Abstract
Research on the digital economy has highlighted the assetization of data. This article argues for expanding existing research on data and datafication processes by focusing on how relationships are made and unmade through and from data. We introduce a general analytic model of “relationing” and show how relationships between users, companies, and products are created in three different moments—entanglement, dissection, and matching—first in the digital economy, then in physical stores. We show how payments with mobile phones connect the digital to the brick-and-mortar economy. Applying our model, we illustrate how a mobile phone's various data streams, money's record-keeping function, and retailers’ loyalty programs produce qualitatively and quantitatively new relations between customers, retailers, banks, app providers, and payment intermediaries. We argue that “relational embedding” captures the inherent relationality between users, their data points, and other economic actors: algorithmically relating users’ data profiles to other users’ profiles yields personalized recommendations, ads, or rebates, continuing the relationship between retailers and customers.
Introduction
When we use our smartphones to pay at the supermarket checkout, we set off a cascade of events involving many actors. The bank learns that we have spent $26 at the supermarket and updates our bank account accordingly. The supermarket learns about our use of the payment app in their store at, say, the train station, and the exact products that we purchased. Because we have stored our loyalty cards from the supermarket in the payment app, the supermarket's loyalty program will credit us 26 cents in cash back points. In this article, we show that the small act of paying digitally at a store's check-out, shapes new and existing relations between consumers, retailers, financial and nonfinancial payment providers, marketing agencies, and app providers. For decades, retailers, marketing agencies, and payment providers have been trying to improve their partial view of customers in physical stores. Digital payments, that is, using apps on phones and other devices, and the data they generate help to complete the picture in forming new and old relations, while also connecting the digital with the brick-and-mortar economy.
Research of the past decade has shown how the digital data economy works (e.g. Couldry and Mejias, 2019; Sadowski, 2019; Kant, 2020; Zuboff, 2020): tech companies utilize users’ transactional data, digital traces, or “behavioral surplus” to predict future preferences; reiteratively, they algorithmically calculate identity profiles based on their preferences and sell access to these identity profiles to marketing companies for targeted advertising. Utilizing browser and platform-based infrastructures, further clicks, likes, or online purchases create ever more digital traces, which are again collected, dissected, segmented, clustered, and reassembled to yield identity profiles for yet another round of targeting.
This article suggests that payment apps used in physical stores are fast becoming central devices to gain insights on the preferences and behaviors of consumers offline, like those obtainable online. Payment apps provide additional streams of user-generated transactional data—the economic assets—including preferences, amount spent, location, and time of purchase. Payment apps may also collect other streams of digital behavioral trace and transactional data, for instance, by directly linking to loyalty programs. Moreover, payment apps produce transactional data that, provided access exists, data and business analytics can link to other streams of trace data from that shopper's smartphone. For the banking industry, value in payments no longer rests solely in transaction and interest fees but rather in the transactional data that payments yield. For retail companies, such transactional data deliver insights into the precise expressions of behavioral consumer decisions. Once parsed, analyzed, and packaged, companies can sell access to these data packages as profiles for targeted marketing, a most profitable part of the digital economy. 1 As a result of this identity-driven marketing, device users receive “personalized advertisements” or “personalized recommendations.” In turn, device users thus become formatted for further consumption and commercial data analyses. Transactional data and new relationships are at the core of how banking industry analysts imagine digital payments to generate economic value: digital payments allow for data monetization and new payment experiences (Mützel, 2021).
Yet, when paying at a physical store using a payment app, customers not only leave additional digital traces. Indeed, the article argues that the study of data and datafication processes must be extended to new and changing relations: retailers, banks, payment providers, fintech companies, and payment app users form qualitative and quantitatively new relations with one another, subsequently, laying the foundation for future data monetization. Because payment apps collect and link payment data potentially to other streams of personal, social, and locational data passing through the shopper's smartphone, they may serve as so far missing unique prisms for companies to observe whole chains of transactions between users, customers, the banking industry, retailers, and brands. To further investigate how such relations are formed, we introduce an analytic model of the process of “relationing,” which shows how data-generating relations and relation-generating data are formed when paying digitally at physical stores.
