PODCAST TRANSCRIPTION SESSION NO. 206  –  RISHI KHOSLA

Welcome to the Lend Academy Podcast, Episode No. 206. This is your host, Peter Renton, Founder of Lend Academy and Co-Founder of the LendIt Fintech Conference.

(music)

Today’s show is sponsored by LendIt Fintech Europe 2019, Europe’s leading event for innovation and financial services. It’s coming up on the 26th and 27th of September in London at the Business Design Centre. We’ve recently opened registration as well as speaker applications. You can find out more by going to lendit.com/europe.

Peter Renton: Today on the show, I’m delighted to welcome Rishi Khosla, he is the CEO and Co-Founder of OakNorth. Now OakNorth only started in 2015, so they’re just over four years old, but already they are the most valuable fintech company in the UK, they raised the largest ever funding round for a European fintech and they’ve been able to do this, they’ve been able to grow very rapidly, at the same time, in a profitable way.

They’re one of the few fintech companies that actually make a profit, but having said that, they are also expanding aggressively internationally and we talk about that in some depth in the show. They’re putting a lot of effort into the US market, we talk about the differences between the US and the UK, how they’re growing their US presence, how they are approaching their international expansion in a very different way than the way they approached the growth in the UK. It was a fascinating interview, I hope you enjoy the show.

Welcome to the podcast, Rishi!

Rishi Kosla: Thank you.

Peter: Okay, so I like to get these things started by giving the listeners a little bit of background about yourself. I know OakNorth is not your first rodeo so why don’t you tell us a little bit about what you did before OakNorth.

Rishi: Sure. In 2002, my business partner and myself set up a company called Copal, it’s basically a financial research company, we set it up with a view of servicing major financial institutions, initially the large investment banks, then private equity funds, asset managers, etc. We actually made good on that vision so fast forward 12 years on we were a 3,000 person organization which we built using a total of $50,000 in terms of capital.

It was very capital efficient in terms of how we got there, our client base included pretty much all the top ten global investment banks, a number of major private equity firms and asset managers and then we sold that business to Moody’s Corporation and 3,000 people were distributed around I think it was 12 offices in [inaudible] countries.

Peter: Okay, so then you sold that. What was the impetus to start OakNorth, what did you see that really led you to attack this which is something that really hadn’t been done before in a digital way. So what was the opportunity you saw?

Rishi: Sure, so the real opportunity…the seed of the idea came through a particular experience that we had. Back in 2005, when we were building Copal, we went to a commercial bank for a line of credit because, as I said, we sort of used a total of $50,000 to build that business so we were balancing pretty finely from a working capital perspective so we went to a commercial bank. They told us very politely where to go and we went to another commercial bank and we had exactly the same experience.

A few months later, we went to an institutional desk at one of the large banks and that institutional desk gave us a credit facility which was for a hundred times the amount we were asking for from the commercial bank (Peter laughs) and they had no issue with it. And we were like, wow, this just shows something is broken somewhere and you could argue that the institutional desk was a little bit too aggressive, but the answer was surely somewhere in between rather than the answer the commercial banks gave us, ie of zero.

So that sort of caused the seed of the idea and as we played out the business of Copal and we were coming out towards exiting it, we kept on coming back to that particular issue we had and saying, you know, is this a structural problem, is it something which applies across markets, is it cycle related, etc. etc.

We became very convicted that, actually, there was a very strong opportunity because this is an underserved segment. It’s been underserved for very specific reasons and therefore, actually a very strong opportunity to build a business within and also building a business which fundamentally has a lot of net positives for communities within the markets that we would be playing in.

Because fundamentally, we’ll be helping put capital into the hands of companies which are growth companies, small, mid-sized businesses which tend to be the drivers of GDP and employment growth in many countries. So we felt like both from the perspective of solving a real problem, solving a problem which is actually a good problem to solve for countries and communities and also clearly building a good business, it sort of ticked the right boxes.

Peter: Before you launched, I mean, obviously you would have done some market research so you said you went out looking in like 2005, did you find out ten years later that it hadn’t changed at all? I’m just curious about the target market you’re after, these fast growth companies, they’re not really being served by the Funding Circles’ of the world because they need more capital than Funding Circle’s max, so what are their options today?

Rishi: Their options are to go into a commercial bank and get a very standardized loan package which is fundamentally based on backward looking data from the borrowers, from the value of the data you provide and backward looking analysis is great if you’re not a growth company because, you know, your next 12 months looks pretty similar to your last 12 months.

But if you’re a growth company, your next 12 months may look nothing like your previous and therefore, actually having a lending institution which is only going to look at your historics so it does not get you very far in terms of meeting your needs so it gets you with a very sub-optimal loan package.

Peter: Okay, so what are we talking about here? Can you tell us the range of loan sizes and the terms, the interest rates sort of thing you’re talking about and what’s the median?

Rishi: Sure I mean, so in terms of the range of loan sizes we’re talking about roughly from half a million up to…depends on which market, let’s say plus/minus $30 million. In the UK it would be plus that, in the US, it would probably be just a little minus that, in some Asian countries it may only be up to $5/10 million, in terms of where the unmet need is. Average term, roughly three years, average loan size $5/6 million, those type of numbers so within that type of range.

Peter: Okay, and what’s the typical interest rate?

Rishi: The typical interest rate is appropriate for the risk which we’re taking on.

Peter: Is there a range, obviously you’re not approving everybody, is there a range that you work within?

Rishi: Sure, I mean the range is anywhere from 4% to 8%, that type of range.

Peter: Okay, okay. And so I know that you’re not just doing loans, obviously, that’s kind of how you’ve built your business primarily to date, but I know you’ve really got two divisions of the company. Maybe my first question is, with this sort of software-as-a-service or banking-as-a-service type of product that you have, was that the intention…first, describe the product and was that the intention always from the get-go to have these two divisions?

Rishi: Sure, so the first thing I’d say is we don’t actually look at it as two divisions. What we were very clear on is that to provide this type of customized loan packages that we felt were needed within this market segment, you needed to be able to do much more detailed analysis on the particular companies that you were looking to lend to and that detailed analysis looks much more like what you would do on a large corporate loan, i.e. if you were doing a loan of a few hundred million.

And we knew how to do that pretty well from our last business because that’s the type of stuff we did day in, day out and we were clear that we needed to bring technology into the equation to be able to apply that same methodology down to smaller ticket sizes. So we actually started by actually saying, we need to build a platform which will enable us to take that corporate approach to be able to provide customized bespoke lending to smaller businesses and mid-sized businesses.

That’s where we started and we were very clear that we needed to prove the market opportunity as well as the functionality of that platform in one market and therefore, the bank is very much…in the UK, where we started actually building that proof and obviously have built a pretty strong loan book here in the UK on the back of it. But as a core theme, it is all about providing customized lending to small and mid-sized companies and globally, we’re doing that through obviously providing our platform to existing banks rather than creating our own bank in other markets.

Peter: Right.

Rishi: So we’ll continue to obviously build our own bank in the UK and continue to use that to test and evolve our own platform that fundamentally, as I like to say, we think of this as a horizontal play around solving this problem on a global basis, rather than a vertical play of building a bank in one country.

Peter: Right, right, got it. Okay, so then I want to just dig in a little bit into the technology you use because you’re making these decisions much more rapidly, I believe, obviously than a traditional bank so how are you using this technology? What’s sort of involved in the underwriting process and what have you automated? How does it work?

Rishi: Sure, here’s the fun part. So not only are we doing things quicker, but we’re doing things quicker on the basis of actually having done significantly more work so if you go back to what I said, most banks in this space actually take a backward looking view based purely on borrower data. Whereas, what we’re doing is we’re taking the borrower data, we’re obviously looking at historics, but then we’re using lots of different data sources to be able to create a forward looking view for that particular business. And that’s something which other lending institutions in this part of the market, i.e. small and medium-sized businesses, tend not to do.

So we’re doing significantly more work in much less time, and ultimately, that’s enabled through the platform. So what does the platform do? The platform effectively helps pull in both the borrowers’ data, but also locks up said third party data sources and when I say lock, on any particular case, it may be 5,10,15, 20, but if I look across the platform, it’s thousands of data sources which the platform will go out to across different cases.

So you pull in all of that data and then you use that data, fundamentally, to create benchmarks on a forward looking view so you can have conviction that your forward looking view has been stressed appropriately to be able then to formulate a loan package which is both appropriate for the borrower and more suitable for their business, but at the same time also clearly meets the risk appetite of the lending institution.

Peter: Right.

Rishi: So that at its core is what the platform does.

