commerce business are handled by third-party logistics companies, so the associated costs could be absorbed into the online hassle cost; in particular, this cost could be directly charged to the customers beyond price. The purchase costs for the same product in the online and offline channel are also assumed the same and normalized to zero despite possibly different procurement routes. Adding these sourcing and logistics costs will significantly complicate our model settings, so we choose to simplify out these factors to focus on the main theme of this paper.

With the above assumptions, the online and offline channels independently determine their retail prices $p_1$ and $p_2$, respectively. A potential online customer buys a product in the online channel only if $u_1\ge 0$. As $\nu$ is uniformly distributed in $[0,1]$, it generates demand of $(1-{\displaystyle \frac{h_1+p_1}{\theta }})$, we first have the following:

Observation 1.

The online channel of the legacy system is effective only if $h_1<\theta$.

It is seen from (1) that if $h_1>\theta$, then $u_1<0$ for any $\nu$ and the online channel is inactivated.

The online channel’s demand, $D_1^L(p_1,x)$, is given by

$$\begin{array}{c}\hfill \begin{array}{c}\hfill D_1^L(p_1,x)=\left\{\begin{array}{cc}x(1-{\displaystyle \frac{h_1+p_1}{\theta }})\hfill & \mathrm{if}{h}_1\le \theta \hfill \\ 0\hfill & \mathrm{otherwise}.\hfill \end{array}\right.\end{array}\end{array}$$

(3)

The online channel optimizes $p_1$ and x to maximize its profit under $h_1\le \theta$ by solving:

$$Ma{x}_{p_1\ge 0,x\ge 0}{\pi }_1^L(p_1,x)=p_1{D}_1^L(p_1,x)-a{x}^2=x{p}_1(1-{\displaystyle \frac{p_1+h_1}{\theta }})-a{x}^2.$$

(4)

When $h_1<\theta$, the optimal solutions to (4) are $x^{\ast L}={\displaystyle \frac{{(\theta -h_1)}^2}{8a\theta }}$ and $p_1^{\ast L}={\displaystyle \frac{\theta -h_1}{2}}$. Correspondingly, its sales quantity is $D_1^{\ast L}={\displaystyle \frac{{(\theta -h_1)}^3}{16a\theta }}$. If $h_1>\theta$, the optimal solution of x in (4) is zero, and the online channel is void.

Likewise, a potential offline customer buys the product when $\nu >h_2+p_2$. As the population size of potential customers in the offline channel is normalized to 1, the corresponding demand is

$$\begin{array}{ccc}\hfill D_2^L\left(p_2\right)& =& 1-h_2-p_2.\hfill \end{array}$$

(5)

The offline channel optimizes $p_2$ to maximize the corresponding profit:

$$Ma{x}_{p_2\ge 0}{\pi }_2^L\left(p_2\right)=p_2(1-h_2-p_2).$$

(6)

The optimal solution is $p_2^{\ast L}={\displaystyle \frac{1-h_2}{2}}$. Correspondingly, the sales quantity for any $h_2\in [0,1]$ is $D_2^{\ast L}={\displaystyle \frac{1-h_2}{2}}$.

As the online channel exists only when $h_1\le \theta$, Lemma 1 compares the prices of the online and offline channels with this condition.

Lemma 1.

When $h_1\le \theta$, (1) if $h_2-h_1>1-\theta$ then $p_1^{\ast L}>p_2^{\ast L}$; (2) if $h_2-h_1=1-\theta$ then $p_1^{\ast L}=p_2^{\ast L}$; and (3) if $h_2-h_1<1-\theta$ then $p_1^{\ast L}<p_2^{\ast L}$.

All proofs appear in Appendix A. Lemma 1 analytically supports the empirical observation of [12] that the online prices of multichannel retailers could be higher, equal to, or lower than their offline prices. Our model setting finds that channel hassle costs and the product digitization coefficient jointly determine this relationship.

By comparing $D_1^{\ast L}$ with $D_2^{\ast L}$, we observe that if the cost of online advertising is low, then online sales exceed offline sales, and vice versa. The market outcome of the online channel is primarily determined by its hassle cost and product digitization and then by the cost of advertising; if $h_1>\theta$ or a is large, this channel may be of little value to the retailer.