The analytic model of the process of “relationing” consists of three different moments—entanglement, dissection, and matching—which delineate a cycle of how relations between companies, users, products, and customers are turned into data and, in turn, how different actors work out these relations to mobilize data as a resource to generate further profitable relations. In this process of “relationing,” two fundamental mechanisms of the give-to-get (Fourcade and Kluttz, 2020) digital economy concomitantly interconnect: personalization and relational embedding. Personalization is based on the relational embedding of disassembled data points, which get reassembled as “good matches” into purchasing identity profiles to produce a “personal experience.” Personalization is thus a thickening of relations. Furthermore, as it engineers not only customers but also entangles and formats relations between customers and retailers, such algorithmically calculated personalization leads to further changes: it also fosters new relations among retailers, banks, payment providers, fintech, and customers as they negotiate data access and usage rights.
Theoretically, the article builds on and contributes to the sociology of marketing, digital data studies, in addition to economic sociology. Specifically, in introducing and discussing this analytic model, the article contributes to analyses of how the digital economy, with its business model of advertising, consumer loyalty, and user retention, seeps into in-store shopping: It zooms in on how the detected “generalized reciprocity” between users, companies, and intermediaries (Fourcade and Kluttz, 2020) works at the intersection of digital data and in-store shopping.
The article proceeds as follows: The first section introduces the analytic model of “relationing” developed for the digital economy. A second section discusses payments and ensuing important innovations in marketing and retail: recordkeeping introduced the possibility to analyze customers’ data. A third section highlights how personalization and loyalty are intricately related in retail situations. A fourth section introduces the apps that allow for digital payment in brick-and-mortar stores. Sections two, three, and four lay the foundation for an analytic shift from online shopping to digitally equipped in-store purchases. Section five applies our model of relationing to digital payments in physical retail, focusing on changing relations between app users/retail customers, merchants, and payment intermediaries. In a digital brick-and-mortar economy the value of payments interconnects with loyalty. The article concludes with a discussion of personalization and relational embedding, which are two foundational mechanisms of the digital economy that the analysis of digital payments sheds light on.
Model of relationing: Turning relations into data and data into relations
As contributions to the field of social media platform studies have shown, the interrelations of users and platform architectures, the technical affordances of algorithms, and the business imperative of data extraction are key for understanding the current state of data capitalism (e.g. Van Dijck et al., 2018; Cheney-Lippold, 2017; Langley and Leyshon, 2017).
Datafication is widely understood as the transformation of social activities into data (Mayer-Schönberger and Cukier, 2013). Datafication's conceptual power rests on the ontological claim that everyday behaviors are now objectively available in digital form (Van Dijck, 2014). Behavior and sociality, however, are not always already data; they must be “infrastructured” first (Alaimo and Kallinikos, 2019). As Zuboff (2020) shows, the production of personal data as an economic resource rests on social, cultural, and technical preconditions: Google first considered data from search meaningless “exhaust,” which required costly server space. In the economic context of the dotcom bust and the cultural context of Silicon Valley entrepreneurialism, this “exhaust” came to be reframed as a resource for targeted advertising based on behavioral traces. The redefinition led to a new economic model of datafication: “the behavioral value reinvestment cycle.” All the data that users produce when typing in a search and clicking on search results, for example, “how a query is phrased, spelling, punctuation, dwell times, click patterns, and location” (Zuboff, 2020: 69), are no longer costly “exhaust” but become “behavioral surplus.” Machine learning algorithms then process this unstructured data into prediction products designed to forecast what a user will feel, think, and—most importantly—do. Then, the cycle starts again.
