The floor is shaking in this sun-drenched, airy downtown Toronto office. Blame that on the two young men, one in shorts and a soccer jersey, thudding around the ping-pong table just off the centre of the room.
About 50 engineers, developers and data scientists—one of whom formerly worked at CERN, the European research facility which discovered the Higgs Boson particle—work in this two-story brick building in the Entertainment District. Their employer is not a buzzy homegrown startup or the Canadian branch plant of a Silicon Valley giant. Paytm Labs is the R&D arm for hyper-growth Indian financial technology firm Paytm. It’s led by Harinder Takhar, a Blackberry alumnus who started this facility in 2014 after three years as the parent company’s first CEO.
Paytm first gained traction by facilitating mobile prepaid phone recharges. Unlike in Canada, where most wireless customers opt for a monthly bill, many consumers in India run on a pay-as-you-go basis, buying packages of voice, text and data which expire after a fixed time. These refills are typically sold by convenience stores and cigarette stands, sometimes in the form of scratch cards bearing 16-digit numbers that users called to top up their balances. “It is so much less convenient than [topping up directly on your phone],” says Takhar. “One of the things we focused on was, ‘Can we make sure that every transaction happens in less than 10 seconds?’” Paytm’s innovations included pre-populating recharge form fields with previously-used payment information, automatically retrying when transactions failed, and notifying users when their packages were on the verge of expiring.
Today, Paytm is facilitating Uber fares and gas and water bill payments, operating an e-commerce marketplace and replacing the tills of bricks-and-mortar merchants with scannable QR codes. The adoption of such digital platforms has been fuelled in part by a government push to bring more of the population into the organized financial system. Shortly after taking office in 2014, India’s Prime Minister, Narendra Modi, kicked off a drive to sign up 75 million poor households for their first bank accounts. New regulations have allowed the creation of digital banks, which can take deposits and sell insurance but not issue loans; Paytm’s launched this month. And in November last year, the government removed the widely-used ₹500 and ₹1,000 notes from circulation. The chaotic “demonetization” process led to a spike in signups, with Paytm adding 20 million new users through January for a total of 177 million, according to Bloomberg.
While the company has a sizeable workforce in India—CEO Vijay Shekhar Sharma has set a target of 20,000 employees by the end of the year—including engineering operations, the Toronto team has produced a number of back-end systems and front-end features that are core to Paytm’s technology. With a game thumping in the background, Takhar explained why Toronto was the right place to put Paytm Labs, what his crew are building and how the company will continue to innovate in the wide-open Indian financial services space.
Why did you start Paytm Labs in Toronto?
In 2014, Paytm had grown to a lot of merchants, and users were using us primarily for bill payments. We wanted to get a new set of merchants that were based in the U.S.—[companies] who wanted to sell to people in India. The examples would be, say, Uber. You want to launch Uber in India, but users cannot save credit cards—there’s a regulatory limitation. So Paytm Wallet works very well, because we can actually do a one-click disbursement. These kinds of deals meant that somebody should be seated over here to work on those partnerships.
Big data was quite a theme of 2014. Everyone was talking about how it could help one way or the other, and we wanted to explore, ‘What does it mean for us as a company?’ In Toronto [there are] lots of smart people who have a very good understanding of how things have to happen at scale, and the banking sector and technology sector come together in one city. Think of banking financial centre points in the world, like London or New York—Toronto is definitely in the top three. Because of that, a lot of these things were already solved by people here. So getting people who had done this problem-solution before was a breeze, compared to say Delhi or any other city. As we were growing, we knew we had to go where the talent was. So Toronto became, ‘Oh, yeah, why not Toronto in fact?’ Don’t you know that so much of effort in blockchain around the world happens in London and Toronto? And every bank over here has a big data team—why not us?’
What does Paytm Labs do?
‘Do something with data,’ that’s how we started. In the early days we thought we’d do reporting: We’ll inform every business of what’s [going on]—iOS is doing less transactions than the Android version, and so on. But very early on we figured out that we had to start at a very different scale than this incremental thing. So we [built] an internal software tool called ‘Datalake.’ We said, ‘Every data piece across the company needs to be in one place, so that we can analyze it.’ We spent a few months getting all sources of data across the company in one place.