If $h_1\le \theta$, the total profit of the system is:

$${\pi }^{\ast L}={\pi }_1^L(p_1^{\ast L},x^{\ast L})+{\pi }_2^L\left(p_2^{\ast L}\right)={\displaystyle \frac{{(\theta -h_1)}^4}{64a{\theta }^2}}+{\displaystyle \frac{1}{4}}{(1-h_2)}^2.$$

(7)

If $h_1>\theta$, the total system profit is ${\pi }^{\ast L}={(1-h_2)}^2/4$. The total sales quantity of the two channels is $D^{\ast L}=D_1^{\ast L}+D_2^{\ast L}={\displaystyle \frac{{(\theta -h_1)}^3}{16a\theta }}+{\displaystyle \frac{1-h_2}{2}}$ if $h_1\le \theta$; otherwise, $D^{\ast L}=D_2^{\ast L}={\displaystyle \frac{1-h_2}{2}}$. We use ${\pi }^{\ast L}$ and $D^{\ast L}$ as key performance measures of the legacy system in comparing this system with the same in the presence of customer migration in the rest of this section and the OMO system in Section 4.

4. The Online–Merge–Offline Channel System

The wave of online expansion into offline initiated by Internet giants tries to build omnichannel systems that can efficiently utilize the online and offline channels. The first step in these efforts is to omnichannelize the customers, i.e., to convert sole-channel users into omnichannel users. However, this endeavour might be constrained if some consumers are non-omnichannel by nature; for example, some consumers are not adapted to online shopping and others may not be able to access physical stores. In line with the motivating case, we study an online–merge–offline (OMO) system, which attempts to integrate the online and offline channel horizontally and vertically along the customer journey. The offline store stocks a specific product, enabling local customers to experience the product in person and acting as a fulfilment centre for any orders placed through the wireless applet of the physical store. The online channel through the B2C store utilizes the offline store or a centralized warehouse to fulfil the products. Shipping for online orders via any route is assumed to incur equal logistics cost, since this cost is normally standardized. In these integrated systems, omnichannel strategies such as buy-online/pickup-in-store and order-in-store-and-ship-to-home could be implemented over customers who are already omnichannelized. Since this paper focuses on the effectiveness of omnichannelization, we will not incorporate those well-studied omnichannel strategies, though including those elements in our model setting is possible.

We depict the OMO system in Figure 2, wherein potential online and offline customers enter the system as they do in the legacy system. We still assume that x units of potential online customers are purchased via advertising through the B2C platform, and there is one unit mass of offline potential customers in the offline store. The wireless applet of the offline store fundamentally alters the routes of the customers propagating through the channel system. This applet facilitates both omnichannelization and webrooming. In the online channel, a fraction, ${\gamma }_1$, of the online customers are directed to the applet of the offline store and become omnichannel users. In the offline store, a fraction, ${\gamma }_2$, of offline customers become omnichannel users by activating this applet. In the online channel, another fraction, $1-{\gamma }_1$, of online customers still order through the B2C platform since some customers just ignore the applet and choose the web-based browser in online shopping. In the offline store, there is also a fraction, $1-{\gamma }_2$, of offline customers who remain offline, as they resist using the applet. We again take the perspective of [9] that customers choose specific shopping modes based on a set of economic and non-economic utilities. Since we only build economic models, we assume ${\gamma }_1$ and ${\gamma }_2$ exogenous, which represents the non-economic reasons for purchasing online versus offline.

Recalling that the incoming customer size through the offline channel is 1, the omnichannelized customers in a number of $x{\gamma }_1+{\gamma }_2$ calculate their anticipated utilities [31] if they purchase the product on the wireless applet or visit the offline store in person and then purchase the product therein. They make the choice of purchase channel by comparing the corresponding utilities. We assume any order placed through the applet can be fulfilled and delivered by the offline store.

Denote the common selling price as p. A customer’s utility of purchasing one unit of product online, either through the B2C store or through the applet, is denoted as ${\overline{u}}_1$, while the utility function for purchasing one unit of product in the offline store is denoted as ${\overline{u}}_2$. These utility functions are defined as follows:

$$\begin{array}{ccc}\hfill {\overline{u}}_1& =& \theta \nu -h_1-p,\hfill \end{array}$$

(8)

$$\begin{array}{ccc}\hfill {\overline{u}}_2& =& \nu -h_2-p.\hfill \end{array}$$

(9)

We have the same product valuations in the OMO system as those in the legacy system, assuming the web-based product visualization is equally effective as that in wireless-based product visualization. With the framed model setting, the OMO system differs from the legacy system in three ways: first, the online channel and offline channel are jointly optimized with the objective of maximizing total profit; second, the product’s selling price is the same across both channels; third, the potential customers in the online and offline channels are omnichannelized at the rates of ${\gamma }_1$ and ${\gamma }_2$, respectively.

Since the online demand is generated through both the B2C platform and the wireless applet, we define the online demand as the sum from these two origins. The offline demand is generated by the customers who visit the offline store.

Region I: $h_2+\theta -1<h_1<\theta h_2$

1.

Demand generated by sole-channel potential customers

The sole-channel potential customers in the online channel who are not omnichannelized generate demand equal to $x(1-{\gamma }_1)(1-{\displaystyle \frac{h_1+p}{\theta }})$ in the B2C store. Likewise, the sole-channel potential customers in the offline channel who are not omnichannelized generate demand equal to $(1-{\gamma }_2)(1-h_2-p)$ in the offline store.