Peter: So these companies that are coming for a loan, obviously they have their own forward looking projections and I am sure you look at those, but from what I hear you saying, you’re saying you create your own based on the data that you’ve seen so obviously you’re not going to rely just on what they say and you’re creating your own forward looking projections, is that how it works?

Rishi: Exactly, so we take obviously what the borrower says, but then we formulate our own perspective.

Peter: Right, right. Okay, so back to the bank for a second, I mean, you’ve been famously quoted saying you’ve experienced zero defaults to date. The unfortunate thing about being famously quoted for that is there’s only one way to go from there (laughs) because you can’t do better than zero.

Rishi:  I one hundred percent agree (Peter laughs), and let’s be clear that there is only one way to go and it’s inevitable that you’ll go there, right?

Peter: Right.

Rishi: So I think what we’re very clear on is that we will perform, both in the current market environment but also in a downturn, at a level which is significantly better than the market in the segment that we play in. So we clearly, as a lending institution, at some point you’re going to have defaults, but we have a very strong conviction that the default rate that we will have and therefore, the loss ratio that we will have will be significantly below market.

Peter: Right, that’s fair enough, that’s fair enough. So one more question about the bank before we get on to your international expansion. Are you funding all of these loans through deposits in the UK, I mean, I know you’re offering some of the best deposit rates. You’ve got a deal with Monzo now I read about, is that how you’re funding the loan book?

Rishi: In the UK, that’s exactly right.

Peter: Right. Okay, so then let’s move along to…I know you’ve opened up a US office recently and obviously most of the listeners of this podcast are based in the US, although we do have a substantial international following, so you’re opening up in the US and I am curious about the…I get the problem is very, very similar.

I’ve heard that this segment of the market is very much underserved in this country as well, but isn’t the approach going to be very…I mean, every country is going to be different, I mean, the data sources are going to be different, I mean certainly the software is different, the accounting software or whatever, so how do you approach that with taking what the core of what you’ve built and applying it into a country that has…it’s got different infrastructure.

Rishi: So the first thing to note is that we’re already live in I think about eight different countries at least, so the challenge of being live in a new country with different data sources, etc. isn’t actually a big challenge for us. Some of the countries which we’re live in are significantly tougher countries than the US to get data. The US is relatively data rich and, I have to say, data quality is relatively strong.

Peter: Right.

Rishi: So the US, for us, is absolutely a core market that we’re looking to delve into. We’ve already started originating in the US as well as actually licensing our platform so we have clients who are live in the US on our platform today, we’re also originating loans for those clients so the client can choose just to take a platform, or they can take the platform and loans as well from us so we’re not putting those loans on our own balance sheet.

The amount of demand in the US for the type of lending we’re doing, I would say if anything, is stronger than even in the UK and I say that fundamentally because the US is a larger market and we’re just seeing larger volumes in a very short period of time than what we had seen on a comparative basis in the UK. So the transportability of the model is very high and actually our ability to execute it as I said, is much simpler in the US than some of the other markets where you have both language, data quality and data availability issues.

Peter: Right, right. So are you just focused on the US banks, is that the market you’re going after for this solution?

Rishi: So we’re licensing the platform to US banks, but we’re also offering the platform and loans to other institutional investors outside banks. So we’re both, fundamentally, appealing to banks and non-banks.

Peter: Okay, interesting. And so I haven’t heard an announcement so obviously there is nothing public yet, as far as who have signed up for your platform, is that correct?

Rishi: That is correct.

Peter: You can make an announcement right now, if you want. (laughs)

Rishi: I’m sure I can. (both laugh)

Peter: Okay.

Rishi: Thank you for that.

Peter: So then are you like …when you look at the US market, I mean, do you see this as…obviously the market is much bigger than the UK, we know that, but where are these people being served in the US right now? Is it still the same, are they going along to banks right now and having the same experience that a similar company in the UK would have?

Rishi: Exactly right. Regional banks and having the same experience.

Peter: Right, right, okay. So then the platform, the website anyway is oaknorth.ai and ai, it doesn’t stand for artificial intelligence from what I gather it stands for analytical intelligence. Is that the platform that you’re going out globally with for this offering?

Rishi: Exactly right, the reason is very specifically that we call it analytical intelligence, but clearly it is a play on artificial intelligence which obviously is already embedded within the platform, but fundamentally, it’s an analytical platform so that’s why…the origins of the name. Actually, we used an algorithm to come up with the name (Peter laughs) just for the fun of it.

Peter: Okay.

Rishi: So, yeah that is absolutely the platform which we’re referring to.

Peter: Okay, so what geographies are you live in today and what geographies are you expanding into?

Rishi: So in terms of where we’re live, obviously the US, obviously the UK, Germany, Austria, the Netherlands, Italy, Singapore, Taiwan, China; that’s where we’re live. And where are we expanding to, I mean, you know, I think that’s already a reasonable amount of breadth, I think what we’re looking to do is gain increased depth within each market. Yeah, I think we’ve got, again, our estimate of total loan stock in this space is around $7 trillion and we think the market is under penetrated so, you know, I think we’ve got a reasonable way to go.

Peter: Right, right, for sure. And the fact is there really is no global brand doing this right now. I mean, it feels like…

Rishi: Except for one.

Peter: (laughs) Right, right, there is one up and coming global brand, yes indeed. So your intention then is not to get a banking license in any of these countries, is that correct, and why not do that, why not get a banking license?

Rishi: That’s correct. Why not, because we fundamentally believe that working with existing lenders in existing markets is a much more scalable way for us to actually take our platform and propagate it across markets, rather than trying to build our own regulatory infrastructure in each of these countries and then dealing with, you know, the challenges of running multiple regulated entities.

And clearly, from a capital perspective, the amount of capital it would take to run a bank in each of these different markets would require us maybe tens of billions and, therefore, it’s just not the most efficient way to either bring in capital, use capital, also not the most efficient way to [inaudible].

Peter: Right, right. So this sort of SME growth business segment is obviously huge, but I noticed recently that you’ve also expanded your product offering in the UK to mortgages for the entrepreneurs, the business owners of these companies. So tell us a little bit about that product, was that driven by your customers, or why launch that product?

Rishi: If you look at what we’ve done within the core, it is very much about how do you create a tech platform to be able to provide customized lending into a segment which isn’t used to receiving customized lending.

Now when you look at what we’re doing in the mortgages space, we’re not doing standard mortgages. We’re doing complex mortgages where we need to be able to look beyond the types of basic parameters that a standard clearing bank would look at. Therefore, the approach which we take has (inaudible) and the underwriting is actually very similar for these types of assets as well as we do in our core.

Now clearly the end customer, instead of being a business in this scenario is a company, but it’s clearly one in the same in the way that a lot of these clients are going to be business owners and therefore, we’re servicing them on their personal requirements, rather than on their commercial requirements. So we felt like it very much leverages the core of our approach and our platform, number one.

Number two, although it’s not technically the same client segment, it’s technically the same client.

Peter: Right.

Rishi: Right, so therefore, it’s actually very much in keeping with the mission that we have.

Peter: So is that something that you intend to roll out across the other markets eventually?

Rishi: Exactly right.

Peter: Okay, so then like I’m just curious then, you’re live in the US and how many people do you have in your US office now?

Rishi: Plus/minus 20.

Peter: Okay, do you think that’s going to be your…what’s the plan for growing that? I was in your office just last month in London and it’s quite a sizable workforce there. Are you intending for the US to grow into that kind of size?

Rishi: Yeah, very much so, very much so.

Peter: Okay. So then talking about the funding round, there was a lot of noise or a lot of a…not noise, but publicity, shall we say, around the funding round that you did. It was the largest ever funding round for a European fintech, $440 million from SoftBank. I’m curious…obviously SoftBank, they’re relatively active in the fintech space, but how are they helping you, or are they helping you beyond just providing capital?

Rishi: Sure, so a couple of things. First of all, you’re right about the $440 million round, SoftBank was the primary investor, but Clermont Group was the secondary investor so it was a combination of the two of them.

To answer your question on SoftBank, SoftBank are helping us and actually I have to say that they’re proving to be a very strong partner for us and they’re helping us in a multitude of ways. One is removing the constraints from your thinking about capital so effectively saying capital isn’t a constraint; how fast can you go, how large can you go whilst still obviously maintaining some level of prudence one needs to to make sure we don’t go off the rails, right, that’s number one.

Number two, they are directly helping in terms of bringing in talent and identifying actually where we may need talent as we scale which is proving very useful.

The third aspect is, you know, they know a few banks who could be clients and they’re being helpful in that regard so fundamentally, strategically, channel acquisition and actually from a client perspective, on all areas they are being helpful and therefore for us are proving to be a very strong partner.

