Before he arrived in Mexico, Delhi-born Arpit Gupta designed and built Formula race cars in Europe. Back in India, he had a brief stint as a product manager at a fledgling digital payments company — one which, at the time, had fewer than 20 employees. That company, Paytm, which he left in 2014, is about to go public in what’s likely to be India’s largest ever rupee-denominated IPO. Gupta, now half a world away in Mexico City, is trying to replicate his former employer’s successes and build a mobile payments company in a country that’s still stoically clinging onto cash.

Paytm began as an online service for mobile top-ups — a great use case for digital wallets, recalled Gupta, who is now 39 and has close-cropped hair and gently graying sideburns, in an interview with Rest of World in the upscale Mexico City neighborhood of Polanco. “Consumed on mobile, needed on mobile, but the transaction still happens offline,” he said. “Why? That’s how it started.”

Even in 2014, Gupta believed there were enough similarities to be able to “copy-paste across any emerging market.” Armed with that belief and heartened by the inexorable expansion of Paytm across India, he moved to Mexico’s capital and launched UnDosTres with his former business school classmates: “a one-stop shop for middle-class Mexicans,” Gupta said, to pay their bills, buy movie tickets and bus fares, and top up their mobile phones. It was to be a mix of Oxxo (Mexico’s answer to 7-Eleven), Elektra (an appliance store with bank and money transfer branches), and Coppel (a retailer and lender for working-class Mexicans). 

“But when we started,” Gupta said sheepishly, “we started naively. And we got hit bad.”

Over the years, Mexico has remained stubbornly wedded to cash, causing even e-commerce juggernauts to buckle and improvise. According to the 2020 McKinsey Global Payments Report, nearly 86% of all transactions in Mexico take place in cash. Even online shoppers who order goods on e-commerce platforms, like Amazon or Mercado Libre, prefer to pay at their nearest Oxxo. This is true for everyday transactions, too, such as cellphone top-ups and utility payments.

This is all well and good if one lives in the right part of town. But, “In middle-tier Mexico, you don’t have Oxxos everywhere — you need to walk 10 blocks,” said Gupta. “A family of four tops up eight times a month; they pay their bills two to three times a month. So that’s 10 to 12 trips a month: a lot of time, energy, and effort.” 

UnDosTres set out to address the onerous nature of everyday transactions in Mexico, but Gupta’s experience in India hadn’t prepared him for a uniquely Mexican problem: singularly high rates of card fraud. According to Mexico’s financial consumer watchdog, the country experiences an average of 463 fraudulent digital transactions every hour. (Card fraud in India is comparatively lower, although it has recently risen dramatically.)

Old and laggard banking infrastructure is to blame, said Gupta. Mexico’s banks have no incentives to control fraud. “It’s the merchants who lose money, mostly. And the regulator doesn’t step in.” In India, there are penalties if card fraud exceeds a certain amount — and if it becomes a trend, banks can lose their licenses. The absence of regulation in Mexico makes “tackling card fraud a completely different ball game here,” he said. 

When it launched in 2015, out of every $100 transacted on UnDosTres, $40 were fraudulent. 

Gupta could see the writing on the wall. Faced with bankruptcy, UnDosTres opted to start using third-party fraud-detection tools. These tools either delayed transactions in order to allow a manual review of user details or relied on standard identity verification, such as security questions. Fraud rates dwindled to between 12% and 15% — down from 40% — but it wasn’t good enough: unless the process was completely frictionless, why would anyone choose to top up their cellphone or pay their bills on UnDosTres? “To have a fast-growing, successful business,” said Gupta, “we had to build our own tech.”

The Mexican payments fintech landscape, argues Gupta, is littered with failed or failing apps that never quite conquered card fraud. But Eduardo Licona, a fintech strategist based in Mexico City, said the real challenge is that to use a mobile wallet, you either need to link it to an existing bank account or top it up with cash. “If you go to an Oxxo to put cash in your digital wallet to pay your bill,” he explained, “you might as well avoid that extra step and just pay your bill directly.” 

In 2018, 63% of Mexicans over the age of 15 were unbanked. Latin America’s second largest economy is a bit of an anomaly in this regard, where regionally the rate is 44.6%. Even during the pandemic, as people across the world shifted toward banking services at unprecedented rates, the amount of cash in circulation in Mexico actually increased, up almost 24% in November 2020, compared with the previous year.

“Mexicans love cash” is a common refrain, but a more accurate explanation may be a severe lack of trust. They don’t trust banks — and so don’t set up accounts — and they don’t trust the rule of law or each other — understandably so, given the high rate of fraud. This trust deficit links both Gupta’s and Licona’s explanations for why digital payments have taken so long to catch on in the country.

Lack of trust manifests in two ways, Barbara Magnoni, a financial inclusion expert, told Rest of World. “One is ‘I don’t trust this company to take care of my financial situation,’” she said. “The second is ‘I don’t trust myself to do a good job navigating this app — if I push the wrong button, maybe my money will disappear.’” Magnoni acts as an intermediary between fintech apps like UnDosTres and cash-reliant Mexicans: she is the co-founder of Noahui Soluciones, a company that tests existing fintech products and trains local women over the age of 35 — Noahuis, or “aunts,” in the native language of Nahuatl — to promote them within their communities. 