2.

Demand generated by omnichannelized potential customers

At the expected amount of $x{\gamma }_1+{\gamma }_2$, the omnichannelized potential customers are heterogeneous in their product valuation, $\nu$, satisfying the uniform distribution $U[0,1]$. These potential customers choose between online purchasing through the applet and offline purchasing in the store by comparing the corresponding utility functions ${\overline{u}}_1$ and ${\overline{u}}_2$. For these two purchasing options, we can identify the indifferent point ${\nu }_0={\displaystyle \frac{h_2-h_1}{1-\theta }}\in (0,1)$. A potential customer prefers the offline purchase if $\nu >{\nu }_0$ and the online purchase if $\nu \le {\nu }_0$. There is no difference between the two purchase approaches at $\nu ={\nu }_0$. In effect, ${\nu }_0$ captures the trade-off between digitization and hassle costs in purchasing.

We divide these omnichannelized potential customers, who are heterogeneous in $\nu \in U[0,1]$, into two groups.

Group A: $\nu >{\nu }_0$ A potential customer in this group chooses the offline store and purchases the product if and only if ${\overline{u}}_2\ge 0$ or $\nu \ge h_2+p$.

(i)

If $h_2+p<{\nu }_0$ or $p<p_c={\displaystyle \frac{h_2\theta -h_1}{1-\theta }}$, any potential customer with $\nu >{\nu }_0$ also satisfies $\nu >h_2+p$. Such a potential customer eventually buys from the offline store and contributes to the demand there. Hence, all the potential customers in Group A lead to a demand of $1-{\nu }_0$ at the offline store.

(ii)

If $h_2+p\ge {\nu }_0$ or $p\ge p_c={\displaystyle \frac{h_2\theta -h_1}{1-\theta }}$, only a potential customer with $\nu >h_2+p$ satisfies the conditions required to purchase the product from the offline store. Hence, the potential customers in Group A lead to a demand of $1-h_2-p$ in the offline store.

Group B: $\nu \le {\nu }_0$

A potential customer in Group B chooses to purchase the product through the applet if and only if ${\overline{u}}_1\ge 0$ or $\nu \ge {\displaystyle \frac{h_1+p}{\theta }}$.

(i)

${\displaystyle \frac{h_1+p}{\theta }}\ge {\nu }_0$, or, $p\ge p_c$, then a potential customer with $\nu <{\nu }_0$ does not satisfy $\nu \ge {\displaystyle \frac{h_1+p}{\theta }}$. Thus, no demand is generated to purchase through the applet from the potential customers in Group B.

(ii)

${\displaystyle \frac{h_1+p}{\theta }}<{\nu }_0$, or, $p<p_c$, only a potential customer with ${\displaystyle \frac{h_1+p}{\theta }}<\nu <{\nu }_0$ purchases the product. Therefore, a demand of ${\nu }_0-{\displaystyle \frac{h_1+p}{\theta }}$ is generated from the potential customers in Group B to purchase through the applet.

3.

Total Demand

Summing over the demand generated from the sole-channel potential customers and the omnichannel potential customers, we find that the online and offline demand functions are segmented by the p value. Namely,

(1)

when $p\ge p_c$,

$$\begin{array}{ccc}\hfill D_1^{M,I}(p,x)& =& x(1-{\gamma }_1)(1-{\displaystyle \frac{h_1+p}{\theta }})\hfill \\ \hfill D_2^{M,I}(p,x)& =& (1+x{\gamma }_1)(1-h_2-p).\hfill \end{array}$$

(2)

when $p<p_c$,

$$\begin{array}{ccc}\hfill D_1^{M,I}(p,x)& =& x(1-{\gamma }_1)(1-{\displaystyle \frac{h_1+p}{\theta }})+(x{\gamma }_1+{\gamma }_2)({\nu }_0-{\displaystyle \frac{h_1+p}{\theta }})\hfill \\ \hfill D_2^{M,I}(p,x)& =& (1-{\gamma }_2)(1-h_2-p)+(x{\gamma }_1+{\gamma }_2)(1-{\nu }_0).\hfill \end{array}$$

The second piece of the above demand functions ($p<p_c$) is exactly the one depicted in Figure 3, where for each omnichannelized customer via the wireless applet, $1-{\nu }_0$ goes to the offline store, ${\nu }_0-{\displaystyle \frac{h_1+p}{\theta }}$ chooses to buy online and the rest of ${\displaystyle \frac{h_1+p}{\theta }}$ does not purchase. The first piece of demand for $p\ge p_c$ is a degenerate scenario.

We also derive the demand functions for another two parameter scenarios of $h_1,h_2,\theta$. Their expressions are presented in the following.