Peter: Right, right. We’re almost out of time, but I want to get a couple of more questions in. You mentioned China and…China is notoriously difficult to crack. I spend a lot of time in China personally with our event, we have a conference in Shanghai. How are you approaching that market, I mean, do you have like….are you partnering with a local business, are you setting up from scratch? How are you attacking the Chinese market?

Rishi: So the first thing to note is in our previous business we did operate in China, we had about 300 people in Beijing.

Peter: Okay.

Rishi: So we’ve operated there and that was an independent operation, i.e. standalone, and therefore operating in China isn’t new to us. We have a small team in China today and we intend to, again, build that, not necessarily at a very high growth rate today, but build it and put all the right foundations in place.

The market which we’re obviously looking at scaling quickly is the US market and what we want to do is take our time in creating the right foundations in China because we do understand how much tougher it is to actually build and penetrate so that’s our approach to China.

Peter:  Right, okay. So you know, I’m thinking as you’re talking, you’ve got a lot on your plate. Like this is the last question then, so tell us how you spend your time? Do you spend half your time focused on the US market, are you focusing on building the team, maintaining the culture; how do you spend your time?

Rishi: All of the above so obviously the US is a key focus. I am the CEO of the regulated bank in the UK so clearly, that continues to be my main focus, but bringing in the right talent across the business in the right places is fundamentally the only way that one can scale. Therefore, that’s inevitably a meaningful focus for us and then, you know, there’s 24 hours in a day and seven days in a week so you can actually stretch the week pretty effectively.

Peter: Okay, well on that note, we’ll have to leave it there, Rishi, I wish you all the best. You’ve built a fantastic business in a short amount of time and I can see that you’re only just getting started. Thanks for coming on the show.

Rishi: Thank you so much, appreciate it.

Peter: Okay, see you.

Rishi: Bye.

Peter: This whole banking-as-a-service segment is truly fascinating to me. It’s one that is really still in it’s infancy. While we’ve seen some companies in the consumer lending space, even in the small business space move beyond their own offering and offering it to other companies in the space, we’ve never seen anyone like OakNorth that are just going out across the globe fairly aggressively, taking their core technology, not just expanding it to one other country, but to multiple other countries at the same time.

If they can pull this off and they certainly have the resources to do it, they’re extremely well funded so if they can execute this, they will be the dominant brand in this space and it’s going to be fascinating to see how all this plays out.

Rishi is actually going to be a keynote speaker at our LendIt Fintech Europe event in September so you can find out more and listen to him at that event.

Anyway on that note, I will sign off. I very much appreciate you listening and I’ll catch you next time. Bye.

Today’s show was sponsored by LendIt Fintech Europe 2019, Europe’s leading event for innovation and financial services. It’s happening September 26th and 27th at the Business Design Centre in London. Registration is now open as well as speaker applications. Find out more by going to lendit.com/europe.

PODCAST TRANSCRIPTION SESSION NO. 205  –  ALEX LINTNER

Welcome to the Lend Academy Podcast, Episode No. 205. This is your host, Peter Renton, Founder of Lend Academy and Co-Founder of the LendIt Fintech Conference.

(music)

Today’s show is sponsored by LendIt Fintech Europe 2019, Europe’s leading event for innovation and financial services. It’s coming up on the 26th and 27th of September in London at the Business Design Centre. We’ve recently opened registration as well as speaker applications. You can find out more by going to lendit.com/europe.

Peter Renton: Today on the show, I am delighted to welcome Alex Lintner, he is the Group President at Experian. Now I wanted to get Alex on the show because Experian has released a number of new products in recent times and I wanted to sort of delve into some depth into these.

We talk about the new Experian Boost product, we talked about their acquisition of Clarity Services, we talk about UltraFICO and how Experian is working to really include more and more people in the credit system and their approach to alternative data and machine learning and how they’re really leveraging everything they can to bring more consumers so you can accurately score as many people in this country as possible and globally, for that matter. It was a fascinating interview, I hope you enjoy the show.

Welcome to the podcast, Alex!

Alex Lintner: It’s my pleasure, Peter, thank you for having me.

Peter: Okay, so I like to get these things started by giving the listeners a little bit of background…I know you’re not from this country, same as me, so maybe you can tell us a little bit about how to came to this country and what you’ve done in your career before you got to Experian.

Alex: Yeah, I’m German born and raised and I got to this country by having the privilege of doing my undergrad and graduate schooling here. After that, I started…I did go back to Germany, but started working for US corporations. I worked for a consultancy that was then purchased by Ernst & Young and then migrated eventually to BCG, the Boston Consulting Group who transferred me to their San Francisco office and as is not untypical of consultants, one of my clients eventually decided to hire me and that was a company called Intuit in Silicon Valley.

After my career as a consultant, I had nine very fruitful years working with Intuit and being part of a journey that I still consider one of the most exciting ones that I’ve had in my professional career.

Peter: Okay, okay. So then what was it that led you to Experian, what was sort of the opportunity, and why did you join the company?

Alex: Well if you think about what Intuit does, they are the makers of QuickBooks, which is a small business accounting software suite and the makers of TurboTax which helps individuals file their income taxes. So I really was in a fintech company, one of the largest, you know, if you will, and Experian called me proactively and really said, well, we’re looking for somebody who sort of understands the sector and has a strong background in technology.

I benefited sort of from the halo effect that during the time while I was at Intuit, we went through a technology transition that I played an active role in and that experience was attractive to Experian who was seeking to themselves evolve and become a leader in the space of delivering data with the most secure and tech-forward technology available. So I now play the role of leading part of that effort and here I am being part of America’s largest credit bureau.

Peter: Okay, so then what exactly are your responsibilities at Experian, what are you focused on?

Alex: I am Group President for Consumer Information Services here, I am operationally responsible for the credit bureau here in the US and recently, also responsible for all of our credit bureaus around the world in terms of knowledge sharing, best practice sharing etc., etc.. So operationally, you should look at me as, you know, the business lead for everything that the US Experian Credit Bureau does.

Peter: Okay, okay. So one of the things that Experian Credit Bureau did recently was, you know, acquire Clarity Services. Actually, I heard you say that it has been a very successful acquisition and obviously, we’ve heard you say many times and others at Experian are interested in expanding the credit box. We’ll talk about Experian Boost in a little bit, but firstly, I want to talk about that acquisition and what have you gained from the acquisition of Clarity Services.

Alex: Yeah, the acquisition of Clarity allowed us to bring credit prediction to consumers who were viewed as unscorable by sort of the traditional credit bureau operations so us and our competition. Clarity Services has files on 62 million people who are thin file or no file and on people who are, you know, near prime and prime and super prime, but who engage in short term lending.

It’s interesting, in the US people think about short term lending as sort of being a type of lending that’s definitely not part of mainstream lending. I would agree with that, but the myth is that they are at the fringe of the population, they live paycheck to paycheck or something like that. That is actually a misnomer, that is not correct.

To give you an idea, one in five Americans gets a hit on the Clarity data, or to tell you the same story in a different way, if you look at the past four years, 100 of the 247 million adult Americans, so people 18 years or older, have actually, at some point, taken out a short term loan of one type or the other, the loan type that’s traditionally not covered by the credit bureaus, but when you get to 100 million people that’s active there, we believe, with our mission of being the consumers’ bureau covering everybody, so we acquired Clarity in order to capture the activity in that type of lending.

Peter: Right, right. So I’m curious…I mean, I don’t know if you know the answer to this, but is there….what was the overlap between Experian’s database, because obviously there were 62 million people at Clarity, some of those would have been already scored by you, but…was there very little overlap? Can you tell us some sense of that?

Alex: Yeah I do have the answer to that question, there was definitely overlap. So, the way you should think about it is the 62 million people who are captured by Clarity, of the 100 million total engaged in short term lending in the last four years, about 60% of them are overlapping with a traditional file. So they were…we already knew the name and there was at least one tradeline.

Peter: Okay.

Alex: …most of those would be thin file. The other 40% are names where we obviously knew the name, the address, date of birth, the kind of stuff that we get from public records, there was no tradeline and Clarity sort of provides the only tradeline at the time. That’s how you should look at it.

Peter: Right, right, okay, got it. Another initiative, I know you talked about this at LendIt earlier this year, Experian Boost is another kind of initiative you to have to really help bring more people into the credit system. Can you just explain what it is and how it works?

Alex: It’d be my pleasure, Peter. Both of these initiatives that you are asking about are basically under the umbrella of, you know, our alternative data initiative. Let me say a few things about that before I answer your immediate question.

