Intermediation of lending: Platforms,mobile money,and data transactions in India

Financial Technology (fintech) is marked by the confluence of digital and financial spheres, a phenomenon that has been taking place worldwide with tech-driven companies accessing data transactions within financial infrastructures and integrating them into their digital platforms. Fintech companies in India, like in several other emerging economies such as South Africa, tend to be venture capital (VC) backed startups (Pollio and Cirolia 2022). Some of the most cutting-edge developments in fintech, particularly in its mobile-first avatar, happened in Kenya with telecom company Safaricom's mobile money platform M-Pesa. Since then, further lending services have been built on the original infrastructure that integrates “mobile telephony and digital data with commercial lending” (Donovan and Park 2022) Various versions of fintech model with crucial regional variations can be seen in a range of Asian nations such as China (Ant Financial's Alipay), India (Navi, KreditBee), and Indonesia (Gojek).

There has been a frenetic growth of the fintech sector in India since demonetization. Unified Payments Interface (UPI), created by the National Payments Corporation of India (NPCI), is an interface for financial transactions that consolidates multiple bank accounts into a single mobile phone application, thereby creating a digital payments gateway. The embrace of digital money was further accelerated when, in 2016, the ruling Bharatiya Janata Party regime banned Rs 500 and Rs 1000 notes in a so-called “demonetization drive.” The fintech companies had been lobbying for this for a while. It allowed them to reach more customers and deliver financial services on their phones.

With the establishment of UPI in 2016 and then the demonetization (notebandi) shocker by the Government of India in November of the same year, the movement and circulation of digital money in Indian society received a fillip. The demonetization move plunged the country into a “transactional firestorm” in the way ordinary people encountered and experienced the sudden devaluation/withdrawl of Rs 500 and Rs 1000 notes they had owned: these notes would cease to be legal tender soon and this lead to anxiety and panic among some citizens who had mostly kept savings in the form of cash. That said, the demonetization drive over time lead to the “expansion of transactional systems” as part of other connected shifts brought by Prime Minister Narendra Modi's “Digital India” initiative (Athique 2019).

The increased spread of digital payments and digital lending in the name of “financial inclusion” has been facilitated by the Unique ID (Aadhar) system, Unified Payments Interface (UPI), and the India Stack Infrastructure (also known now as Digital Public Infrastructure). Many Indian citizens who had previously been unbanked for a long time were asked to open Jan Dhan (People's Money) bank account, and then their bank accounts were authenticated through the Aadhar (unique ID) and connected through mobile phones setting up the “JAM” (Jan Dhan-Aadhar-Mobile) payments pathway (Ranade 2017).

While Aadhar's role in delivery of governmental subsidies and welfare benefits has been touted, Aadhar has done much to create new mobilities of data in the many different market domains including digital payments and lending. During the loan approval process, the Aadhar card is key basis for the eKYC (Know Your Customer) checks as part of customer authentication and verification. The biometric enacting of Aadhar transformed the Indian resident's body into storable and retrievable data, and this instituting of “individuals as data points within databases” has not only served the so-called welfare beneficiary (labharthi) but also served the market players and data economy (Ranganathan 2020, 53). The identification, and then authentication infrastructures, that were created with Aadhar, have then been made part of the India Stack Application Programming Interfaces (APIs) representing the consent layer and cashless layers. Through Aadhar and the India Stack, the Indian state has been involved in deriving valuable data from the very large Indian population, and circulating that data to create more data, and therefore has been a participant in “enacting the logic of capital” (Prasad 2022, 808). To put it succinctly as Nayantara Ranganathan (2020) does, Aadhar and India Stack have made Indian citizens’ data “market-ready.”

Digital Transactions certainly have been speeded up in India, as a consequence of these infrastructures, whether that is using the mobile handset to pay for vegetables in open-air bazaars through QR codes, or receiving instantaneous deposits of credit in one's online bank account through a loan app. Given that the doctrine of financial inclusion places such emphasis upon credit expansion, the focus of this article is on loan apps and the rapid growth of digital lending as one part of the fintech sector in India. I examine the intermediation process in the loan app ecosystem. That is, how and why loan apps mediate between lenders and borrowers and, in a functional sense, precisely how loan apps are serviced by fintech infrastructures allowing for automation of tasks such as risk score checking and identity verification. These processes are a crucial part of data and money transactions that involve loan approvals and disbursals. These are decision-making processes that were once assigned to human beings by both formal banks and informal moneylenders.

Writing about transactional records and electronic money, Rachel O’Dwyer (2019: 2) notes, “Many of the new channels for payments can record detailed transactional data alongside a range of other demographic, psychographic, social, and even biometric details about payers and payees” where “how we transact” is correlated with “other personal and social information.” UPI, India Stack, and JAM did not only afford the movement of digital money across mobile phones in India, but also helped grow the amount of data stored and recorded about monetary transactions, which could also be coupled with or related to other kinds of transactions, including social media and e-commerce activities engaged by mobile phone users.

In their operations, loan apps (lending platforms) enact what Carola Westermeier (2020: 2052) sums up as the “platformization of financial transactions.” This involves the linking of “transactional data to other kinds of data such as location, past behaviour and social connections” which “enables all kinds of future value propositions.” Loan approvals and loan disbursals enacted by lending platforms in India are often not solely based on credit scores because for many new-to-mobile phone users, such risk scores simply do not exist because many have never had a credit card. Loan apps have to then rely on alternative datapoints such as social media activity and phone records to determine financial transactions such as loan disbursals and approvals, as well as evaluate/ascertain future repayment windows and interest rates.

Accelerated by cheap data plans provided by the telecom company (or “industrial conglomerate”) Reliance Jio in 2016, a whole new group of first-time mobile phone users emerged in India (Mukherjee 2019; Athique and Kumar 2022). Many among the new-to-mobile phone users were the erstwhile unbanked, who could now be included in formal banking and financial operations through mobile digital practices like opening a bank account on the phone or using a mobile wallet. Thus, goals of digital inclusion and financial inclusion converged through mobile internet in India.