The “behavioral value reinvestment cycle” highlights the extractive, asymmetrical character of datafication between users and a company. Our model shifts the analytical gaze towards how “relationing” comes about, that is, the work of making and unmaking relations between users, companies, and products. We argue that the study of data and datafication processes must be expanded to include a focus on new and changing relations. In our analytic model of relationing, we identify three moments pertaining to the making and unmaking of relations:
First moment: Entanglement
In a first step, companies design their relation with users in such a way that users generate data. The digital platform's architecture entices users to type into the search bar, browse results, and like a post. Platforms thus use digital offerings as “loaded gifts” (Elder-Vass, 2016) to “captate” users, which involves “the care and effort put into establishing a bond” (Cochoy, 2007: 205). As users enter data, by clicking, linking, and tapping, with the aim of using the platform's services, they also provide gifts to the platform. Platform-generated user data are “gifts whose acceptance or use automatically entails a return” (Elder-Vass, 2016: 179) in the following instance, perhaps search results, recommended products, and the like. The exchange of the user's original gift and the company's countergift are interdependent. Indeed, the “obligation to repay” is already embedded “into the original gift itself” (Fourcade and Kluttz, 2020: 3). By giving away “digital gifts” of using their platforms “for free,” platform companies
Particularly social network companies, such as Meta, tap into their users’ sociality. Relating to others via likes, comments, or mentions is a kind of phatic communication serving the same function as the gift: establishing, acknowledging, or maintaining relationships (Romele and Severo, 2016). “Infrastructuring sociality” is a “delicate engineering accomplishment” that transforms everyday behaviors as digitally executable scripts so that they can be performed by users (Alaimo and Kallinikos, 2019: 304). A user's action is thus not simply electronically noted and registered. Rather, each of a user's data point is interrelated and entangled with others: whenever a user performs such a scripted action, say a “liking” of “a product,” a relation of a like between the user and the product is recorded.
Second moment: Dissection
In a second moment, recorded data get transformed and separated from its original context: users’ data traces get dissected into data points and “repurposed” (Beauvisage and Mellet, 2020a). In this moment, the generated behavioral data traces are no longer mere “gifts,” but instead become valuable “goods” for a company when data are transformed into a context-independent though relation-rich commodity, which can be used for further analyses. These goods carry the implicit meaning of what users desire and prefer—and how they are entangled with others. However, ex-ante “encoding” procedures (Alaimo and Kallinikos, 2017) or ex-post datafication techniques like vectorization (Rieder, 2020) ignore this surplus of meaning. Instead, these techniques allow for the decoupling of subjective and data-driven meaning, making a wide range of users and actions comparable and calculable—and treat them as digital data tokens that can be cross-referenced, aggregated, and combined with other data tokens (Alaimo and Kallinikos, 2019). One prominent example of such data tokens are “cookies.” While invented to memorize contents of online shopping baskets, they are now used and shared to display advertising across websites (Mellet and Beauvisage, 2020). Another example is the ubiquitous “like” button that helps to incorporate other websites into Meta's platform ecosystem and collect data points on users and nonusers (Gerlitz and Helmond, 2013). As Fourcade and Kluttz (2020) highlight, this repurposable customer data is at the center of a cycle of “generalized reciprocity” between different companies as well as website and platform users.
Third moment: Good matches
In a third moment, these momentarily context-cleansed data points get newly re-entangled to obtain further insights. Digital behavioral data provide information about users’ past actions. However, the value of this data only emerges when user profiles are algorithmically related to other users in order to derive new and potentially profitable relationships. In turn, this leads to new relational entanglements, interventions, and the generation of more personal data (Fourcade and Johns, 2020).
Data marks individuals as individuals of a certain type, for example, as deserving or undeserving (Eubanks, 2018) or as high-lifetime-value customers (Turow, 2017). This marking is only the means to an end: the goal is to detect “good” combinations of users and things. Such “good matches” (Zelizer, 2011) consolidate the relationship of users and companies, keep it going, and ideally extend it into the future. In Zelizer's case, culturally competent actors make good matches in everyday life. In the digital economy, however, this job is delegated to recommendation systems that bring users and things together, suggest personalized recommendations (Unternährer, 2021; Seaver, 2022), and thus consolidate the economic relationship between users and companies. As the analytic model of relationing suggests, personalization works as a thickening of relations. Users are algorithmically related to each other to compare, evaluate, and select the most profitable combinations of users and things. Thus, good matches enable the continuation of economic relationships.
Payments and recordkeeping
Before applying this analytic model to the brick-and-mortar economy, we first ask, where do data in the consumer economy come from? The answer is: from payments. The next sections thus discuss payments from a sociological perspective and point to the role of recordkeeping.