The reason why we’re passionate about and determined that we are going to be the lead in alternative data is because we have learned that lending history, which is the traditional data  that credit bureaus use, is certainly a proven method to predict a consumer’s/an applicant’s likely ability when it comes to taking out a certain type of loan. But sometimes there isn’t enough data for all people.

In order to address people who may not have a lending history or not sufficient lending history, let me give you some examples, so it could be the super wealthy, those could be people who just turned 18 years old, those could be immigrants, those could be [inaudible] where it was a single income household taking the spouse’s name who didn’t have the income.

It could be heirs who, you know, live off the wealth of somebody who had passed away but who in the past didn’t handle financial dealings so there are populations that could be very similar to me, Peter and not people living on the fringe for whom lending history data is not available. And then we turn to what we call in this country alternative data. By the way, around the world, we refer to alternative data as data…

Peter: Right (laughs).

Alex: …because very few people in emerging markets actually have a real lending history.

So if you look at the fastest growing part of the world, the Asia Pacific markets, 5% of people there have any lending history, 95% need to be scored in other ways. So we have asked ourselves, how can we add data about people to our file if they’re comfortable with us having that to allow us to correctly and accurately represent their credit reputation. Could they pay back the type of loan they’re applying for, what type of risk do they represent because we do stand for fair lending, we want to help the lender and we want to help the consumer.

So, Boost is the outcome of more than three years of effort that we had here at Experian to find a way to incorporate consumers’ bill payment history; bill payments about utilities, mobile phones, cable bills, etc. into their credit report. The most popular scores out there like FICO, that was always foreseen [inaudible] that those would be part…we just had trouble having the utility companies and the mobile phone companies and the cable companies report that data. They didn’t.

So we have opened the door for consumers by allowing them to contribute that data and virtually all consumers have utility payments, mobile phone payments, many of us have cable TV payments and it has not been factored into the credit score. We’re thrilled that we have found a way to do that with Experian Boost and it basically works by consumers giving us their online banking credentials. Through that, we get access to the bank’s last statement for that consumer and then we have what is called a categorization engine that identifies those type of payments and then they get incorporated into the existing scores.

This is, you know, risk-free because there is no impact to the consumer, if they so choose. So we show the consumer their current score, then we look if there is a tradeline that we could add to their file. If there is, we tell them what are the tradelines, we show them what the tradelines would do to their score and then they can choose whether to add the tradeline or not. And we are delighted to say that after, you know, what is now ten weeks or something like that…we have 700,000 people…

Peter: Wow!

Alex: …who have signed up for Boost, just to tell you from my own history, Mint which is an Intuit product and this is the most popular personal financial management software in the US with now 24 million users. It took Mint one year to get to a million users, we are ten weeks in, we’re at 700,000 so the velocity of adoption here is fantastic.

Peter: And so then what is a typical…you’ve got that 700,000, what is the average or typical increase in score that they’re seeing?

Alex: Yeah, so let me give you a couple of statistics. About 70% of all consumers do get a boost to their scores so their score increases. The average increase that people get is 13 points and coincidentally, 13% of the people who do get an increase jump risk bands.

Peter: Right.

Alex: So, you know, obviously this skews heavily to the no file and thin file. These consumers benefit the most because if you jump a risk band that makes you eligible for a certain type of loan, or if you were already eligible, it will qualify you for lower interest.

Peter: Right, right, okay, got it. So that’s super interesting. One thing I heard you say at LendIt as well is the consumer can opt out at anytime, right, so you opt in for this. Say someone decides for whatever reason they’re not comfortable anymore, they can just log in and opt out, is that how it works?

Alex: Absolutely, it is consumer permissioned and consumer permissioned means exactly what you have just said which is, you, the consumer, are in control. You can opt in and use it and if you don’t want to use it anymore, you do opt out.

You know, of course, we do encourage people to stay “boostable”, if you will, because if they keep Experian Boost activated, consumers can build a stronger payment history over time which would be to their benefit and we encourage it, but if they elect not to, they are 100% in control and can opt out at anytime.

Peter: Right, right. Okay, so I want to talk also about a new score that you are involved with and that’s UltraFICO. I want to be clear and we talked about this again at LendIt in depth, but how does that interface? What is UltraFICO, tell us that, and how is it different to Experian Boost?

Alex: Yeah, so up to now a positive behavior from consumers in terms of meeting their monthly financial obligations in relation to their income was hard for an underbanked consumer to demonstrate. The UltraFICO score is essentially a new score that incorporates transactional data from consumers’ checking, savings and money markets accounts, thus providing the opportunity for an underbanked consumer to access credit.

UltraFICO as opposed to Boost, which I talked about earlier, they will have attributes created around cash flow of the consumers whereas Boost incorporates the utility, mobile phone and cable bills. This does a cash flow analysis and then comes up with a totally new score. We estimate that that score will apply to about 15 million consumers that were previously unscorable which, you know, us and FICO together we hope that millions of households will have the opportunity to gain access to mainstream banking products which is, you know, a less expensive alternative financial product than the means that they turn to today and our collaborations with FICO and also with Finicity, who is the company that provides the account aggregation technology that’s required to make this possible.

Peter: Right, right. And so I know it’s pretty new in the game, do you have pilot lenders out there using UltraFICO today? Is there anything you can share on that?

Alex: No, UltraFICO is going to launch this summer so, Peter, we’re probably…I would say six weeks too early to give you data on that, but if we chat again in a couple of months then I’m happy to share.

Peter: Okay, fair enough. So something else, while we’re talking about scoring here…this was done a little while ago and you introduced the Clear Early Risk Score. We actually wrote about it on Lend Academy and I think that was about 18 months ago now, can you explain what that is and how does that fit in with other things like UltraFICO?

Alex: Right, so earlier we spoke about Clarity data, the Clear Early Risk Score builds Clarity data and it allows lenders to gain a previously unavailable view of a consumer loan and payment activity spanning both the mainstream and alternative financing spaces. Thereby it provides the most comprehensive consumer insight in the industry whilst providing financial access to more consumers, as I explained earlier.

Peter: Right.

Alex: The score has some unique analytics inside and applies those to leverage both Experian’s national credit bureau data and Clarity Services specialty data. It predicts the consumers’ credit worthiness over a 12-month period, it focuses in particular, at scoring risk of the first payment default and why that’s important, in short term lending, the loans are not installment loans, they are single payment loans so you think about a title loan or something like that.

You know, people borrow a certain amount of money and then agree to pay it back at a certain time; our own definition is shorter than 12 months usually so sometimes they pay it back next week, sometimes in two months, sometimes in a year, depends on the type of loan, and it’s a single payment, not an installment. This score predicts, if they were to take that type of loan, what would be the propensity to fail that payment as agreed upon. That’s the Clear Early Risk Score.

Peter: Okay, okay. So then the companies that are doing these single payment loans, they’re reporting the data into Clarity, is that how it works?

Alex: Yes, that is how it works.

Peter: Okay, okay, interesting. What about rental payments? We haven’t talked about that yet and obviously, you know a good chunk of the country doesn’t have a mortgage and they’re renting. To me, that’s something that would be also a highly predictive data point. What are you working on and how are you working with rental payment data?

Alex: Yeah, so a subject close to my heart exactly for the reason that you have just named…if you don’t have a mortgage, your rent is often your largest payment every month. We have an asset here at Experian that’s called RentBureau which we launched in 2010.

Just a little bit of context, nearly a third of all Americans are renters and Experian RentBureau has data on more than 20 million residents. That’s a great asset, to be frank, I’d like it to be even larger and it is a positive rental payment reported by all those landlords who are added to the Experian credit database.

It’s very helpful to 60% of consumers, those who are in the near prime segment, in particular. When they make those payments, it helps them qualify for a mortgage, or it helps them get to a new risk band. Rental payment we attempted to also get those through account aggregation so, you know, give us your online banking credentials and we can look and through a categorization find them, but we have learned some very interesting things that kept us from doing that.

What we’ve learned is essentially, somewhere around 45% of all rental units in the country are owned by institutional renters, large landlords who have sometimes millions, or sometimes only in parenthesis, tens of rental units. We would recognize those because they have a corporate name. The other 55% of all rental units are owned by people, people like you and me, so if a payment goes to Peter Renton or Alex Lintner, it is hard for us to recognize that it’s a rent payment.

Peter: Right.

Alex: Therefore, the categorization engine technology doesn’t work, like it would work when you pay to a company, a utility company, your mobile phone company, etc., etc. So we are still dependent on the reporting of renters and we work with our partners to grow that stake, our plan is to double it in the next three years and we’ll see whether we achieve that.