Government of India's governance and regulation of banking activities is mostly carried out by the Reserve Bank of India (RBI). The RBI initially did not want to rigorously regulate, or pay attention to, the fintech credit sector for two reasons. Firstly, the very small size of the (unsecured personal) loans made them not that significant in comparison to the large transactions handled by banks. Secondly, the government wanted the unsecured short-term credit business to grow and reach lower middle-class populations, prompting a considered silence in policy on micro-lenders. This regulatory slackness has begun to change lately (Manikandan 2024).

Amidst lax regulatory guidelines, payment and lending platforms mushroomed in India to such an extent that they created unease and anxiety among consumers and regulators, as some predatory loan apps flouting lending licenses and deploying aggressive collection methods started harassing customers who could not repay on time. These predatory lending platforms made calls to the defaulters’ relatives from the phone contact list (part of the data collected during the initial onboarding), and engaged in blaming and shaming in a very public way (Christopher 2021). Even legitimate loan apps, who wanted to scale up by gaining more market share and increasing their disbursals, started to lend too ambitiously and indiscriminately, and the number of bad loans increased, especially amidst the economic downturn during the Covid-19 pandemic.

The RBI issued a circular on Sept 2, 2022 that laid down several restrictions regarding premature withdrawal of regulated entities like Non-Banking Financial Companies (NBFCs) and platforms selling loans on the secondary market. The lending platforms were also cautioned to stop advertising guaranteed returns to attract customers without explicitly mentioning the risky nature of investment (RBI Circular 2022).

In this article, I focus on the digital money and data transactions with respect to loan apps, particularly those loan apps that provide unsecured short ticket size loans and/or microloans to individual consumers. The lending platforms operating in India that I examine are Navi, Fibe (EarlySalary), and CASHe. Some of these loan app companies also offer other kinds of loans like long-term home loans, and are not restricted to just short-term small ticket loans. In several instances, these companies partner with external lenders (banks and NBFCs) to offer loans. Some of them also provide loans from money raised by their own inhouse NBFC arm that is funded by promoter equity. For example, Navi and KreditBee's respective NBFC arms are Navi Finserv and KrazyBee. Navi and KreditBee do not always depend on Navi Finserv and KrazyBee, but enter into co-lending arrangements with other NBFCs and banks (Noronha 2023).

I also delve into the differences between the work of legitimate loan apps and those predatory loan apps that emerged during the Covid-19 pandemic in India which not only charged exorbitant interest rates and processing fees, but also deployed abusive recovery techniques entrapping borrowers. Both legitimate and illegitimate (predatory) loan apps scrape phone data, like contact lists and phone call records. Legitimate loan apps scrape data from phone devices to ensure fraud detection, that is, to make sure they are not dealing with fraudsters. Predatory loan apps use the customer contact list as a collateral to call, or threaten to call, the customer's relatives in the event that they do not repay, tantamount to an act of blackmail. While there are significant differences between legitimate loan apps and the predatory loan apps, there are times when the lines are blurred. One reason for this is because of the architecture of lending platform ecosystem where the loan app intermediates between lenders and borrowers as well as between lending institutions like NBFCs and fintech as a service (FaaS) companies. Some of this intermediation creates opacity.

The dependence on FaaS players that offer various services of loan management and credit scoring through their APIs and software development kits (SDKs), provide the much sought after accelerated velocity to digital financial transactions and the decision-making behind the almost instantaneous loan approvals and loan disbursals. Such a dependence becomes a cause of concern, especially if these FaaS players decide to retain customer data and not share it with the loan app and NBFC. These are the stakes of instantaneous digital transactions, that on the frontend of the app interface for a borrower might seem to comprise of a few selections and scrolls, but in the backend involves a complex assemblage of human and nonhuman intermediaries, all part of an intricate fintech infrastructure that supports loan apps. While the dependence on external FaaS players (third-party developers) is more in the case of predatory loan apps, even legitimate loan apps depend on third-party algorithms for some aspects of loan collection or risk verification. Hence, this criterion of dependence simply cannot result in an absolute categorization.

Another aspect of digital monetary transactions is the collection of phone data from the borrower's phone. In the case of legitimate loan apps, this is not used to later harass customers as predatory loan apps do. However, this kind of collected data from social media activity and phone records becomes part of alternative datapoints created by machine learning and predictive analytics algorithms that help legitimate loan apps decide whether a borrower is credible enough to be offered a loan. This is where behavioral finance meets Artificial Intelligence as fintech companies in India (and worldwide) try to make the case for how social behavior of customers can be related to their long-term financial behavior.

The predatory loan apps and the financial scams ensued by them are more than just aberrations from or exceptions to contemporary capitalism (Poster 2022). The scams perform critical boundary work to lay bare the workings of financial infrastructures in place that circulate money and delivery of financial services. Furthermore, while the legitimate loan apps are not predatory by explicit intention, their financialization logics can lead to predation. The decisions of such lending platforms about who is eligible or not eligible for loans are based on alternative datapoints and algorithms that while promising to expand access to credit can perpetuate “predatory inclusion” which “is the logic, organization, and technique of including marginalized consumer-citizens into ostensibly democratizing mobility schemes on extractive terms” (McMillan Cottom 2020). Since the loan apps often times are lending based on the borrower's intent to pay rather than their capacity to do so, there remain chances of borrowers falling into debt traps. After all, access to credit can be about supporting education, health, and aspirational lifestyles, and access to credit can also transform into access to debt.