The very first thing that we started working on was fraud prevention. Around that time, we were already doing a couple of million transactions a day. It’s a needle-in-a-haystack problem: Out of those two million, definitely 20,000 are bad transactions—but how do you find those? So we started looking for anomalies. We built an internal tool called ‘Maquette,’ which finds out for every transaction whether it is good or bad. Think of it as a credit score against a transaction, designed in real time. The next thing that we picked up was recommendations. ‘Midgar’ was our personalization effort. When you go shop on Paytm, [it recommends] this watch or whatever, these are what I would be likely to purchase. That requires a lot of computation and lot of understanding of who you are, what you’ve bought before, what you’re likely to buy in this session and so on.
How much of your job is head office saying, ‘We want to implement this feature, figure it out,’ and how much is your team saying, ‘This is what the data says there’s an opportunity to do’?
I think I can fairly say of the team over here that about half of the problems have been, ‘I understand the Paytm business, I think this should be solved. Let’s go ahead and solve it.’ We have an internal score for every buyer—we call it ‘Karma Score’ internally. It’s sort of our credit score equivalent, because India doesn’t have a very strong credit score system. We built that over here—and not because somebody asked us to. It can be used for fraud, and for arbitration in the customer care team. Let’s say there is a dispute; to know the scores of both sides can actually help the customer care agent do a much better arbitration than guess-based arbitration.
It’s challenged my definition of what it is to be an engineer. We are solution people now. But it’s taken a long time for this office to be empathized with [the Indian operation]. It is not organic. Think of the average person living here. How does someone who hasn’t lived in India understand the reality of 500 million people moving into the mainstream economy, when they can’t even log into the app because they don’t have an Indian sim card? So we make a lot of effort to integrate them. We make visits—10 people from here will go there and spend two weeks, not particularly for meetings but just to assimilate. Out of that empathy, these things sometimes happen.
How do you attract people to work here if they’re not likely to encounter the product organically?
It’s not an ‘Indian company’ problem. It’s, how do you sell your story to somebody who doesn’t know about it? It’s not different than any other startup or big company. Our focus here, including in the interview process and onboarding is, ‘What is the problem we are solving?’ There is an inefficiency in the marketplace, cash is actually a very bad way [to conduct transactions], and people pay for it in many different ways. How do we make it better? For example, because of Karma Score, so many people who could not get a loan can now get a loan. Or if the fraudsters are removed from the system, it can actually grow, and the good guys don’t lose out because of one bad guy. It’s actually a very worthy goal.
Let’s say you want to do fraud prevention in Canada. There is a credit score, which you have to work around—half of the answers are already there. So the road is there and you can only change the vehicle. But [in the case of India], you actually get to design where the road should go—or should there even be a road? When you get to design from a blank canvas, it is a ten-times more interesting problem. It’s not a problem that you get at Google and Facebook, because maturity always leads to tighter boundaries. But for us, these things have never been solved in India, so we can have the audacity to say, ‘Let’s build our own score.’
There are a lot of copycat companies in the Indian market, which makes sense: A lot of global tech companies weren’t operating there, so people were just replicating those products for a domestic audience. But given the breadth of things Paytm does, there’s not necessarily a model for you to learn from. Without that roadmap, where do you go next?
‘What is the competitor doing?’ is practically irrelevant in our discussions. If we were doing that, we would not have done payment via QR codes. In the first few years, we had a very simple way to find out what to do next: Look at customer care complaints, identify the top three, and that is your roadmap. Just think about that for a bit. Customers are telling you, ‘This is not working,’ so you solve it. You do this ad infinitum, and you’ll get somewhere.
But you have to be very creative about how exactly you solve the problem. For example, I would call the Paytm Cash Wallet one of our most pivotal inflection-point innovations. In the early days, we found that when people did a mobile prepaid recharge or a bill payment, some transactions failed. For the minor percentage of transactions which it failed, we would do a chargeback, but that does not reach the user for up to 21 days. So people who had that problem said, ‘You didn’t do the transaction, at least give me my money back.’ And we said, ‘We’ve sent the request, and we don’t know what is going to happen.’ It was a black box for us. Nobody said, ‘Please make a wallet.’ But our solution on that day was: Let’s park it in a wallet, and we’ll give the user the option to retry [the transaction] or get the money back in their bank account.
FFB is a not-for-profit platform that engages different types of financial institutions of the country; assist them in their journey to reach next level