Peter: Okay. So I want to talk about privacy for a minute because, obviously, it’s a real concern with credit bureau data and even, I am sure, some of the customer contributed data that we’ve already talked about. What is your approach to privacy and how do you make your customers comfortable with that?

Alex: To state the obvious, I’m a consumer too…

Peter: Right (laughs)

Alex: …I am concerned about my own privacy and I respect that our clients and our end-users are concerned about their privacy, they should be and we want to help them with that concern. Having said that, our mindset is that for consumers contributing data, it’s not such a new thing. If you think about data that we contribute to social media, we contribute all kinds of data so why not make a contribution, something that helps boost your credit score.

That’s why we like account aggregation technology, the consumer permits us to do that, they can choose when they want to do it, they can choose when they don’t want to do it which is privacy. I choose what I do when I want to do it and I only show the people who I want to show. Hopefully we earn their trust, we do everything that we can to keep the data safe, we do everything that we can to make the experience, you know, frictionless so it’s easy for them to do that, and we do everything we can to present them with that opportunity at the time that’s most opportune for them.

For example, when they apply for a loan, or for example, when we know they have slipped credit bands and, you know, are working themselves out of a tough situation so Experian feels responsible for this data that was contributed. We will never charge for it, it’s under control of the consumer, available at all times and it’s free and we, nevertheless, but we don’t make money on it, we give it the same security treatment as we do with all our large data furnishers, the lenders in the country.

Peter: Sure, sure. So then you’re working with massive datasets at Experian, you’ve got billions of data points, millions of daily transactions and you know, it screams out for the use of machine learning and artificial intelligence because the datasets are so large, so maybe could you just comment on how you’re using machine learning in the development of some of these new scores?

Alex: Yeah, so we have an Extended View score here at Experian that we’ve had for a while that was generated using machine learning. You know, machine learning is one of those things that not many people understand so let me tell you a little anecdote that helps illustrate how we think about machine learning.

I have a parent of a very dear friend who was diagnosed with brain cancer, actually, and I asked her…I’m sorry to hear about your diagnosis, what are you going to do? Her answer was interesting to me, her answer was, I’m going to a doctor who is an oncologist, who uses machine learning. Why machine learning? She goes, well it’s very simple, this person is a mathematician, so she said, well, the average oncologist with over 10-years of experience gets the diagnosis right in 97.3%. The average oncologist with ten years experience who leverages machine learning gets the diagnosis right 99.8%. Alex, if you had brain cancer, what would you do?

Peter: Right, right.

Alex: Obviously we all know the answer. The point of the story is that the human thought leadership in putting together a score doesn’t get lost, but we have such vast amounts of data available today for a certain type of scenarios, for certain segments of the population, would there be other data elements that better describe the individual so that the individual gets reflected to the potential lender in a fair and accurate manner. That’s what we aim for.

If we think back about our own founding days, we were founded by a man called Si Ramo, the story that I tell about Si Ramo is very simple. He was a pharmacist in Nottingham, England who had a big heart, he gave pharmaceuticals to people who couldn’t pay for them immediately, he kept manual books, who did he give pharmaceuticals to so that they pay him back at some point.

He became the most popular pharmacist and keeping manual books didn’t do it anymore so he created a scoring system so he knew who he would get his money back, who to do that for so he could get his money back and who not to do that for, where he would likely not see his money. That evolved into the credit system.

Si Ramo initially knew people, he asked, well, who is your mom and dad, where do you work, and so he knew the local community understood the individual. What we do now with artificial intelligence…when we look across 147 million American adults we want to make sure that we, as Experian, represent the individual as who they are so that the lender understands the person sitting across from them is not some anonymous individual, it’s a real human being.

And that is what machine learning can help with so to give it more “gestalt”, I can use one German word (Peter laughs), more “gestalt” of who is sitting across from them and what should I take into consideration as I consider whether I give a loan and if so, at what interest rate which represents the risk the applicant represents.

Peter: Right, right. Okay, we’re running out of time, but I’ve got a couple more questions that I really want to get your thoughts on. First one is just the state of the American consumer because you obviously have great insight into that and, you know, we’ve been on this economic expansion for a long, long time and the consumer still seems to be doing pretty well, but there’s definitely been talk in the media over the last couple of months that maybe delinquencies are ticking up in certain segments so what’s your take on the American consumer today?

Alex: Peter, we just published our latest State of Credit Report, as we call it, it was released a little more than a week ago, and it takes a 10-year look back at consumer credit behaviors from Q2 2008 until Q2 2018. If you remember, Q2 2008 was the year when we headed into the financial crisis. If you take a look at this past decade, the report shows that over the past ten years we’ve seen a decrease in delinquencies.

My interpretation of that is we also see the average debt load for the consumer increasing, but delinquency is decreasing which means that I believe consumers are using credit more responsibly and that is reflected in their credit scores. The average score now is 680 on the Vantage scale and that’s a prime score so I think that’s good news for the American consumer.

Peter: Okay, so last question then. We’ve talked a lot about the thin file consumer and I’d like to get an idea of your goals there and what’s it going to take to get these tens of millions of people who are still unscorable, what’s it going to take to get that much closer to zero?

Alex: Yeah, we talked about it briefly when we touched on the topic of alternative data. Our goal is, you know, we regard ourselves as the consumers’ bureau, is to determine the credit worthiness of any consumer seeking access to credit. To move those tens of millions of people into mainstream credits will take innovation and alternative data is one of the ways which we can do that.

We spoke about Experian Boost where we get certain type of bills included and we spoke about UltraFICO which looks at the cash flow of the consumer, but we’re not done. Those are not the only things that we’re going to do. Improving access to credit is part of our fundamental mission and our innovation aims at finding more ways, more alternative data to continue to improve the financial health of consumers around the globe. That’s all of our charter and I think if we talk again in six months, in addition to Boost and UltraFICO, you’ll hear more news about how that can be accomplished.

Peter: Okay, it’s a great mission and I really appreciate you coming on the show today, Alex.

Alex: It’s my privilege, thank you very much for having me.

Peter: Okay, see you.

Alex: Take care.

Peter: One of the things that I’ve talked a lot about on the show over the years and particularly, in recent times, is just the importance of bringing more data to bear, bringing more consumers who are unscored or who have a thin file into the credit system where they can…the bottom line is they can save money. I’ve said before, it’s expensive to be poor and you may have not much money, but you’re always making your payments on time and you have a very low credit score, but you would be a good credit risk for some of these products.

What Experian has been working on, I think, is fantastic in bringing more data to bear to give a more accurate picture of those consumers in the thin file type population. I really hope that we can continue to expand that and it’s going to make a huge difference for those people who can suddenly obtain less expensive credit than what they have been used to.

Anyway on that note, I will sign off. I very much appreciate you listening and I’ll catch you next time. Bye.

Today’s show was sponsored by LendIt Fintech Europe 2019, Europe’s leading event for innovation and financial services. It’s happening September 26th and 27th at the Business Design Centre in London. Registration is now open as well as speaker applications. Find out more by going to lendit.com/europe.

PODCAST TRANSCRIPTION SESSION NO. 204  –  HUSAYN KASSAI

Welcome to the Lend Academy Podcast, Episode No. 204. This is your host, Peter Renton, Founder of Lend Academy and Co-Founder of the LendIt Fintech Conference.

(music)

Today’s show is sponsored by LendIt Fintech Europe 2019, Europe’s leading event for innovation and financial services. It’s coming up on the 26th and 27th of September in London at the Business Design Centre. We’ve recently opened registration as well as speaker applications. You can find out more by going to lendit.com/europe.

Peter Renton: Today on the show, I am delighted to welcome Husayn Kassai, he is the CEO and Co-Founder of Onfido. Now Onfido are a fascinating company, they have become almost the standard now in identity verification for the fintech space. Certainly in the UK and Europe, they are used extensively and becoming more so in the US, but anyway, I wanted to get Husayn on the show because they are really doing interesting work.

I wanted to talk about how their system works, what’s wrong with identity today, why the current system is so flawed, we talk about the challenge of user experience and reducing friction, we talk about how banks are using this, we talk about artificial intelligence, and how that is playing a major role in improving these models, and we talk about what it would take to actually have a perfect world with very little identity theft. It was a fascinating interview, I hope you enjoy the show.

Welcome to the podcast, Husayn!

Husayn Kassai: It’s great to be on.

Peter: Okay, so I like to get these things started by giving the listeners a little bit of background about yourself and…you know, you haven’t had a long and storied career because, as we were just chatting before we hit record, Onfido was your very first job out of college. So maybe just tell us a little bit about why you decided to do that instead of going and getting a job and what are some of the things you studied in college that you felt prepared yourself for this.