One of the promises of demonetization and financial inclusion drives (that include mobile payment, storage, and lending technologies) has been increased efficiency and transparency through immediation. Such a “politics of immediation” can obscure the potentialities in the intermediations to be found in everyday social relations that enable (and are enabled by) in-person money exchanges and cash circulations (Zabiliūtė 2020). For lower-income population groups in India with temporary and unstable jobs, handing over a certain amount of daily cash to household members is (or has been) an obligation of the head of a family which makes it difficult to save those amounts in some digital accounts. Indigenous lenders who had a role in India's economy before the British colonial government arrived (and other global banking outfits sprung up), continue to exist today as part of key merchant communities, such as Marwari, Nat-tukottai Chettiar, and Multani (Baker 2021). Lucy Baker (2021:1817) argues that traditional/indigenous moneylenders are able to take more risks than digital lenders because they have “developed social networks of “informers” to reach and prompt” defaulters to repay previously agreed upon monthly installments. This is different from tele-calling recovery agents in the case of loan apps. In this paper, I have argued against a popular assumption that with increased use of digital lending we have moved from intermediation to immediation. This is because a range of financial intermediaries exist in digital lending and the loan app itself is structurally not a neutral intermediary.

Approaching loan apps as multi-situated intermediate spaces

I begin by explaining my research methodology as well as the theoretical framework of intermediation through which I approach lending platforms, and what it means to study digital transactions by putting on such a conceptual lens. Focusing on the lending platform landscape in India, I analyze the loan app transactions at three levels.

Firstly, I analyze how the instantaneous loan approval process is demonstrated in loan app advertisements and how this process appears on the front-end interface for borrowers onboarding through the loan app.

Secondly, I examine how the loan app as lending platform acts as an intermediary between borrowers and lenders (lending institutions like banks and NBFCs) and how that impacts interest rates, repayment windows and processing fees.

Thirdly and finally, digital lending involves several financial transactions that create new payment and credit records, and check past payment, credit, and social media activity histories involving application programming interfaces (APIs) and software development kits (SDKs), all part of Fintech as a Service (FaaS). SDKs help build applications and APIs help communicate across applications ensuring interoperability. These APIs and SDKs work with/in the backend transaction chains enacted by network economies to provide the convenience of instantaneous loan approval and money transfer but can also be used to potentially siphon off user data. Data analysts from legitimate FaaS companies such as Think360.ai, a company that provides data anlaytics services to loan apps and other clients, were quick to emphasize and clarify that they are "data processors" and not "data owners." Data scientist Amit Das, who founded Think360.ai, mentioned to me in a research interview that he did not want to be involved in data ownership as that would require him to build a customer facing business and provide customers with a compliance layer. As an AI analytics firm, he reasoned that there are smarter non-personally identifiable information ("non-PII") ways of abstracting data which could then be used to improve the ML layer.

I will analyze the role of APIs and SDKs in the lending platform ecosystem with an emphasis on the role of “alternative data.”

The more established and legitimate loan apps like Fibe and KreditBee, who have been in the digital lending business from 2016 onwards, carry out most of the loan disbursal and credit scoring operations through their own in-house stacks, thereby reducing their dependence on third-party developers. Some others like RapidRupee run their entire lending business end-to-end on Finflux's suite (Ramanathan 2021). A FaaS company, Finflux's operations include loan disbursal and loan management. Loan disbursal involves plugging into the banking APIs to disburse money into borrowers’ accounts. Much of these automated digital transactions, thanks to FaaS company's APIs and SDKs, makes the loan approval and disbursal very quick, to the point of being almost instantaneous.

The so-called “alternate” or “alternative data” collection and processing, much of it involving behavioral data captured from and flowing through phone and social media activity, that governs loan app decisions about eligibility, payment windows, and interest rates remain under-discussed in the public domain (Langevin 2019). Yashoraj Tyagi, chief technology officer of lending platform CASHe mentions in an interview “We would never extend credit to somebody (a customer) who does not have the intent to repay because even if somebody has the intent and they lose the ability downstream, they will sometime repay you the money because the intent is always there” (SME Venture & CXO XPERTS Interview, 2021). What is at stake in this self-admitted risk or gamble that micro-loan companies are taking by trusting, or banking on, the intent to pay of a consumer/borrower vis-à-vis their actual ability to pay? I explore this question in a later section of this article.

The loan app, self-styled as lending platform, catering to a variety of stakeholders (NBFCs, borrowers, third-party developers), often paints itself as an intermediary merely facilitating or mediating transactions between various customers. In order to examine this lending platform intermediation as a key process of digital financial transactions, I compare the role of intermediaries in erstwhile practices of informal analog money lending with those involved in digital lending. The lending platform as intermediary is fintech's answer to the notoriety and usury of moneylenders/middlemen/loan sharks/brokers who historically used to be the major lenders (and also at times, the intermediaries) in informal monetary transactions in India.

Much has been said about the “distinctive intermediary logic of the platform, which is to make multi-sided markets and coordinate network effects” (Langley and Leyshon, 2017). It is equally important to point out that the loan app as the lending platform connecting NBFCs and borrowers is not the only intermediary involved in digital lending. There are recovery agents as part of the online lending ecosystem who send text messages and make phone calls gently nudging, and at times coercing, defaulters to pay if they do not want to face the consequences. The AI algorithms as well as the APIs and SDKs are the non-human intermediaries along with the human intermediaries, recovery agents, in this digital lending ecosystem. The credit checks, authentication of identity, and loan disbursal which were once performed by human intermediaries are today being conducted by, or automated through, AI algorithms and SDKs. Indeed, one can say, the intermediation of lending platforms, is a form of reintermediation in lending process: “old” intermediaries are being replaced by “new” intermediaries in money and data transaction chains.

Overall, this inquiry seeks to uncover the complex data-based interactions, recordings, communications, and decision-makings that happen across various human and non-human intermediaries (including FaaS players’ APIs and SDKs) in the process of materializing almost instantaneous digital monetary transactions. Borrowing from Dieter et al.'s (2019) “multi-situated app studies” research approach, I situate loan apps within multiple infrastructural settings, including the financial infrastructures operated by credit scorers and identity verifiers as well as the role of human recovery agents and non-human collection agents who make automated calls to overdue borrowers. This multimethod research approach helps open up the opaque intermediation process of lending platforms, which involve a lot more intermediaries than initially become visible (or legible) to the borrower, or in some cases, the lender.