Region II: $h_1>\theta h_2$

In this region, the online and offline demand functions are

$$\begin{array}{ccc}\hfill D_1^{M,II}(p,x)& =& x(1-{\gamma }_1)(1-{\displaystyle \frac{p+h_1}{\theta }}),\hfill \\ \hfill D_2^{M,II}(p,x)& =& (1+{\gamma }_{1}x)(1-h_2-p),\hfill \end{array}$$

which are the same as the demand functions in Region I when $p\ge p_c$. Nevertheless, their origins are different, as one originates from the problem parameters while the other originates from the pricing variable.

Region III: $h_1<h_2+\theta -1$

In this region, the online and offline demand functions are:

$$\begin{array}{ccc}\hfill D_1^{M,III}(p,x)& =& (x+{\gamma }_2)(1-{\displaystyle \frac{p+h_1}{\theta }}),\hfill \\ \hfill D_2^{M,III}(p,x)& =& (1-{\gamma }_2)(1-h_2-p).\hfill \end{array}$$

The demand functions corresponding to the OMO system are different in the above three regions. We study this system in two stages: first, by setting ${\gamma }_1={\gamma }_2=1$, we analyse the system under full omnichannelization; second, we numerically solve the system with general values for ${\gamma }_1,{\gamma }_2$.

5. The Perfect OMO Model

Adding the wireless applet can omnichannelize the sole-channel customers entering the system; however, as we explained, this is not always achievable for every customer. It is intuitive that ${\gamma }_1={\gamma }_2=1$ is the most ideal case for the OMO system since all the customers can choose between the two channels. This situation is referred to as a perfect OMO model hereafter, whereby it is the target for an omnichannel system to achieve. In this perfect system, we can fully solve the OMO model and analyse the impact of the problem parameters on the system performance. Under these ideal conditions, it can be checked that some customers may purchase from both channels, while other customers may purchase from only one channel. Note that in this case, all the online orders are placed through the wireless applet. These observations are formally stated in the following:

Observation 2.

(1) If $h_2+\theta -1<h_1<\theta h_2$, then both channels are viable purchase options; this is denoted as Region I.

(2) If $h_1>\theta h_2$, then only the offline channel is viable; this is denoted as Regions II-a and II-b.

(3) If $h_1<h_2+\theta -1$, then only the online channel is viable; this is denoted as Region III.

These results are illustrated in Figure 3. If the product is perfectly digitized, i.e., $\theta =1$, Region I degenerates into the diagonal line in the square, while all the customers purchase from either the online channel or the offline channel depending on whether $h_1<h_2$; if product digitization is completely ineffective, i.e., $\theta =0$, then all the customers purchase from the offline channel.

Note that when $h_1>\theta$, no potential customers adopt the online channel in the legacy system (Observation 1). However, in the new omnichannel system, an online potential customer under the same condition can switch to the offline channel through omnichannelization, whereas in the legacy system, that potential customer’s only option is to exit the market without a purchase. The OMO system creates potential demand from the online channel that produces no sales opportunity in the legacy system.

5.1. System Performance

By letting ${\gamma }_1={\gamma }_2=1$, we reduce the demand functions for the three regions in Figure 3 as functions of hassle costs and product digitization. A retailer must determine the retail price p and the potential customer number x to acquire from a platform based on the associated advertisement cost $a{x}^2$. This is achieved by maximizing the profit functions corresponding to the three regions, which are indexed by I, II, and III in Figure 4:

$$Ma{x}_{p\ge 0,x\ge 0}{\pi }^{M,i}(p,x)=p{D}_1^{M,i}(p,x)+p{D}_2^{M,i}(p,x)-a{x}^{2}i=I,II,III.$$

(10)

Note that when both channels are viable, i.e., when $i=I$, the demand functions are piecewise segmented in p. By solving (10) based on the demand functions of the three regions in Figure 3, we can characterize the performance of the new omnichannel system. Specifically, we compare the total sales and total profit of the OMO system with those of the legacy system in Section 3.

Market Outcomes in Region I

In Region I, where $h_1\le \theta ,{\displaystyle \frac{h_1}{\theta }}\le h_2\le 1+h_1-\theta$, both the online channel and the offline channel may have potential customers, dependent on the corresponding price p. The solutions to ($x,p$) are denoted as ($x^{\ast M,I1},p^{\ast M,I1}$) when ${\displaystyle \frac{h_1}{\theta }}\le h_2\le h_2^{\ast }$ and as ($x^{\ast M,I2},p^{\ast M,I2}$) when $h_2^{\ast }<h_2\le h_1+1-\theta$, where $h_2^{\ast }=1-\sqrt{\theta }+{\displaystyle \frac{h_1}{\sqrt{\theta }}}$ satisfies ${\displaystyle \frac{h_1}{\theta }}\le h_2^{\ast }\le 1+h_1-\theta$. While the details are relegated to the online appendix, Lemma 2 presents these solutions and relates them to those in the legacy system.