Husayn: Of course, so in my last year at university, I was fortunate enough to be the President of the Oxford Entrepreneurs Student Society and during that whole year, I got to see very many startups coming and pitching and presenting and asking for funding so I got to learn quite a bit during that process. We had a clear metric…my Co-Founder later also was the Vice-President of Oxford Entrepreneurs at the time and we had a metric around wanting to see how many successful spinouts we can have.

Despite us doing a lot of activities and trying very hard, towards the end of the year we still only had maybe one or two actual startups spin out and do it full-time. So we had learned so much and we genuinely felt that there is a big need to solve this identity problem that we felt that we should do it ourselves because the norm is you kind of have to conform to expectations, you know, you go and earn a salary, you have a to get a mortgage, you have to tick these boxes. We had experience and we saw that counter to what the expectations are, doing a startup is never necessarily, there’s never an ideal time for it. The ideal time is when you see a real problem and you feel like you can work to solve that so we felt it was the right time and that’s why we went straight into getting the company going.

Peter: So then what was the thing you saw that led you to the idea that became Onfido?

Husayn: So fundamentally, it’s about building trust and Onfido in Latin is a mix between fido which is “trust” and confido which is “confidence” and we saw that the way that identify verification process was carried out was essentially what we had deemed as broken.

We have our different stories as co-founders, but when I turned ten, my parents moved from Iran to the UK and I remember quite well that it took them a few months to be able to open a bank account in their own name surely because they weren’t registered on the credit bureau. Growing up, I found out that credit bureaus essentially had a log of everyone’s date of birth, name and address and it excludes half the world of the adult population who are underbanked, unbanked and are therefore not on the credit bureau so it has an exclusion and access problem.

On the other hand, you have a security problem whereby almost all of that data is already breached and in the dark web anyway so a bad actor can just steal your date of birth, name and address and open an account so you don’t get security and this kind of loss of currency. So given that it was excluding a lot of people and also not offering security. We felt sure there was a much better way as the world is moving online for businesses to be able to verify their customers in a user-friendly way, but also secure and to give access to as many people as possible. That was the initial mission which we’ve been working on now for seven years.

Peter: So tell us a little bit about those early days. I’d be curious to sort of get your take on did you have a particular product in mind when you launched, were you trying to sort of redo the entire process? What were the early days like?

Husayn: So when we looked at how are you able to prove someone’s legal identity, we first distinguished how we are going to focus on your legal identity. So we could see how Facebook has standardized the way everyone shares their social identity, your likes and preferences and so on and how LinkedIn has solved the problem of everyone who ever wants to with an account be able to share their professional identity, your academic career or job career and so on.

So our focus was on your legal identity, that is your date of birth, name and address so that you’re compliant if you want to carry out some financial transactions, sign up to a lending platform or similar and then at the same time if there is a breach and a bad actor is found, then law enforcement can be told precisely who they are so that was our focus around legal identity.

Our third co-founder, Ruhul, is our technical co-founder. His university thesis was using computer vision and machine learning to spot wildlife in 25,000 photos of the jungle so this pattern recognition research that he did, we felt very well mapped on to government ID checks. So when you want to open a bank account, for instance, you go in person and there is a bank clerk and an individual that asks to see your passport or driving license and deems it to potentially fraudulent or not and then compares your face to the photo on that.

With our technical co-founder’s research, we could see that we could bring that bit and make it digital and bring it online so not only is it more secure, but it’s actually a lot more convenient too. So that was the first version of the software and naturally, we’ve been improving it ever since.

Peter: Okay, interesting. So then what is it that is specifically wrong today, is it just merely the fact that your data is out there to be stolen, is it the fact that really we’re looking at, you know, a system that…it’s all about these legacy products like a driver’s license or a passport, is it merely the fact that these things are so easily stolen, is that the biggest problem?

Husayn: So there are a few and identity fraud, as you say, is a big one so in the US, for instance, it’s the largest crime and the fastest growing crime across all of fraud, specifically online frauds. The common denominator is identity theft because a fraudster does not want to get caught so the first thing that they do is impersonate someone else, or come up with a fictitious identity so that they’re able to commit a bad act.

It is a large and growing problem because 2 to 5% of the world GDP is laundered money, that’s between $800 billion and $2 trillion which are United Nations figures so it’s a real shame that less than 1% of that is actually caught because money laundering is used for human trafficking, drug trafficking, terrorist financing and so on so not only is identity theft problem a large one, it is getting worse. When you consider the reason why it’s so bad and getting worse is because the current systems in place to address it aren’t fit for purpose.

On the one hand, you have the option of going face-to-face in a bank branch and being verified, but as we know, that’s not accessible to everyone and most consumers now prefer to do things digitally and online and there is a big convenience factor that is the driver. But on the other end of the spectrum and the other approach is to verify people online and online verification, historically, has been predicated on using credit bureaus and that credit bureau model is just a centralized database, but because it’s leaked or because it’s been breached, that data is accessible to bad actors and hence, it has lost its currency.

So there are one or two other sort of smaller alternatives, using social media which has a privacy issue and sort of is not seen as the right thing, nor should it be, or using knowledge based questions such as, you know, what’s the third character in your last utility bill for instance, which again has a lot of friction because users aren’t necessarily accustomed nor used to storing or saving or looking up utility bills.

So our approach, which was from the outset focused on a user showing their government ID, either a photo or a recording of their face, was designed so that we could do a very good job of determining if it’s fraudulent or not because we run our machine learning models on whether that government ID is fake and then secondly, whether the person showing that ID is the owner of that by taking a photo of their face and comparing it with that. So that’s our problem, is identity theft, the current solutions are in our view, are not fit for purpose. Third, is that the world is moving online and businesses no longer see their customers face-to-face, and hence they need to have this in place.

Peter: Right.

Husayn: So we have four building blocks that we focus on which is fundamentally to help the 98% easily access services; second is to get very good at determining which 2% are the bad actors and they are two sides of the same coin; the better you become at detecting fakes, the more easily you can onboard that 98%. Third is to uphold privacy, and fourth is to make it as frictionless as possible so it is an easy and friendly user experience. That’s been our chief focus to solve this identity problem.

Peter: So then take us through a typical example of how your product is implemented. I know you’ve got several clients in the fintech space and the banking space, but let’s just go and take us through…someone is opening up a new account, let’s take LendInvest, for example. I listened to a podcast where you were on with I think it was Christian Faes of LendInvest so I’m an investor, I go on to LendInvest, I open an account and then what happens as far as the verification, what are you doing?

Husayn: So as part of the registration process at LendInvest, you either, on the portal or on the app, at some stage, probably stage two or three after you’ve entered the fact that you’re interested in taking out a product and so on…that’s the point at which our SDK kicks in. Essentially, it’s an experience where you are then asked to take a photo of your government ID, it can be a driver’s license, passport or similar and then once that is done the next stage is to take a photo or a short recording of yourself.

Once you click you’re ready then you hold the phone or it can be your webcam on your computer for instance and you’re asked to read three randomly generated numbers so for instance you would read 2,5,6 or whatever the numbers happen to be and you’re then asked to carry out a task. It can be turn your head to the left, turn your head to the right or something simple and then you click submit and that’s it.

We then take it from there and then the checks are processed and the results are sent via API to the team at LendInvest, but behind the scenes what’s happening is when we have a photo of your ID like your driving license, we have seen tens of millions so we know what patterns to expect, we know what fonts, characters and if it has been digitally tempered with and so on and so forth.

So the machine learning models can determine whether it seems fraudulent or not and secondly, the photo on your driving license is compared to the face that you took either a photo or a video of and we can sort of determine if you’re the same person or not so that LendInvest would know that, not only has a person been verified, part of their KYC or Know Your Customer checks, but the person presenting the details are who they claim to be and the actual owner of the identity.

Peter: Right, right. So then how confident are you…is it 99.9% confident that the person who owns the ID is the person, particularly if you’re doing a video, I imagine video, particularly if you ask them to do a task, I mean, I can’t imagine that that could be hacked very easily, so how confident are you that you’ve got the right answer whether this person is who they say they are?

Husayn: So in some ways we’re kind of like an anti-virus software where there will be sophisticated fakes that cheat the system. So what we’re able to do is two things; one is to say we are significantly better than manual, human teams doing it, either comparing templates or in person and secondly we’re better than any other digital offering.

The reason for that is because of the machine learning models whereby as an ID comes through if it’s fine then it gets processed, but if we suspect that it might be fraudulent..for example, a password image might be stained. Now when it’s stained we have fraud experts that double check and when the fraud expert double checks, they’ll be better able to determine is it stained because it was accidentally put into a washing machine or does this look like a sophisticated fake pattern or anomaly that we’ve seen in the past and therefore, it is scored accordingly.