Loan disbursal, or approval, is therefore not one digital transaction, but rather a set of connected transactions (what Athique calls “transaction chains”, see editorial introduction) across platforms and network actors, many of these happening in the intermediate spaces of the multi-situated fintech infrastructure that supports the loan app. This is not just a question of space, but also time. The on-demand platform economy not only requires on-demand delivery of foods and goods, but also on-demand distribution of electronic/digital money. This makes loan approvals and disbursals very swift.

The movement of money is not just a question of space, but also time. Focusing on the “intermediate spaces” allows us to slow down the velocity of financial transactions in order to comprehend loan disbursal or loan approval not just as a single instantaneous process but also as being “marked by moments of suspension and transition,” thereby “revealing how value shifts across a series of linked transactions” in the wider assemblage of the fintech infrastructure (Tankha and Dalinghaus 2020). During the loan approval process, prior monetary transactions, social media activities, and consumer lifestyle purchases stored as tokens and records are scanned in search of alternative datapoints that guide decision-making. This is an example of how digital transactions continuously transform money into data and vice-versa (Athique, editorial introduction, this collection). These remediations of data into value form are translated through scales and temporalities in the multi-situated intermediated spaces of loan apps.

The delivery of loans by lending platforms like many other fintech products might seem unmediated and direct, but through careful research, one will find myriad intermediaries in the fintech infrastructure that drive automated decisions. Mrinalini Tankha and Ursula Dalinghaus (2020: 345) characterize the “intermediate” as the “topography of in between spaces, and the multiple agents, technologies, and socialities that serve as intermediaries.” They explain the “mapping the intermediate” as a “crucial means of approaching the politics of financial transactions and their social and material stakes” (Tankha and Dalinghaus, 2020:345).

Methodologically, the mapping of the lending platform ecosystem that I present in this article is drawn from formal and informal conversations in fintech festivals and conferences, interviews with investigative financial journalists, professionals in loan app organizations, consumer rights advocates, and forensic data experts. I have also undertaken document analysis of several regulatory reports and circulars, analyzed loan app ads, and extensively read financial and technology newspapers, magazines, and industry periodicals.

“Instant” imaginaries and new socialities of dependence and anonymity

I will begin on the consumer side with a survey of the advertisements and front-end interfaces of both legitimate and predatory loan apps, and then move to the next section discussing the inner workings of loan apps and backend fintech infrastructures including the political economy of the lending platforms and their positioning with respect to customers and lenders.

The Navi loan app roped in former Indian cricket captain MS Dhoni as its brand ambassador in 2022. In one ad, Dhoni is shown to be tied to a chair with several gangsters threatening him, demanding Rs 20 lakh rupees as part of a kidnapping set up. Dhoni is making calls on his phone to a loan service company, and asks for a loan immediately, but the loan app company official on the other side of call says that in order to verify Dhoni's bank statement, salary slip, and identity verification, it would take about 7 days (3 banking days and 4 working days), after which the money will be deposited to his account. The gang leader cannot wait for 7 days. He needs the ransom money now and conveys his impatience by angrily crushing an apple with his bare hands. This makes Dhoni even more perplexed and anxious. The situation improves when Dhoni is addressed by a Navi agent's voiceover who says Dhoni can receive loan at his time, on his conditions, and he just needs to download the Navi loan app, and through a paperless process obtain loans upto 20 lakhs. (Navi Ad, 2022). Navi's personal-loan ticket sizes range between Rs 5000 and Rs 20 lakh, and typically, the loans tend to be much smaller than 20 lakhs.

Some of the earliest micro-loan digital players like Fibe, which started with the name EarlySalary in 2016, often advertized use-cases which were about early career professionals facing cash crunch close to the end of month. They had still not received the salary, and hence were struggling to buy something they required. They could then avail of EarlySalary loan so as to buy a household appliance or purchase a holiday plane ticket. Several other loan apps such as CASHe and KreditBee also have similar use-cases such as small ticket loans for laptop repair, scooter to motorbike upgrade, buy a sari during Diwali or not having to deal with the fear of missing out (FOMO) on airpods. In each such advertized case, there is an emphasis on how quick the loan approval process is. Such claims of instantaneity are proudly declared in fintech conclaves and pitches. Akshay Mehrotra, Co-Founder and CEO, of Fibe (EarlySalary), in a podcast interview with Founder Thesis claimed that Fibe was the “fastest lending product available in Asia” with an “average loan disbursal window” of “122 s.” (Founder Thesis Interview, Mar 15 2024)

Another pattern across various loan app ads is that they promise to mitigate the risk for borrowers of being publicly embarrassed or humiliated by the lenders, who can be their friend or father. CASHe's ad starts with a couple enjoying their date at a bowling alley with popcorns. Their date then gets interrupted by the guy's friend who sarcastically embarrasses the guy by letting his girlfriend know that while they are enjoying a great lifestyle, he has not yet repaid the loan the friend had provided him. The girlfriend seems irritated and clearly the date has gone awry. The CASHe ad notes that one should never ask a human for loan, but rather ask CASHe. CASHe will help a borrower receive a loan without losing their “izzat” (honor/reputation) (CASHe Ad, 2016). These ads emphasize the embarrassment that borrowers suffer if it is made public that they are a defaulter or unable to repay back their friends and relatives. This is an important aspect of socialization around money lending in India. Enjoying an amazing lifestyle on borrowed money has been looked down upon in the country. The loan app affords a far more discrete and anonymous way of receiving loans, where one no longer risks being accosted by an annoying friend or relative in a restaurant or bowling alley, asking for their lent money back.

This myth that loan apps were an embarrassment-free mode of receiving money was decisively broken when during the Covid-19 pandemic, the borrowers who defaulted found out that the predatory loan apps had called/messaged their friends and relatives whose phone numbers they had gathered from the borrower's contact list. This contact list was accessed by the predatory loan apps during the loan onboarding process through sometimes an explicit, and sometimes an implicit, consent provided by the borrower themselves.