Sociology has paid much attention to money as a currency for economic exchange (e.g. Carruthers, 2010). However, payments do more than just transfer value. Payments give meaning to relationships and shape relations (Zelizer, 1996). Value transfer requires an array of technologies like accounting, cards, or terminals (Maurer and Swartz, 2017). Moreover, payments generate a specific “excess” due to the “work involved in settling an exchange” (Maurer, 2012: 17): intermediaries in the interaction charge fees for payments as well as the services and infrastructure they provide. From a sociological and anthropological perspective, payments thus involve interactions, technologies of transfer, and intermediaries.
Making payments and maintaining detailed records of every transaction are closely linked. Using a range of financial technologies and infrastructures, sellers and buyers have recorded completed payment transactions and owed deeds for the past 5000 years (Maurer and Swartz, 2017), generating data points. Two types of innovations are especially important in the current situation: keeping detailed records, for payments and granting credit, and using the gathered data to learn about customers. In the 1920s, US department stores began recording sales amounts, purchase dates, and customer information in large ledgers. These entries then served as the basis for granting store credit (Lauer, 2017). This system allowed stores to categorize customers based on creditworthiness, thereby creating new value and enabling them to learn about customer preferences and potentially recommend future purchases.
These processes were updated with the introduction of credit cards issued by nonbanking companies, which were initially issued as charga-plate cards and later equipped with magnetic stripes, eventually evolving into the modern credit card (Evans and Schmalensee, 2005; Mandell, 1990; Stearns, 2011). This not only altered the technology and infrastructure required to make a payment with a monthly credit line. Credit card payments also produced new transactional data points for customers, such as the itemized date, amount, location, and a brief description of purchase (Lauer, 2020: 6), as well as additional transactional data for retailers, payment providers, and customers beginning in the 1980s. Account details were now encoded on magnetic strip cards. Point-of-sale technologies, such as card reading machines have intertwined the act of payment as a transfer of value and the record of the event and the person: the transfer is the record, that is, the update of the record of who owes what to whom.
At the same time, card companies also began systematically mining the captured transactional data to gain insights into their customers’ preferences and creditworthiness, sorting them into tiered segments based on socioeconomic characteristics and spending patterns, and selling access to the identified segments to other companies through joint-marketing arrangements for further personalized offers (Swartz, 2020).
Loyalty and personalization
Typically, personalization refers to efforts to address individuals as singular individuals: personalized medicine promises an individualized treatment for a specific disease, while personalized marketing suggests advertisements tailored to individual customers based on their individual preferences and needs. This concept of personalization is based on knowing specifics about an individual.
However, in retail, personalization is relational, because it is connected to loyalty, which is a social bond between the retailer and the customer. Retailers may know the preferences of their loyal customers, which in turn helps to develop a personal retailer-customer relation. For instance, merchants who remembered customer preferences or recorded their customers’ previous purchases may alert their loyal customers to new in-store arrivals that fit their interests. In addition to this type of direct marketing, customers may also receive discounts or coupons for their next purchases. Foundational to this personal retailer-customer relation is a customer, who, since the beginning of mass retail in the 1930s, has been carefully crafted as loyal by marketing instruments such as offering rebate stamps, special discounts, and shopping opportunities. Indeed, the creation of loyalty is a key mechanism of social control for retailers: “through loyalty, merchants aim to lead customers to engage in everyday activities that strengthen the merchants’ businesses” (Turow, 2017: 20).
Supermarkets were in a peculiar position vis-à-vis loyalty and personalization since their idea of self-service was intended to increase efficiency while also democratizing, anonymizing, and depersonalizing shopping. They had little systematic sense of who the customers were, what they wanted, and if they were returning regulars – even though market research regularly surveyed supermarket customers on their preferences (Turow, 2017). Retailers had to find new ways of getting to know their customers. Faced with these challenges, supermarket chains developed loyalty and reward programs in the late 1990s, drawing on the success of airline reward programs (Swartz, 2020). These programs rewarded customers with points for their purchases when they linked them to their reward program identity at checkout, regardless of payment method. The programs were designed to learn more about their customers while granting rebates and offers (Zwick and Denegri Knott, 2009).