So a client typically, let’s say LendInvest would get 90% coming back as fine and then about 5% would be classed as consider. So either it’s because it’s a black and white document or it’s expired or a number of different reasons and their team will double check and every business has its own threshold as to what they would or wouldn’t accept, but these are triggers for them to be able to double down and focus on the 5% to be able to automate the 95%. And then 5% are typically rejected, either the person has not been able to effectively upload or they’ve stopped the process or something similar.

Once businesses integrate us and they compare their metrics prior to them integrating us, they can see the fraud rates going down sometimes in a pretty substantial way. So Zoomcar, for instance, which is the largest peer-to-peer car rental service in India, their car theft halved once they integrated us. So some of those 5% that drop off, they can do so as a result of legitimate reasons, but some may be bad actors that is because they don’t want to expose themselves or because they know there’s no way to cheat the system then they essentially self select out.

Peter: Interesting, interesting. So then…I keep thinking about, you described that process and…how do you think about the friction? There really is friction between a good user experience, a seamless user experience and going through this identity verification because the person sitting at the computer…I know I’m a good guy, I shouldn’t have to go through all this rigmarole, particularly if you’re not all that adept at using your cellphone. How do you approach that?

Husayn: Most people tend to expect to have to go through some sort of compliance check, especially if they’re asking a financial service, particularly if it’s lending, so part of it is because it’s integrity in the system…if it’s very easy for people to come and borrow money, we’re doing very little checks then naturally, you might get the impression that it’s easy equally for bad actors and therefore, the platform can not sustain itself.

The best thing with the wave of fintechs that we’ve seen is fundamentally…I mean, lending has happened for hundreds if not thousands of years which as you know better than I for example, but more recently what the fintechs have been able to do very well is put the customer experience at the heart of everything that they do and as a result, they are able to operate digitally or fully online and as a result, we from the outside have had to adapt and focus and partner because we also were aiming to make this as easy and frictionless a process as possible.

And so when it comes to showing your ID and a photo of your face, the good fortune is that smartphones and other technologies have now in many ways normalized it, especially taking a sort of a selfie or a photo we often…more and more of us are now using it just to unlock our phone. So because it’s been normalized and because more and more consumers are seeing this repeatedly every time they’re using one of the large fintechs or rental car services or online communities they’re seeing us so it’s in many ways normalizing. Does that help answer the question?

Peter: Yeah, that makes sense. So how long does it take typically, what’s the average user time like they get into your system, they’ve got their driver’s license, they’ve taken a photo, a selfie, is this a 30-second process, a 90-second process? What’s the average kind of time it takes?

Husayn: So from our side, as technology goes, if there are no issues with the ID or the selfie match, then it is typically sub 20 seconds so less than 20 seconds. If there is something that needs to be double checked, the average is two minutes so that’s when a fraud expert for instance needs to double check something, the average return is two minutes.

Peter: So in those situations you’ve got a human standing by going through this in a manual process, is that correct?

Husayn: Correct, so hundreds of fraud experts when they need to double check something, but this is the time it takes on our side. You have to remember that, depending on the business, they have additional checks, additional processes…

Peter: Sure.

Husayn:  Sometimes it might take up to 24 hours before reaching a conclusion, but in other use cases if it’s a home sharing platform or if it is a travel platform then you kind of want to know within 30 or 40 seconds. So speed matters a lot specifically in those examples.

Peter: Right, right. Okay, so you mentioned machine learning a couple of times already, I think, can you explain…is the machine learning really in the recognition of the face itself? How are you using the technology?

Husayn: So in a number of ways, neural networks in particular, it is at different layers. So one is the classification of the documents so is it a passport, is it a driving license is it…for example, a California driving license and so on. Second is extraction of the characters; third is determining a range of different fraudulent patterns; font, style, it can be image tampering, it could be copy/paste and so on and so forth, but when we release a new model then as the results come through, we typically have to have humans double checking to annotate and correct, for instance, and that is what trains it. So it’s human-assisted machine learning, especially for the fraudulent patterns or fraudulent factors.

And then there are the examples, as we talked previously about, the very sophisticated fakes that if they happen to be able to cheat our system then our client, two weeks or four weeks later, when someone hasn’t paid their credit card or whatever it may be will flag to us that, for instance, this fake has cheated our system.

What they like the most is we use those samples to build the next version so that you can essentially fool us once, but it’s going to be increasingly hard for you to fool us again. What businesses, in particular, care about is those sophisticated fakes. That is our pace of innovation and pace of improvement that really has helped set us apart and we’ve only been able to achieve that fundamentally as a result of our machine learning approach.

Peter: That’s interesting. So you’ve got humans that are really helping the model improve and then is the model improving by itself anyway as well?

Husayn: So there are different models for different functions…when it comes to where human assistance is particularly useful is on the fraud factors and so when an ID comes through, if it needs to be double checked or if their thresholds haven’t been met, humans are double checking and we have internal feedback loops.

Those internal feedback loops with our own team help significantly improve accuracy on the models because they are kind of eliminating the false positives and then we have external feedback loops with clients. Where again if we happen to have missed a sophisticated fake, they then flag it to us and then we use those samples to build future versions. So when it comes to what sets us apart which is fundamentally our fraud detection capabilities being so strong, that is the result of, in large parts, feedback loops both internally with our fraud experts, but also externally with clients.

Peter: Interesting, interesting. It seems like it’s a bit of an arms race in some ways, or is it? Maybe you can answer that question because the fraudsters have access to the latest technology as well and they can hire people who are really smart as well, so do you find it an arms race, or do you feel like they won’t go to your clients, they might go to somebody else who has less sophisticated type of, you know, type of fraud detection?

Husayn: Yeah, it is actually an arms race, but it is one that we are now able to effectively sort of counter. So in some ways, before us, the way I look at it is like they were armed with tanks and the solutions offered were knives at best because you don’t need to be a sophisticated fraudster to just guess or steal someone’s date of birth, name and address to be able to cheat and open an account.

But now, you have to show an ID and show your face and you have to make sure that all the fake techniques that capture all the attack vectors are prevented and so on and so forth. So what fraudsters, what they want to do the most is the sophisticated fraudsters want to commit fraud at scale so they want to cheat hundreds or impersonate a hundred people every day, for instance, and they are extremely sophisticated so they have…you can you imagine the super sophisticated ones often work out of office blocks, they have performance related pay, sick pay and everything else, very sophisticated machinery, very sophisticated techniques. But what they want to do more than anything else is actually to be able to cheat systems at scale, as I mentioned.

So what we do, especially with clients that you use our liveness check for instance, is you have to do a 10-second recording, you know, read three randomly generated numbers and do a task and so on. Now that isn’t very scalable, first of all, you can’t cheat that by doing or pretending to be someone else or holding up a photo, or something similar because it’s a challenge and it’s been designed to prevent that, for instance, and secondly, as a bad actor looking to cheat a platform, why would you go through one that is A. very time consuming and B. super difficult to cheat as opposed to many other alternatives that are much easier to cheat.

That’s in many ways, why we’ve become the platform sitting behind a lot of the fintech community. All the communities are going online, including some of the mainstream banks. That’s because by us sitting in the middle of all these different businesses, our models are learning from all the data and all the attack vectors that are sort of looking to be broken across all these different businesses. All models get stronger and all clients benefit as a result of a bad actor trying to cheat one system and therefore, we learn that pattern. So if that bad actor has sold a hundred fake IDs, we would catch the other 99 coming through other clients’ for instance.

Peter: Right, right. When you’re going into banks today, are you…this might be their first facial recognition or any kind of thing…it wasn’t very long ago, you would open up a bank account or you even get a bank loan and there was nothing automated at the bank, it was all a manual process, so are you going into banks and this is their first foray into your kind of product and you’re sort of starting from scratch, or are you going in and replacing what’s existing?

Husayn: It’s a good question, in many ways both. So three or four years ago, the mainstream banks would talk to us, but we didn’t feel as though they took the wave of the neo online challenger banks too seriously.

Peter: Right.

Husayn: In some ways some of the early assumptions were, you know, this is a new wave, it’s a bit of a fad and how can you have a purely online bank. Surely, you’re going to get all the fraudsters disproportionately and therefore the model is going to crack or break. What everyone has been able to witness, especially in the last two or three years, is that these online banks not only have they been able to prove that they’re scalable in on-boarding very many customers very rapidly, but more importantly, they don’t have any excessive or any out of the norm fraud exposure than the mainstream banks.