During the course of my research, when I tried using some of these loan apps such as TryCash and FlashRupee (both of them no longer exist as they were taken down) on the Apple and Google stores, I was prompted for government ID as well as a video selfie, including explanations for how to take the selfie. After that I was asked for a range of device permissions including contacts, call history, and SMSes and I could not go further to the next steps of loan onboarding unless I agreed to those permissions.

The public shame some of the borrowers encountered when they found out that their relatives and friends knew they had defaulted lead some to even commit suicide. As the materiality of money shifts and as modes of circulating digital money changes, new socialities of dependence, obligation, and anonymity around money also keep mutating. The loan app ads seem to be prescribing that their customers now shift their dependence on loans from relatives and friends to lending platforms. Some ads show protagonists who feel obliged to publicly clap on and appreciate an uncle's poor jokes or an aunt's bad singing. This obligation that comes with taking a loan from an uncle or aunt, according to a Navi ad is like carrying baggage or load with oneself, something that taking a loan from a loan app relieves one of. The Navi ad sums it up by saying that borrowers “need not take load” (“loadmatlo”) and only “take loan” (“loanlo”) (Navi instant personal loan ad 2021).

Such loan app imaginaries exhibited on ads and interfaces attempt to create particular kinds of financial subjectivities in their address to potential customers, but do not say much about the actual inner workings of a loan app.

Operating relations: lending platform and NBFC

To go deeper into the transactional architecture of loan apps, I will explore the operating relations between lending platforms and NBFCs, as well as lending platforms and third-party developers (FaaS players) who provide the technological infrastructure for automated payments, risk score checks, and loan management. These operating relations are typically strategic and contingent, and shape the data and money transactions involved in mobile lending (even more specifically: small ticket size personal loans) in India. In today's fintech landscape in India, to set up a lending platform, one needs to enter into an arrangement with one or more banks and NBFCs so that the loan app has access to capital for providing loans. The loan app is, after all, a technology company, and needs to either invest in in-house technology for credit scoring, loan management, and collection processing, or pay third-party developers (FaaS players) to have their SDKs offering these services be plugged through APIs into their system. So, the relationship with NBFCs is one dimension of the digital lending business and the other dimension is to manage the technology side of things. In some cases, as I have pointed out earlier in this article, a loan app might have an NBFC arm but even in such cases, given the large amount of loans to be transacted, the loan app needs to enter into arrangements with numerous NBFCs. Furthermore, the money raised from VC funding cannot all go to the NBFC arm of a fintech loan app because a significant chunk of that funding is required to be invested in tech innovation. These investments in technology include improving AI algorithms or building inhouse loan management capabilities or hiring data scientists. For now, I will begin with NBFC-loan app relationship and then move to the loan app-FaaS players relationship.

The NBFC-lending platform relationship becomes a way of understanding how legacy banking architectures and regulatory norms in India interact with capitalization of fintech startup firms. Many of the loan apps as fintech startups are backed by venture capital funds. More than banks, it is the NBFCs who have shown much interest in lending to loan apps as they see opportunities for themselves as well. A lending platform as a startup can always raise VC funding, but it cannot use that money right away to lend by itself. They have to either obtain an NBFC license or tie-up with a NBFC. This is because in India, only banks and NBFCs as regulated entities are allowed to lend money.

What becomes an area to focus on then are the kinds of risk sharing arrangements that loan apps get on with NBFCs to make the deals lucrative for both. Some lending platforms, plush with investments from VC funding, are eager to show their investors that they have reached out to so many customers and that they have disbursed a large number of loans. Given the investor backing, they are happy to continue lending even to customers who do not repay back in time. The NBFC though can get worried if a large number of loans are not getting repaid because according to the protocols set by RBI, it is NBFCs who are finally responsible in terms of covering losses if the borrowers default (RBI Circular 2022). As part of the lending platform and NBFC negotiation, there have been situations when to win over the NBFC's hesitation, the lending platform has gone ahead and said that they will cover the losses in case the borrowers will default (Singh 2021).

During the Covid-19 pandemic, the NBFCs were especially worried that the borrowers will not be able to repay, and so some lending platforms following aggressive tactics, assured them that they will completely cover the losses. An almost unwritten rule is that the fintech loan app company can offer to cover up to 5% of the losses in a first loan default guarantee (FLDG) clause as part of the contractual agreement between the lending platform and the NBFC. During the pandemic, as investigative financial journalist Arti Singh (2021) wrote, some loan companies offered NBFCs “100% FLDG” which meant that the “fintech partner, which finds the actual borrowers, will take on all of the potential losses, an excellent proposition for an NBFC.”

Some loan companies were essentially ready to part with their money to NBFCs as collateral for taking care of 100% losses right at the beginning, a very attractive offer for NBFCs, who with this kind of arrangement were no longer engaging in lending operations but earning off of their lending license. Also, fintech insiders, who spoke off the record to me, noted that NBFCs are much less regulated than banks, and while the top 100 NBFCs might be less interested in compromising on rules, thousands of other smaller NBFCs who are not always under the Reserve Bank of India's radar, are fine to adjust a bit, especially when they see the potential for significant profits. Noting these indiscretions, the RBI has reiterated in recent circulars that NBFCs are required to adhere to FLDG guidelines where a third party cannot compensate beyond a certain point of the loan portfolio in loss sharing arrangements (RBI circular 2022).

This kind of deep-dive into the political economy of the NBFC-loan app interactions and negotiations is crucial because they condition the kind of digital financial transactions possible. A loan company startup would proudly tout how they work with many different NBFCs as lending partners, which provides them with multiple options when they have to find the best possible match for a potential borrower in terms of interest rates and repayment windows. What they do not disclose are the backchannel conversations that are part of attracting the NBFCs to them. Without these contingent negotiations, which some industry insiders I interviewed termed “workarounds,” NBFCs would not participate in lending in the unsecured personal micro-loans sector, and this would mean fintech lending startups would not be able to scale up their number of borrowing customers. Reaching out to as many customers as possible, to reiterate, is a key target for lending platforms in order to please their VC investors.