To be sure, the new loyalty programs work differently than the old rebate books. In addition to enticing consumers to return to the store and spend more, they want personal data, which includes information about who bought what, when, and where. As loyalty shifted to an electronic database, it became increasingly reliant on a give-to-get logic to “entice consumers into divulging a range of personal information” (Pridmore, 2010: 565; see also Elmer, 2004). For frequency marketing, loyalty programs became a “stealth weapon” (Turow, 2017: 83). Supermarkets began to mail out “specific promotions to members, matching what they had previously bought with participating marketers’ coupons” (Turow, 2017: 87), regardless of how accidental the data basis was. When a customer returned to the store, a new cycle of entanglement, dissection, and good matches began.
Customers, however, resisted being formatted, much to the dismay of marketing. They used their loyalty cards on occasion, but they also forgot or used other people's cards. This resulted in a very fragmented entanglement of records for retailers. Furthermore, payment and loyalty records were siloed in different databases.
Digital payment apps
The rise of smartphones and digital payment has transformed the retail landscape: shoppers now bring in technology that can be used to reliably identify each shopper. The “lowly mobile phone” has become “a marketing, loyalty, and payment device” (Turow, 2017: 101). Now, shoppers pay instantly at the point-of-sale with a smile, the touch of a button, or simply by waving the phone. Their payment apps use near field communication, QR-codes, or bluetooth to connect to a terminal. Numerous payment apps are available, including digital wallets (e.g. Google Pay and Apple Pay), loyalty apps (e.g. Starbucks), retailer loyalty apps (e.g. Safeway for U, Kroger's Boost), (neo-)bank apps, or peer-to-peer payments apps (e.g. Venmo in the US, Swish in Sweden). Most of these apps leverage existing financial infrastructures, such as credit card networks and bank accounts (Bernards and Campbell-Verduyn, 2019; Gießmann, 2018). These “things of payments” are more than convenient technologies. For one, payment apps, just like other apps, track users and log their purchasing preferences, amount spent, time, and location (e.g. Beauvisage and Mellet, 2020b; O’Dwyer, 2019). They shape the economy and help to digitalize and reconfigure consumer practices (e.g. Cochoy et al., 2019).
Market research expects the transaction value of digital payments, using wallet and other apps, at the point-of-sale to nearly double from 2020 to 2025 to US$9.5 trillion. The COVID-19 pandemic has served as an accelerator for the adoption of contactless, digital payments around the globe and has further pushed what some call the “retail nirvana” of frictionless payments (Gregg et al., 2015) toward reality.
In the online world, the tech giants and the online marketing industry have shown how “personalized experiences,” targeted marketing, tailored ads, and building relationships with customers are achievable and profitable. As we show in the next section, paying digitally brings the digital economy into brick-and-mortar stores, thereby altering the relations between retailers, customers, and payment providers.
Moments of relationing: Digital payments in brick-and-mortar stores
Above, we introduced an analytic model of three moments of relationing, developed for the digital economy. In this section, we apply the model to digital payments in physical retail stores. When retailers adopt the logic of the digital economy, relations between retailers, customers, payment providers, and app providers are reformatted to data-generating relationships, resulting in relationship-generating data.
Especially because of competition from online vendors, physical retailers pay special attention to providing a positive customer experience, which in turn contributes to loyal customers and enforces a “personal brand relation” between a specific store or a supermarket brand. Customers return to retailers they can trust in the selection of brands and quality of products, with whom they are familiar in terms of location and products, and where the overall shopping experience is enjoyable. Payment becomes instantaneous, seamless, and ambient. The so-called “payment experience” of social and communicative interactions with a seller, the product, peers, and potential future buyers moves to center stage (e.g. Tkacz, 2019).
What appears to be a seamless, frictionless payment experience front stage is a data-generating and relationship-generating machinery backstage: customers gift payment data, customer identity, and purchased items together to banks, retailers, and marketing agencies all at once. Purchases, customer identities, and payment information get encoded into data. The desire for a seamless, frictionless shopping experience, including the payment situation, mutually benefits all those involved who are interested in the production of data.
Notably, electronic payments via debit, credit, or loyalty card already connect disparate data infrastructures of banks, payment processors, and retailers, each of which has a unique, fragmented view of their customers: A loyalty card customer's identity paying via app is linked to all items purchased, including where and when they were purchased, for how much each time, and which bank account was used. From the perspective of a bank, a customer paying via app is linked to all transactions of their accounts, which also may include merchants’ bank accounts and insights on total customer wealth (Botta et al., 2017).