Lemma 2.

When $h_1\le \theta$, (1) for $h_2\in [{\displaystyle \frac{h_1}{\theta }},h_2^{\ast }]$, $x^{\ast M,I1}={\displaystyle \frac{{(1-h_2)}^2}{8a}},p^{\ast M,I1}={\displaystyle \frac{1-h_2}{2}}$; (2) for $h_2\in [h_2^{\ast },1+h_1-\theta ],x^{\ast M,I2}={\displaystyle \frac{{(\theta -h_1)}^2}{8a\theta }},p^{\ast M,I2}={\displaystyle \frac{\theta -h_1}{2}}$; and (3) $x^{\ast M,I1}>x^{\ast L},p^{\ast M,I1}=p_2^{\ast L}>p_1^{\ast L}$; $x^{\ast M,I2}=x^{\ast L},p^{\ast M,I2}=p_1^{\ast L}<p_2^{\ast L}$.

As is the case in the legacy system, the purchased number of potential customers in the OMO system decreases in the cost of advertising, a. A high a may bring more online customers to the OMO system than to the legacy system because a high a decreases online sales more severely in the legacy system, where potential customers cannot transition to an offline channel; however, such a transition is possible in the OMO system if the hassle cost of the offline channel is not prohibitive.

When $h_2$ is low, the online store can buy more potential customers, as some online potential customers with high product valuations, namely, $\nu$, will shift to the offline store, thereby pushing p higher. In contrast, if $h_2$ is high, some potential customers with low product valuations in the offline store may shift to the online store, pushing the price down. Note that the price in the OMO system is equal to either the online store’s price or the offline store’s price in the legacy system. The reason for this is that the profit functions in (10) are decomposable in p and x, while p is bounded by the two channel prices of the legacy system.

With the optimal retail price and the number of purchased potential customers, the total sales quantity obtained online and offline can be denoted as $D^{\ast M,I1}$ when $h_2<h_2^{\ast }$ and $D^{\ast M,I2}$ when $h_2>h_2^{\ast }$. We then compare these quantities with those of the legacy system.

Proposition 1.

(1) If $h_2\in [{\displaystyle \frac{h_1}{\theta }},1-\sqrt[3]{\theta }+{\displaystyle \frac{h_1}{\sqrt[3]{{\theta }^2}}})$, then $D^{\ast M,I1}>D^{\ast L}$; (2) if $h_2\in [1-\sqrt[3]{\theta }+{\displaystyle \frac{h_1}{\sqrt[3]{{\theta }^2}}},h_2^{\ast })$, then $D^{\ast M,I1}<D^{\ast L}$; and (3) if $h_2\in [h_2^{\ast },1+h_1-\theta ]$, then $D^{\ast M,I2}>D^{\ast L}$.

Proposition 1 characterizes the total sales quantity in the OMO system relative to that in the legacy system. When $h_2<h_2^{\ast }$, the price in the offline channel of the OMO system is the same as that in the legacy system, leading to demand that is equal to or greater than that in the legacy system. Such a price is too high for the customers in the online channel of the OMO system, so the online channel generates no sales and the potential consumers either choose to buy in the offline channel or do not buy. These high-value potential online customers will choose the offline channel and possibly even generate demand higher than that in the legacy system if $h_2$ is relatively low. The behaviours of the customers in the two channels lead to the first part of the above proposition.

However, if $h_2$ is high but less than $h_2^{\ast }$, intermediate-value potential online customers may find that no channel is suitable; this is because the price is too high for them in the online channel, while the hassle cost is too high in the offline channel. These customers, who can generate effective demand in the offline channel when $h_2$ is small, are lost when $h_2$ is large. The conditions between $h_1$ and $h_2$ in the second part of this proposition span a triangle in Region I with an area equal to $(\sqrt[3]{{\theta }^4}-\sqrt{{\theta }^3})/2$, a function that is concave in $\theta$ while reaching a maximum of 2.2% at $\theta =0.4972$ but zero at $\theta =0$ or $\theta =1$. A product with an intermediate $\theta$ is neither highly digital nor highly nondigital; therefore, neither channel is the perfect option.

When $h_2>h_2^{\ast }$, the total sales of the OMO system exceed those of the legacy system. This is because some potential customers who were not qualified to buy offline in the legacy system can now do so due to the decreased offline price; thus, more offline potential customers are absorbed. Due to the high $h_2$, the online channel of the OMO system purchases the same volume of potential customer clicks and sets the same prices as that of the legacy system. Moreover, omnichannelization creates additional cross-channel sales opportunities. These reasons jointly contribute to the third part of Proposition 1.