The lending platform therefore acts as the “performative business model of capitalist intermediary enterprise” (Langley and Leyshon 2022). As an intermediary between lenders and borrowers, the lending platform does not provide the details of its interactions with NBFCs to its borrowers: all it typically says in the form of a disclaimer during the loan transaction process is that the loan is subject to the credit policy and terms & conditions of the registered NBFCs/Banks [lender(s)] participating.

Beyond the money earned from interest rates, loan apps can also significantly benefit from processing fees, the fees levied to cover the cost of evaluating these loans. In a recent piece for The Ken Anjali Jain (2024) found out that the lending platform Kreditbee charged borrowers “anywhere between Rs 500 (US$6) up to 6.5% of the total loan amount” as processing fees. This charged fees was separate from the additional onboarding fees, documentation fees, and a host of other charges the loan app charges. These partially legible costs accompanying money and data transactions need more research.

At the extreme end of the spectrum of predatory loan apps, there exist such lending platforms that have neither their own funds or any tie-ups with NBFCs. They depend on interlinking their various problematic (shady) enterprises including lending, gaming, and crypto-trading, where the loan disbursal for one customer comes from the top-up amount filled/gambled by a gamer: this is a form of peer-to-peer (P2P) lending happening across inter-connected lending and gaming/gambling platforms. While I have not found a way of verifying whether such practices exactly unfold like this, some hackers and consumer rights groups pointed out to me that they think this is what is happening because the loaned money they saw coming to customer accounts were not being received from an enterprise's bank account but rather from the personal account of another individual.

As we move from the financial side of the lending platform ecosystem to its technological side, we realize that the financial and technological stakeholders are related to one another, and it is the loan apps that mediate between them, and in some cases, selectively so. On the technological side of the loan apps are the third-party developers for APIs and SDKs for risk scoring and loan management services, also called fintech as a service (FaaS). The NBFC partner may not even know what FaaS player the loan app is partnering with for a particular service. This shows how a multi-situated study of loan apps is so important to understand software as infrastructure and how apps are situated within varied infrastructural relations, such that the work of FaaS remains almost hidden from the NBFCs.

FaaS players: tech Side of lending platform ecosystem

Let us now shift the conversation to the technological side of the loan app that is the fintech players. It became increasingly apparent over time that a company interested in building a loan app could approach different fintech players and they could just plug the different services through APIs and SDKs into this lending platform. These predatory loan apps which aggressively approached the fintech market during Covid-19 were mostly white-label based, meaning they are essentially deployed out of the same templates. If and when one loan app was identified as fraud and brought down on Google PlayStore, the company behind it remained the same, and that company could just partner again with the same set of stakeholders. They could just approach once more the three or four white-label providers, and then produce an app with a slightly different name and brand with the same stakeholder SDKs/APIs working in the back end. These predatory app companies were tying up with NBFCs on the other side, and these NBFCs had little idea of what was going on the tech side.

During the microloan controversy which happened around the time of the pandemic, an NBFC like Inditrade remained unaware that some of the loan apps it was partnering with were using identity-verification services from the “Advance AI,” a company working on AI-driven eKYC (Know Your Customer) (Ramanathan 2021). Some of these verification methods, as explained by Cashless Consumer experts Suman Kar and Srikanth Lakshmanan, entail taking a selfie at the time of loan onboarding, which can potentially be transferred to a server in another country. This server then checks whether the borrower is a real person, after which the information is passed back to the app. These flows of data across national boundaries raise concerns, as some of these ID-verification techniques involve liveness detection and facial recognition data, which can then be mirrored in other databases (Mukherjee 2024)

Scholars examining the relationship between apps and infrastructures stress that apps are not fixed objects but are entangled in data flows across digital infrastructures, and thus contingently transformed "through interactions with diverse socio-technical situations" (Dieter et al. 2019:1). The activities of loan apps thus go beyond the individual or company level and can possibly register effects at the level of transnational geographies and infrastructural platform services. As Arundhati Ramanathan (2021) notes, “The use of SDKs and APIs by these FaaS players has allowed the Indian startup ecosystem to be competitive. It allows startups to take on digital behemoths by giving the same level of service.” At the same time, these FaaS intermediaries may actually be retaining data that they are just supposed to pass on to the NBFC or the loan app. Furthermore, the loan app company as a digital lending platform seems to be selective in the way information moves between the various stakeholders it is mediating between, that is, here, between NBFC and the third-party developer.

It is important to add a caveat that while all loan apps depend on third-party FaaS players to some extent, the lending startups who are absolute beginners tend to rely more than others. Some of the legitimate and established loan app companies like Fibe and KreditBee have over time built some of the fintech services inhouse. The smaller fintech loan apps tend to be API mashups of various third-party providers and as they mature, they get internal data pipelines, inhouse risk scoring, and custom data-sources based credit scoring within their tech stacks.

Just the sheer number of stakeholders in the FaaS infrastructure thoroughly contradicts fintech's lie (or promise) that there are no intermediaries or human mediators between the borrower and the seller in digital lending. Intermediaries, both human and nonhuman, have increased in fintech operations. In the case of predatory loan apps, these intermediaries play crucial roles in shaping the shifting financial subjectivities of borrowers who suddenly realize that their user data has been compromised—that their parents, friends, and relatives know they are defaulters. The irony is that recovery agents might even know about the defaulter's inability to pay, and they are simply forcing defaulters to ask their relatives and friends for money. The recovery agents are themselves under pressure from their bosses, and so they keep on with their mediated threats through calls and text messages.