Digital payments are “shuffling who can see what” of this data (Swartz, 2020: 125) as technology companies and fintechs try to “embed financial transactions within their data streams” (Westermeier, 2020: 2) of locational, personal, and social data (e.g. McKinsey, 2022: 12). As consultants suggest, “[p]robably the greatest potential of data monetization comes from merging cardholder data with data from the merchant side to gain an end-to-end view on transactions that can unlock additional value” (Botta et al., 2017: 2). Within the limits of privacy regulations, banks, retailers, and third-party marketing agencies can gain insights. For instance, the payment provider Revolut writes in its privacy statement that they gather different data from the users’ social media accounts: “If you allow us to, we will collect information such as friends lists from Facebook or similar information from other online accounts” or device information like “contact information from your address book, log-in information, photos, videos or other digital content, check-ins” (quoted in Ferrari, 2020: 529). Payment apps are located at a critical intersection of data streams between customers, retailers, and payment providers. These intersections become locations for re-negotiations of data access to allow for more comprehensive views of customer identities. 2
For marketing, this merging of data because of digital payment at a physical store is one more milestone towards its holy grail of knowing a user's tastes, preferences, habits, and desires ex ante to sell fitting products and services and deliver a “personalized experience.” For banks, payment data and their monetization—that is, selling access to customer data to third parties, learning more about their customers, and adding supplementary data points to internal data—shifts their business model: previously profitable due to fees, payment transactions are now profitable because of the data generated and the relations forged are turned into “assets” (e.g. Birch and Muniesa, 2020; Birch et al., 2021). With payment apps, the discrepancy between observable fragments of consumer choices and so-far unobservable consumer behaviors and choices collapses into an entire stream of data, which pleases the insatiable data hunger of machine learning processes (Fourcade and Johns, 2020).
At the heart of this search to find good matches are machine learning processes that convert assortments of data points on customers, such as items purchased, date of purchase, location of purchase, into personalized predictions, that is, algorithmic personalization, based on different matching techniques and domain-specific business needs. For retailers’ loyalty apps, product recommendation for cross-selling and upselling is key (Coll, 2016; Mole and Nadeau, 2016). Such recommender systems are mathematically rooted in collaborative filtering algorithms (e.g. Ricci et al., 2011). They suggest probabilities of next, most likely choices based on the relations between two different types of things, say, data points generated on customers and data points generated on items purchased.
By linking data points of identities to behavior, marketing can find a “good match” based on homophily for the next most likely consumer choice. For customers, a suggested ad or product is personalized because it is “relevant just for me” based on algorithmic calculations. Meanwhile, and behind the façade of the relevant-for-just-me, a customer's singular choice has been linked to a slew of relationships between other customers and things, other customers and retailers, and other customers and retailers, generating additional data points for the next round of “personalization.” Previously, retail stores used personal data to maintain good relations with their customers; now, relations between people and things generate further data.
To be sure, this type of personalization has nothing to do anymore with a coherent individual, the person, or “the personal.” From the perspective of marketing firms, an individual is an ever-drifting set of data points, which form bundles of vectors that get disassembled and reassembled in algorithmic processes. Individuals are transformed into dividualized constellations of data (Lury and Day, 2019). Only when algorithmic calculations of similarity and difference based on relations yield predictions on profiles of identities, that is, when decontextualized data is translated back to social parameters with which marketers can work, do we find “repersonalization.”
Such profiles represent a new type of capital: eigencapital (Fourcade and Healy, 2017) . It arises from an individual's “position and trajectory according to various scoring, grading and ranking methods” (p. 14). As consumers respond to personalized offers, their profiles are updated based on their choices. This type of capital is the result of user-generated, trackable, and traceable data, which is already interrelated based on previous choices made by others and which has been algorithmically calculated using various methods of differentiation and pattern search.