The online and offline stores of the OMO system might both sell less or more than those of the legacy system; this is decided by the trade-off between these two stores and the cost of advertisement to acquire online customers. As market coverage is in the interest of firms, as well as that of investors, Proposition 1 helps explain how hassle costs and product digitization affect total sales.

We then define the profit functions in the OMO system as ${\pi }^{\ast M,I1}$ when $h_2<h_2^{\ast }$ and ${\pi }^{\ast M,I2}$ when $h_2>h_2^{\ast }$ and compare them with their counterparts in the legacy system.

Proposition 2.

(1) If ${\displaystyle \frac{h_1}{\theta }}\le h_2<h_2^{\ast }$, ${\pi }^{\ast M,I1}>{\pi }^{\ast L}$; (2) if $h_2^{\ast }<h_2\le 1-\theta$, ${\pi }^{\ast M,I2}>{\pi }^{\ast L}$.

The implications of Proposition 2 provide strong theoretical support for the OMO system from a profit perspective. Ideally, if we can convert all sole-channel (online or offline) customers into OMO customers, then we can achieve higher system-wide profit than we can in the legacy system. This is true even for intermediate-$\theta$ products, which may generate fewer sales in the OMO system than in the legacy system.

Market Outcomes in Regions II and III

We can solve the profit optimization problems in (10) for $i=II,III$ subject to the corresponding demand functions. The optimal prices and potential customer number are as follows:

Lemma 3.

(1) When $h_1>\theta h_2$, $p^{\ast M,II}={\displaystyle \frac{1-h_2}{2}},x^{\ast M,II}={\displaystyle \frac{{(1-h_2)}^2}{8a}}$; (2) When $h_1<h_2+\theta -1$, $p^{\ast M,III}={\displaystyle \frac{\theta -h_1}{2}},x^{\ast M,III}={\displaystyle \frac{{(\theta -h_1)}^2}{8a\theta }}$.

From Lemma 3, we can derive total sales and total profits. Comparing these quantities with those of the legacy system, we have the following:

Proposition 3.

(1) When $h_1>\theta h_2$, $D^{\ast M,II}>D^{\ast L};{\pi }^{\ast M,II}>{\pi }^{\ast L}$; (2) When $h_1<h_2+\theta -1$, $D^{\ast M,III}>D^{\ast L};{\pi }^{\ast M,III}>{\pi }^{\ast L}$.

In Regions II and III, both total sales and total profit exceed those of the legacy system, while sales only occur in one channel. Moreover, omnichannelization expands the potential market, as stated in Observation 2.

We have shown that under any feasible region of $\theta ,h_1,h_2$, the level of profit generated in the OMO system is higher than that generated in the legacy system. This result is based on the condition that all sole-channel customers are omnichannelized (${\gamma }_1={\gamma }_2=1$), which provides any customer with more options in this system; nevertheless, the uniform-pricing rule may undermine the role of omnichannelization because this pricing policy is not unconditionally optimal. The firm in the case examined in this study utilizes this pricing rule to avoid intracompany price competition that customers may take advantage of. We have shown that a perfect OMO system with a uniform pricing policy outperforms the legacy system, which has sole-channel customers subject to individual optimal prices. In Section 6, we investigate the robustness of these rules in a general OMO model, where ${\gamma }_1$ and ${\gamma }_2$ are not equal to one.

5.2. Value of Omnichannelization

We measure the value of omnichannelization by the profit differences between the OMO system and the legacy system, namely, $\Delta {\pi }^i={\pi }^{\ast M,i}-{\pi }^{\ast L},i=I,II,III$, in the three regions in Figure 4. We conduct a comparative analysis of these profit differences with regard to $h_1,h_2,\theta$. Combining the regions with the same comparative analysis results, we repartition the three regions shown in Figure 4 to create another three regions as shown in Figure 5: Region A is the area defined by $h_1>\theta$; Region B is jointly defined by $h_1<\theta$ and $h_2<1-\sqrt{\theta }+{\displaystyle \frac{h_1}{\sqrt{\theta }}}$; and Region C is defined by $h_2>1-\sqrt{\theta }+{\displaystyle \frac{h_1}{\sqrt{\theta }}}$.

We further have the following:

Proposition 4.

(1) In Region A, the value of omnichannelization decreases in $h_2$ but is independent of $h_1$ and $\theta$. (2) In Region B, the value of omnichannelization increases in $h_1$ and decreases in $h_2$ and $\theta$. (3) The value of omnichannelization in Region C decreases in $h_1$ and increases in $h_2$ and $\theta$.

In Figure 4, we use arrows pointing in different directions to represent the monotonic properties of the value of omnichannelization versus $h_1,h_2$, and $\theta$, where → means no dependence, ↑ means increasing, and ↓ means decreasing.