Why there is a lack of trust in companies is clear to Magnoni. “Mexico is a country with a very strong bank lobby,” she said. This has impacted financial digitalization by discouraging people from opening accounts and by creating a regulatory ecosystem that deters newer competitors. “There are a lot of disincentives,” said Sandra L. Suárez, professor of political science at Temple University. “Fees are high, and, in most banks, you need to have a minimum in order to not be charged.” 

“[Mexico is] using India as a model of success without understanding the limitations.”

In 2016, Suárez published a comparative analysis of mobile payments in Kenya and Mexico: upward of 50% of the population used mobile payments platforms in the former, compared to barely 2% in the latter. Kenya has less-developed infrastructure and lower GDP per capita but has still managed to become a pioneer in fintech. Both countries had similar conditions for mobile payments to take off — high rates of financial exclusion, widespread mobile phone use, and high rates of remittances leading to a latent demand for financial services — but what stifled growth in Mexico, Suárez argued, was that banks were allowed “to insert themselves into the regulatory-making process for mobile payments from the very beginning.”

Gupta agrees that trust is the main missing link from what is a preexisting digitally savvy Mexican middle class.“More than 50% of our users had never made a digital transaction,” he said. “They spend a lot of time on Facebook, on YouTube, on the free version of Spotify, but won’t transact online.” Digital payments are intended as a gateway, he explained; if UnDosTres could just nudge these banked Mexicans, Gupta could sell them an array of other financial products, such as access to credit — which UnDosTres provides, by connecting its users to third-party lenders as well as through its own credit card, currently in pilot mode. 

According to Licona, mobile wallets are only really able to remain afloat if they diversify into other financial services, such as lending — rather than focusing exclusively on card fraud. Paytm has been trying to do exactly the same in India.

Still, Magnoni is nonetheless wary that Mexico is “using India as a model of success without understanding the limitations.” She remains skeptical of the premise that India’s digitalization has been as transformative as is touted in fintech and startup circles. “That’s not to say financial technology isn’t working in India,” she said, “but it’s not reaching the types of groups that the narrative claims.”

Gupta and his co-founders have clearly heeded this warning: They aren’t looking to cater to the vast number of Mexicans who remain unbanked. Rather than targeting “the next billion users,” they want to reach middle-class Mexicans who already have bank accounts and use the internet for communication and entertainment.

The moment you download UnDosTres and fire up the app, it begins collecting information about you: from straightforward identifiers, such as location, device, card, and bank, to how quickly you type, where you’re clicking, whether you logged in with Facebook or Google. 

Dissatisfied with the rate of card fraud, the company dispensed with third-party software early on and instead began developing a self-improving algorithm, employing dozens of parameters to screen each transaction. 

The technology has helped whittle the rate of card fraud to below 1.7%. — the industry benchmark, Gupta said, is about 6% or 7% — and maintain a high rate of successful transactions. “Of every 100 payment attempts on UnDosTres,” Gupta said, “more than 80 go through successfully. The industry average is 30%–40%.”

UnDosTres also employs user data — whether you’re paying your electric bill on time, for instance — for its lending services: its middle-class clientele is normally deemed too risky and subsequently neglected by banks. “But we have alternate data and information patterns on our users that we can use to prevent fraud as well as to underwrite loans,” said Gupta. 

“People don’t quite realize how much data is being collected.” 

This data-driven solution has allowed UnDosTres to thrive, but it may also be the source of its own undoing in a market undermined by chronic mistrust. “There’s no regulation against automated decision-making in Mexico,” said Grecia Macías, a lawyer with Red en Defensa de los Derechos Digitales (R3D). There’s also no transparency on the part of financial institutions, she added.

UnDosTres tracks behavioral as well as device data — whatever can help in combating fraud. Critics warn that in the long run, this may perpetuate existing biases: algorithms can easily use alternative data as proxies for protected groups, such as zip code to infer race, or height and weight for gender. Gupta argues that the likelihood of bias is low because there is no human intervention in its lending algorithm. 

Not everyone may want to make the Faustian bargain of relinquishing privacy for convenience. “When people say, ‘There’s no privacy in 2021, ha ha,’ it always gives me a heart attack,” said Macías. “People don’t quite realize how much data is being collected.” To make the bargain all the worse, in Licona’s experience, most of the data collected isn’t a particularly good predictor of people’s ability or willingness to pay back loans, but conventional wisdom in the tech sector dictates that some data is probably better than no data at all.

Using data, UnDosTres — with upward of 2.5 million users in 2021 — may have overcome its fraud problem, but the broader digital payments landscape in Mexico still appears rocky. Just before the pandemic, Mexico’s central bank launched Cobro Digital (CoDi), a real-time mobile payments platform using QR codes. UnDosTres was only part of the pilot program, but Gupta watched the rollout of CoDi closely, to get a sense of which way the financial digitalization winds were blowing.

He was disappointed. Within a year of launching, the initiative garnered 6.4 million users, far short of the goal of 18 million. The platform registered just over a million transactions so far; the central bank was aiming for 28 times that amount. Trust in digital payments is, for the moment at least, still a pipe dream.