Loan recovery agents in the fintech world operate through tele-centers or call centers making calls to borrowers, reminding them that they need to pay the interest on time or that they have defaulted. While there are some bot nudges through robotic calls or text messages, in most cases, it is humans who call. The recovery agents, thus, are a crucial intermediary of the lending platform ecosystem along with NBFCs and the fintech APIs and SDKs. Even in the case of informal moneylending, there are recovery agents, who are often stereotypically portrayed as “loan sharks.” The traditional formal banks also tend to send collection agents to defaulter's homes, if need be, to send warnings. The one key difference in the fintech lending scene is that the recovery agent appears as a voice on the other side of the phone calling the customer/borrower/defaulter. The anonymity granted to the recovery agent in the form of the phone call (vis-à-vis in-person visit) can be abused to pronounce threats. This anonymity was opportunistically used as part of the aggressive playbook deployed by predatory loan apps to extract as much money as possible from hapless unemployed defaulters during the Covid-19 pandemic (Christopher 2021).

Several legitimate loan apps have also been found, at times, to have resorted to questionable collection strategies as revealed in the BBC documentary on loan app scams (BBC Documentary, 2023). Some fintech insiders I conversed with noted that legitimate loan apps after a point tend to outsource the collection work to a telecenter. Some predatory loan apps consider the collection aspect to be the key place of profit and keep it inhouse. In fact, in many instances, the only office spaces which I could locate of fraudulent loan apps were the call centers they (previously) ran to extort money from hapless clients during Covid-19 pandemic. PC Financial, one of such key predatory and problematic lenders, had an office space in Gurugram Udyog Vihar Phase 1 Sector 20, where the company label on the top of the building read “iEnergizer (business process excellence).” Several other offices of PC Financial had closed by then as it was being investigated by the regulators.

Transaction, trust, and behavioral finance

Whether it is informal or formal lending, the fundamental question is always whether the borrower can be trusted to repay. In informal lending, this assessment is based on whether a person in a locality is known by others, or knows friends who can vouch for them. If the person, for example is a migrant, writes Tripta Chandola (2018), and therefore is new to a locality with no past history of having lived in the area or any immediate friends who know them, the terms of advancing a loan would be different. The migrant would have to deposit their voter id card or ration card as collateral. This is so that the moneylender can trust this migrant to stay and not leave/flee the neighborhood with the money. The migrant, being a stranger to the neighborhood, might be asked to pay a higher interest rate (“biyaaz”) on the loan as well, but over time, if the migrant turns out to be a dependable customer, the interest rates will be decreased each time a new loan is disbursed. This is what Tripta Chandola found out based on her extensive long-term ethnographic fieldwork in the Govindpuri resettlement colony in South Delhi. Chandola (2018) makes a crucial argument:

“the informal nature of these monetary exchanges does not imply an informality of the social structures and networks within which they unfold. In fact, these informal monetary exchanges have a history in the slums, which in turn accords them their longevity and robustness as a viable and reliable economic practice, as also accruing the ‘trust’ both amongst the borrowers and the lenders.”

At one level, we are told that fintech loan apps are formalizing monetary exchanges because they are digitizing them and there is a record of these transactions. At another level when first-time borrowers are undergoing the loan onboarding process, the loan app has little idea of the history of social structures and networks that this borrower is part of, especially if the borrower mostly socializes offline. These social structures in a deeply hierarchical society like India might be repressive, and the loan app might be way out of these oppressive socio-economic hierarchies. At the same time, it is also possible that these social structures, as part of quotidian lived embodied relations of this person can be networks of care, trust, and reciprocity for them.

This also brings up the related question of how these fintech loan app startups decide who to provide loans and who not to? A popular answer to this question for some time now has been that they base it on “alternate data” or “alternative data.” With not much credit scores data available for most people in India, but with so many people on their phones leaving traces of their social media, e-commerce, and other app activity, it is not surprising that the fintech industry desires to examine all kinds of unstructured (behavioral) data which they are now able to gather through access to customer's phone information. Given that the unsecured personal loan market is so precarious, fintech companies are trying to devise algorithms so as to gauge risks and consumer intent better. CEOs and CTOs of Indian lending platforms have often openly confessed that in many cases, they are not providing the loan based on the capacity of the borrower to pay but rather on their intent to pay. In a recent interview, the CTO of CASHe, Yashoraj Tyagi explains how his lending platform from 2016 onwards built out an algorithm that goes beyond the usual credit scores to assess large cohorts of customers’ data sets that are completely independent of financial attributes:

Suppose I don’t see how you are on bill payments, credit card payments, and loan payments. I just see what kind of places do you shop at? Where do you, how do you transact every day? What kind of places do you stay in? … How do I just look at just broad lifestyle indicators and create a score out of that, that could be correlated to your financial behavior with respect to lending…If you use premium lifestyle apps on your phone like for example if you use a Vistara app on your phone compared to using a Make My Trip…it tells me that a customer is in a certain lifestyle band…it tells the model a lot more about the customer…what is the propensity of a customer to repay credit…what is the trustworthiness indicator of a customer (Tyagi in an interview with Ayush Wadhwa, 2024).

When a customer is trying to get a loan by being on Tyagi's CASHe app, the loan app has ways of finding out the other apps the customer has installed on their phone. It makes a difference if the customer uses Vistara, a premium lifestyle/travel app vis-à-vis Make My Trip. The pin code (zipcode) of the person's address or where they shop makes a difference for the purposes of determining loan amount, interest rate, and repayment window. Aggregating data across these touchpoints helps CASHe put together what they call a “social loan quotient” (SLQ). SLQ relates social behavior to long-time financial behavior. In other words, SLQ represents the calculation and quantification of the customer's intent to pay. This calculation and quantification of customer's intent is being dynamically undertaken in real-time as the customer is being onboarded onto the CASHe loan app. This calculation and quantification requires automation provided by the plugged SDKs.

Discussion

In this article, I have mapped out the spaces of intermediation comprising the platform lending ecosystem as a way to slow down the instantaneous monetary transactions involving loan approvals and disbursals, and explain just what goes on at the backend fintech infrastructure when such transactions take place. I have analyzed the relationship between the various third-party developers and the loan apps in terms of the financial software services provided through SDKs and APIs, while also exploring the political economic relationships between NBFCs and lending platforms.