Propositions 2 and 3 have verified that the OMO system is superior to the legacy system in terms of profitability. Proposition 4 has revealed the extent to which such a change is effective. We also have explored the impact of the advertising cost on the sales outcome in the two channels. In the perfect OMO model, ${\gamma }_1={\gamma }_2=1$ and all online customers employ the wireless applet and place orders there or visit an offline store before purchasing. From Observation 2, we know that only Region I of Figure 4 might contain both online and offline demand. Specifically, we find that both channels have nonzero demand only when $\theta >h_1$ and $h_2\in [1-\sqrt{\theta }+{\displaystyle \frac{h_1}{\sqrt{\theta }}},1+h_1-\theta ]$ and that the demand decreases when the cost of advertising increases. Under these conditions, we can show that the ratio between online sales and offline sales increases in $\theta$ and $h_2$ and decreases in $h_1$. These results are intuitive but have direct implications for increasing the online sales of a physical retailer. Retailers are tapping into online customer markets by setting up virtual fitting rooms in stores, launching live streams hosted by salespeople on social media, providing convenient delivery and return services for customers, etc. An online store on a B2C platform functions as an engine to infuse new customers from the Internet, which may enlarge the potential customer population of an offline store due to omnichannelization.

6. General Omnichannelization Rates

The omnichannelization rates (${\gamma }_1,{\gamma }_2$) dictate how the sole-channel users entering the two channels are converted into omnichannel users, making our models fundamentally different from the omnichannel models in the literature. We solve the OMO model with general omnichannelization rates numerically and compare the key performance measures of the OMO system with those of the legacy system.

6.1. Properties

In the OMO system, omnichannelization and uniform pricing are the counter forces driving the system profit. In contrast with the perfect OMO model, we first have the following:

Lemma 4.

When ${\gamma }_1\to 0$ and ${\gamma }_2\to 0$, the OMO system performs no better than the legacy system.

The above lemma is based on a simple observation, namely, if customers in the streams of the two channels do not transition each other, then the uniform-pricing policy applied in the OMO system should not be optimal. This leaves the question of whether the perfect OMO model remains superior to the legacy system in the case of general omnichannelization rates and whether the uniform-pricing rule is robust in the OMO system.

We have shown that bidirectional transitions can be altered by price p only when $h_2+\theta -1<h_1<\theta h_2$. In many practical cases, online hassle costs are perceived to be lower than offline hassle costs because traveling is often the first issue that is considered in terms of shopping online vs. offline. Without loss of generality, we assume $h_1=0$, and the condition of $h_2+\theta -1<h_1<\theta h_2$ is reduced to $h_2<1-\theta$. However, to solve (10) with the piecewise demand functions derived in Section 4, we must solve high-order equations for p, which is intractable for further analysis. We therefore resort to numerical methods to solve the OMO model with general omnichannelization rates. The solution is divided into two scenarios with segmented demand functions.

Lemma 5.

(1) If the retail price satisfies $p>p_c$, the optimal number of potential customers to acquire in the online channel and the retail price can be calculated based on the following set of equations:

$$\begin{array}{ccc}\hfill x& =& -{\displaystyle \frac{p[p+p(-1+\theta ){\gamma }_1+\theta (-1+h_2{\gamma }_1)]}{2a\theta }}\hfill \end{array}$$

(11)

$$\begin{array}{ccc}\hfill 0& =& 2a{\theta }^2(1-h_2)+2p^3{(1+(-1+\theta ){\gamma }_1)}^2-3p^2\theta (1-h_2{\gamma }_1)(1+(-1+\theta ){\gamma }_1)\hfill \\ & & +p\theta (\theta {(-1+h_2{\gamma }_1)}^2-4a\theta ).\hfill \end{array}$$

(12)

(2) 

If the retail price satisfies $p\le p_c$, the optimal number of potential customers and the retail price can be calculated as follows:

$$\begin{array}{ccc}\hfill x& =& {\displaystyle \frac{p(\theta -p)}{2a\theta }}\hfill \end{array}$$

(13)

$$\begin{array}{ccc}\hfill 0& =& 2p^3-3p^2\theta +p\theta (\theta -4a(\theta +{\gamma }_2-\theta {\gamma }_2))+2a{\theta }^2(1-h_2(1-{\gamma }_2)).\hfill \end{array}$$

(14)

Referring to Figure 2, we can determine that the first part of Lemma 5 applies to a case involving high p, which exists when $h_2$ is low. In contrast, low p occurs when $h_2$ is high, when all online customers choose to make online orders whether they browse the applet or not.

For the given problem parameters, we solve the two sets of Equations (11)–(14) and obtain the optimal solutions for ($x,p$). We explore the effect of omnichannelization rates on the total sales and profit changes of the two channels in the OMO system with respect to their counterparts in the legacy system.