Conclusion: Personalization and relational embedding
Traditionally, banks and other payment providers obtained value from payments by incurring transaction-processing fees, which are levied on money's passage from merchants to customers and then to the merchants’ bank account. Yet, following the financial crises and the advancement of digital transformation, banking consultants identified new possible revenue-generating strategies: monetizing customer data, providing access to customer data, and personalizing services while adhering to privacy and compliance requirements (Capgemini, 2012, 2015; Gregg et al., 2015). Data monetization emerged as one of the future imaginaries banking consultants envisioned for banks (Mützel, 2021). This idea spread to supermarkets as well. At checkout, this article has shown, payment types and loyalty programs interconnect and supercharge each other. When paying digitally, the production of transactional data highlights new and changing relations between payment app users, retailers, banks, fintech intermediaries, and marketing agencies.
Our paper intertwines substantive and conceptual contributions. Using an analytic model, we have argued that digital payments assist in bringing foundational principles of the digital economy to brick-and-mortar shopping. Smartphone apps and wallets supercharge existing loyalty programs: Merging data on payment and purchases in loyalty programs, dissecting, and reassembling them for good matches, allows retailers to predict how profiles of customers will respond to shopping situations, including discounts and rewards.
In common parlance, personalization refers to efforts to address individuals as singular individuals. Instead, this article suggests that “personalization” is a thickening of relations and, furthermore, reshapes relations between a multitude of actors involved in the digital consumer economy. When tracing transactional data such personalization efforts are based upon, we will find algorithmic processes that calculate relations of similarity and difference between users and user-clusters and yield predictions of future decisions—instead of individually tailored, targeted recommendations. Thus, personalized ads that are presented as a perfect match for just me, are based on calculated generalizations about who I am, that is, which parts of my behavioral trace and transactional data available are similar or different from others, who are also estimated based on their data profile. The idea of personalized offers must be viewed as an engineered means of maintaining “user engagement” in order to keep the data flowing and to continue the uneven, but generalized exchange between users and economic actors in the digital economy.
This article also proposes “relational embedding” to capture the inherent relationality between users, their data points, and other economic actors. On the one hand, relational embedding is a process dual to personalization. With users’ data points, typically in partial and packaged data profiles, their predicted preferences are relationally estimated, compared, and contrasted to other data profiles. Transactional data thus turns into algorithmically calculated predictions on users’ future relationships with companies, for example, as loyal or creditworthy. On the other hand, relational embedding refers to the multitude of relations the circulation of transactional data involves. In following the stream of transactional data from users to other economic actors, patterns of relational embedding become evident.
Future research into algorithms and platforms may benefit from our understanding of relational embedding. Take “embedded payments,” a recent buzzword in the payments world as part of the larger push towards “embedded finance” (e.g. Dresner et al., 2022). On the surface, embedded payments are seamless, frictionless, or even invisible payments that are easy and convenient for consumers. Consider renting a bike via an app; the ride begins and ends on the app—payment “simply happens” in the background once the ride is completed. Or consider Amazon Go, Amazon's store where payment is initiated when the store's cameras register a customer leaving the store. In the background, the act of consumption involves multiple processes of relationing: Users or consumers leave data traces that entangle them and other users in thickening relations of similarity and difference to evaluate likelihood of fraud or propensities to buy a particular product—with associated actions like blocking an account, better offers to sweeten the deal for potential churners, or recommendations and discounts. On the other hand, the twin efforts to monetize data and payment involve the rearrangement of relations between economic actors: Embedded payments via apps, in particular, involve not only payer and payee, but also other payment intermediaries, each of which is interested in a piece of the payments or, alternatively, the data pie.
Moreover, our analytic model of relationing sheds light on fundamental principles underlying the current datafication and digitization in physical retail stores. The identified processes of relationing, that is, entanglement, dissection, and good matches, invite further research, for example, on how shoppers are encouraged to use a retailer's digitally infrastructured payment and loyalty apps, that is, how do retailers “train people to give up personal data willingly” (Turow, 2017: 168)? Also, who are the economic actors in payments that can negotiate access to valuable consumption data, such as item-level purchase data, how do they re-negotiate access, and who remains excluded in the cycle of relationing?
Additionally and with an empirical shift to the inner workings of the payment industry, future research could focus on how data and payment infrastructures work and are maintained, including a focus on human workers who help to classify and train data or work to avoid breakdown and malfeasance.
By focusing on the different and changing relations of payments and data and introducing an analytic model, our paper contributes to a sociologically nuanced picture of payment processes.
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Swiss National Science Foundation (SNSF) (grant 10001A_200764).