While I have traced the distribution supply chains of payments and lending through the fintech infrastructure, I have not done away with representational analysis. Rather, I believe in juxtaposing the digital materiality of infrastructural backends as well as representational dimensions of platform frontends, like interfaces and ads. This allows us to remain attentive to the cultural politics of digital lending in India, as signified within the advertisements and interfaces of loan apps. Digital lending is new in India, and the ads of loan apps are a way of inviting millions of new-to-mobile phone users to be part of novel kinds of transactional communities. In these emerging transactional communities activated by the on-demand platform economy and new monetary circulations across mobile phones, promises are made of digital records replacing paper as an inscription (medium) of transaction (Swartz 2020).

One place where the various intermediaries involved in the lending platform ecosystem and fintech infrastructure can be found to gather together is the fintech conclave. I attended the Fintech Festival India held in early March 2024 at the Yashobhoomi convention center in Dwarka, New Delhi. Here is where I found stalls set up by loan app companies such as Fibe advertising that they cater to the aspirational youth as well as FaaS players such as Mobilewalla who promised AI driven APIs and SDKs to approve thin-file customers and thereby enable financial inclusion. The statements they had on their pitch deck dashboards could be dismissed as glib corporate talk, but at the same time, there were eager professionals representing third-party fintech developers who seemed keen to take me through a flowchart of how they conduct face match and liveliness check as part of their offered suite of instant verification APIs for onboarding and digital KYC. The pictures of those who had their Aadhar (unique ID) cards made early had to be matched with their recent face pictures, and this required a “live” selfie to be taken during the loan app onboarding process, and more nuanced algorithms were needed to be able to make sure this was the same face of the same person on the Aadhar card. As much as this sounds like a techno-utopia, this was also a certain kind of normalization of digital surveillance. These acts of data extraction as part of the loan onboarding process are a part of both legitimate and predatory loan apps, and so it becomes difficult to distinguish between the two, given the similarity of surveillance practices involved.

The kind of state-corporate support driving fintech cannot be underestimated if one sees the execution of such a festival in the state-of-the-art convention center in Delhi. That said, some of the hype and boosterism that followed fintech lending in India around 2016 took a scary turn with the rise of predatory loan apps during Covid-19 pandemic and the regulatory backlash from RBI following that. Some financial journalists have opined that fintech like some other sectors such as Agritech, Edtech, and short-video streaming platforms, has failed to live up to the initial hype that was catalyzed by VC funds and investor optimism (Manikandan 2024). At the same time, fintech companies like Fibe and CASHe feel confident that their investments in tech (AI algorithms and alternative data) have brought them new customers, who have now become repeat customers, and have developed a credit score over time (since 2016) when initially they had no credit history.

What this perhaps suggests is that only the more robust lending platforms with less financial dependencies on external NBFCs and less technological dependencies on external FaaS players will survive. In studying fintech lending, I have attempted to trace the (un)intended transformations of data into value form in platform economies, which are marked as much by optimizations and conveniences as they are by predatory inclusions and debt traps. Skeptics of the Indian fintech lending story, some who are even industry insiders, have in interviews with me, questioned the tall claims of financial inclusion that are touted by lending start-ups as part of their presentation decks in front of their VC investors. Their frustration is understandable given the continuing gender inequities with respect to accessing loans and a growing perception within the industry (and even RBI) that loan apps are mostly catering to urban millennials, and not really serving the marginalized Indians in whose names they initially entered the Indian lending arena. Some of the same skeptics also express hope that as the fintech industry matures, they will be able to devise better loan underwriting models that more comprehensively take into account the on-ground (actually existing) social relations that shape lending in India.

It is important to take into account what socio-economic processes are being set through the fintech mandate marked by a tight coupling of NBFC-loan apps that serve both public–private interests. For some time now, by promoting easy and a wide variety of unsecured loans (based on behavioral phone data) and buy now-pay later consumption models, the NBFC-lending platform tie-ups are financially socializing, that is encouraging, mobile phone users into making use of mobile payments and loan apps. The government has shown regulatory laxity in the initial stages to encourage market demand, help microentrepreneurs, and increase digital money circulation, though the RBI has lately tightened its regulatory screws around the NBFC sector, especially regarding unsecured loans amidst risks of credit bubble and unmanageable debt. Saloni Shukla (2025) reports that fintech companies lately have been verifying the end-use of loans, and offering unsecured loans for medical aid, major life events, and for assisting small business owners, and have cut down on personal loans for buying consumer durables. Based on RBI's encouragement, lending platforms are utilizing the services of account aggregators and AI to carefully track end-use of loans based on financial data pulled from banks, insurance companies, and payment services. As the principal monetary regulator, RBI thus has facilitated active consumption in Indian society at an earlier point and now seems to be shifting towards funding productive sectors.

Corporate ads about mobile payments and digital lending seem to extol instantaneity, immediacy, and convenience almost to the point where users are tacitly encouraged to provide consent without bothering about surveillance or privacy. This is further aggravated by the incentivization of particular behaviors/consumption patterns enabled by clicks, swipes, and taps. However, Beni Chugh et al. (2017) and her Future of Finance team at Dvara Research found that—contrary to conventional wisdom that Indian citizens (many of them new to the internet) are focused on access and convenience and are indifferent about surveillance—“people in India care deeply about their personal data and privacy.” Some borrowers might be okay to trade their social media data for credit but not their identity documents and locational information, especially when this data could be disclosed to third-parties.

Noting that the SIM card might be taken on a wife's name, but most of the day the phone is being used by the husband, a lending company professional having worked in the fintech sector for twelve years admitted to me in an interview that the loan underwriting models for women have to be very different if they have to truly reflect their phone based social or financial behaviors. Here we find another layer of intermediation not in the fintech backend, but in the way women's access to phones and loan apps is mediated by male relatives amidst gendered and caste hierarchies in India with implications for calculations of creditworthiness. The changes and continuities in both mobile usage practices and lending processes have to be traced in any critical evaluation of the reconfigurations of data, intermediaries, and financial technologies in India's mobile ecosystem.