6.2. The Effect of Omnichannelization Rates

We do not intend to conduct an exhaustive numerical study but have covered the representative parameter conditions. In this numerical study, we set $\theta =0.7$ and $a=0.1$, while $h_2$ is chosen to satisfy $h_2\in [0,1-\theta ]$. Section 5 has well characterized the OMO model in the extreme scenario, i.e., ${\gamma }_1={\gamma }_2=1$. We can verify that if both rates are close to zero, the OMO system performs worse than the legacy system. To characterize how ${\gamma }_1,{\gamma }_2$ affects the OMO system and its relation to the legacy system, we conduct two sets of numerical studies with ${\gamma }_1=0.2$ and ${\gamma }_1=0.8$ while setting ${\gamma }_2=0.1,0.5,0.9$. The main results are presented through a set of figures.

The changes in the total sales and profit of the OMO system versus the legacy system are presented in Figure 5 and Figure 6. When $h_2$ is either very low or very high and at least one omnichannelization rate (${\gamma }_1$ or ${\gamma }_2$) is high, total sales and total profit may both be improved greatly. In these extreme cases, the OMO system can always shift sufficient consumers to the channel that satisfies the majority of consumers well, i.e., either with an easy purchasing process and low price or a strong product fit and direct shopping experience. If both omnichannelization rates are low, then the change is small. The worst case is when $h_2$ is at an intermediate level, namely, when neither of the channels appeal to the majority of consumers. In these figures, we can see that the profit of the OMO system is slightly higher or close to that of the legacy system, while the total sales of the OMO system may even be lower than that of the legacy system. Figure 5 and Figure 6 cover a set of representative (${\gamma }_1,{\gamma }_2$) values, indicating that the total profit of the OMO system is generally improved in relation to that of the legacy system but that the total sales of the OMO system might be worse in some cases.

As we have set $\theta$ and $h_1=0$, we view products with high $h_2$ as relatively digital products that do not require direct experience during the purchase process and products with low $h_2$ as relatively nondigital products that require travel and touch-and-feel during the purchasing process. Examples of the first type are standard products such as books or music, for which traveling to a physical store is unnecessary and costly. Examples of the second type include certain nonstandard high-end products for which traveling to a physical store is necessary and less costly. Through our analytical and numerical analysis up to this point, we have shown that there are many situations in which the OMO system may be valuable but that this phenomenon is not universal, particularly for products that are neither highly digital nor highly nondigital, namely, those with intermediate $h_2$.

We repeat the above numerical studies by varying a with $a=0.01$ and $a=1$, proving that all the major qualitative observations are unchanged. Our most critical observation is that for cases where ${\gamma }_1=0.2$, there are situations in which the total profit of the OMO system is slightly less than that of the legacy system. From Lemma 5, we know that these phenomena should exist when both ${\gamma }_1$ and ${\gamma }_2$ are close to zero. These results demonstrate that the OMO system can function effectively when omnichannelization is mildly effective.

7. Conclusions

The retailing industry is evolving into an omnichannel world, and the boundary between physical and online spaces is vanishing. Motivated by offline expansion of online giants, this paper presents an online–merge–offline (OMO) approach to omnichannel integration via a wireless applet. In this system, pure online and offline customers can be omnichannelized, and complete omnichannel integration can be realized. Within such a boundless system, a potential customer needs only to balance their hassle costs and product digitization in choosing their shopping methods. To validate the effectiveness of this online–merge–offline system, we set up a benchmark system to approximate the legacy system in the examined business. We use the customer omnichannelization rates to characterize the online–merge–offline system. If these rates are both equal to one, the profit generated in the OMO system always outperforms that generated in the legacy system; however, this is not true if these rates are low. The system profit and sales also improve when these rates increase. We also find that the OMO system is most valuable if the product is highly digital or highly nondigital. In addition, converting sole-channel customers into omnichannel ones requires investments in integration, digitization, and information. These costs certainly reduce the system’s profitability and help explain why omnichannel projects implemented by Alibaba and other companies may not always be profitable in the short run. Scalability of the retail chain in terms of store number and product assortment are critical to success in this endeavour.

We list a few directions for future research. First, we may consider the pooling effect of the inventory at the warehouse when the online and offline channels are integrated; meanwhile, the product sourcing routes in the two channels in the benchmark system might be different. These additional factors can integrate the retail operation with supply operation in the retail chain system. Second, the omnichannelization rates of the sole-channel customers are exogenous, while a mechanism to endogenize these parameters should further uncover the role of webrooming. Third, we may analyze the operation output of the channel-integration projects initiated between leading e-commerce platforms and some brick-and-mortar retailers, as well as their diagnostics in China, since this retail innovation was started in 2015 and substantial real cases should be ready.