So when these mainstream banks looked under the hood and saw that technology such as us in identity verification or (inaudible) for behavioral stuff or iovation or I guess, ComplyAdvantage for watchlist searches and different types of machine technologies being put together, assembled to help with the on-boarding compliance process, they’ve started to adopt since. And to your question as to whether is it new for them or not, what the mainstream banks are now doing and have been doing for a few years now is either they are buying an online bank or spinning up a completely different online bank division and for that it has been designed to be sort of new age and tech.

Once that model is proven out, it is reverted back to the bank. So if most of the mainstream banks…their tech stack is not necessarily equipped to be able to do work in an agile way, the way you’d expect a fintech or a neo bank would, but they are all moving in that direction and that’s why we are partner with the Salesforce Financial Services Cloud, or equally BBVA might be one of the exceptions where as a bank it is public knowledge that their whole tech stack has been improved end-to-end to ensure that they’re able to plug in the latest technology, specifically if it’s machine learning and API-based.

Peter: Right, right, okay, fair enough. So I’m curious about the fact that…you know, you’re based in San Francisco, but your company is based in London. Tell us a little bit about where do you operate and why are you here and not in London, tell us a little bit about the structure of your company.

Husayn: Yeah of course. So, we’re 240 people, the growth in sales in our largest market, our fastest growing market is in the US, hence, why I’m based here, but our engineering base and the technology side is predominantly in London, but also we have a team in Lisbon as well. So we have 7 offices in total, we have three or so colleagues in Paris, three in Delhi and also two in Singapore.

To your question of why I’m based in the States in particular, in large part because in Europe or at least western Europe, we are pretty much the leading provider in many ways. In the US, we are known in the community, but not as well known as we will be hopefully over the coming years so that’s essentially why I’m here. As you can imagine, it’s in many ways, a much larger market too.

Peter: Right, right, okay, fair enough. So I was reading recently about this Global Financial Innovation Network, GFIN, that cross border regulatory sandbox initiative. The FCA in the UK was one of the real driving forces behind this because this is a global problem that obviously knows no borders. I mean, the people hacking into these systems could be in Russia, they could be India, they could be in, you know, in Chile, we have no idea where these people are.

I’m curious about the fact that you’ve got all these different countries, there’s certain best practices with fraud so maybe, with anti-fraud, I should say, so from a regulatory perspective how important is it for an initiative like this to kind of come together and really combat fraud in a global way?

Husayn: There is a big need and the regulators can play a big role so we’ve been very fortunate in that, having the majority of our product and technology team based in London with the Financial Conduct Authority and their sandbox approach and how they have both competition and innovation as their focus area. They’ve been able to help us and companies in a similar space as us work hard to get the right balance between using the latest technologies and achieving really strong outcomes and results. So we were part of the first FCA sandbox, for instance, and it was sort of announced a few weeks ago we’re part of the current batch again so cohort 5.

With GFIN, GFIN is essentially, like you said, a similar setup to the FCA sandbox, but it’s across multiple jurisdictions so we’re part of the current GFIN batch too. It’s the first time it’s being done. So we’re continuously communicating and engaged with the regulators because we want to help as many countries as possible enable an online digital ecosystem, but underpinned by strong fraud prevention techniques because more and more countries, not necessarily emerging, even to develops like BaFin, for instance, in Berlin and in Germany. They’re looking at London and they’re looking at Paris and New York and they can see the progress that fintechs have made that has been far more than what they’ve been able to achieve.

Peter: Right.

Husayn: One of the underlying reasons which is in Germany the regulation is still a little bit outdated, it requires like a video Skype call for instance for you to be able to sign up to a fintech whereas digital technologies can be used in most other jurisdictions. So we’re continuously engaging and developing both learning and teaching, or sharing at least, and these sandboxes are the perfect forum to help facilitate that because the regulators are very much interested and very much want to learn, again, not just across the range of..not just the fintechs, but the platforms that are enablers to these fintechs.

Peter: Right, right, okay, makes sense. So I’d love to get your perspective on something. It’s an issued that I’ve been thinking about for a long time…you know, obviously the Equifax hack a couple of years now and the whole idea of having your identity or details about your identity stored in a central database is…it already seems old fashioned, but we haven’t really yet come up with a better way to store this data. How would you recommend we store identity data?

Husayn: So the first is to recognize how important and how sensitive it is and secondly, as you mentioned, it’s how it currently operates which is not fit for purpose. So essentially, for the current model it’s predominantly centered around credit bureaus’ centralized leaking database. The antithesis or the solution or the counter to that would be the opposite which is a much more decentralized solution where the consumers own and control their legal identity.

So there are different models and we are partners to quite a few of the different technology solutions looking to solve this identity problem essentially calling it a portable identity or decentralized identity. Some are using blockchain, others are not, so different approaches, there is a wave of activity now that has been initiated and that is making progress.

The fundamental objectives are first of all, for the user to own and control their legal identity and share that with the businesses that they choose to so the actual way that’s stored. There are different models and different options, again, the consumer could choose. One would be your smartphone, for instance, in a secure way, an alternative would be your local bank or the bank that you trust with your finances. They could be a custodian of your legal identity, a third could be AWS, Azure or a cloud service, for instance.

So there are a number of ways and naturally, if it’s in one place, you control and you are able to manage, it’s going to be far superior and far more secure than what we have currently. What we have currently is you want to sign up to your local gym, you want to sign up to supermarket’s loyalty program or anything, any account that you want to open, you have to share so much data.

That exponentially increases the likelihood of one of those businesses getting breached and therefore your data leaking on the dark web and therefore, bad actors being able to very easily use that data to open accounts in your name. So we have no doubt that the future is going to be much more of a decentralized trust and a network in the sense of portable identity and because we believe in it, we’re actually very involved in helping different platforms looking to solve that problem.

Peter: Right, right. So we’re almost out of time, but a couple of things I want to get to. You know, I think about this, do you think that in the future, in the distant future, do you feel like identity theft is going to be just a solved problem where it’ll be so rare as to be very unusual because tools like yours get so sophisticated, or do you think we’re always going to be challenged by this?

Husayn: I think that there will be…I can’t say it can’t get worse because it’s actually getting worse, (Peter laughs) but there is now effective technology out there to address it. So it’s going to be…you’re probably going to see two camps; those who are deploying effective technology and therefore are able to be protected and those who are not which increasingly are going to find it difficult, both costly and so on, to continue. This is ripe for there to be a network of trust built that gets stronger, the more data there is and therefore is able to offer all the benefits because of different drivers there is increasing regulation around privacy and around security, there is consumer demand for personalization and efficiency.

There is a growing…I guess, the power of machine learning only in the last five/six years has really been able to show itself and all the benefits that it brings and so on and so forth. So the answer to your question is that we are going to very much solve this identity problem, but it’s going to take some time, but we’re definitely on the right path.

Peter: Right, right. Okay, last question then, what are you working on right now, what are some of the things that you’re excited about that might be coming down the track?

Husayn: So in some ways, I’m kind of interested anything with similar use cases, or interesting patterns in different geographies. So one of the companies I am following closely…one of our partners is Tala, for instance, in emerging markets like the Philippines, Indonesia and elsewhere, they’re able to do micro loans so $20, $50 lending and they are able to do so because it’s all online and they’ve been able to reduce the on-boarding cost to a few dollars as opposed to tens of dollars if they were required to see everyone face-to-face.

On the other end of the world you have…if you look at Southeast Asia, examples of MoneyMatch or similar ones that are doing very, very cool stuff. So across the board the fintech wave in many ways is still the first wave and it’s catching on in different parts of the world so it’s very interesting and exciting to see that.

Secondly, back here in the States it is, and sort of the UK, we’re running some pilots on portable identity so the topic we were just discussing. Ultimately, it’s going to be the user that owns and controls their legal identity and they’re able to decide which businesses they want to take that to so that’s going to be pretty exciting to see it progress.

Peter: Okay, yes, it’s going to be fascinating to see it. I find this whole topic just so interesting that we are here in this process of moving to a completely different way of doing things. Anyway, Husayn, we’ll have to leave it there. I very much appreciate you coming on the show today.

Husayn: Likewise, thank you for the time.

Peter: Okay, see you.

Husayn: Bye for now.

Peter: You know I was reflecting on this after I hung up the phone with Husayn and, you know, I’ve recently opened up a couple of new accounts, I’m not going to name names, but sometimes the systems to open these accounts even with fintech players, they’re not as sophisticated as I expect, particularly when it comes to identity verification. I’m sometimes now expecting to have a video or sophisticated photo verification done with ID and oftentimes, that’s not the case.

I think in Europe it’s probably more so, but here in the US I think we still have a long way to go before we get an identity verification system that everybody uses and I think companies like Onfido are really going to help drive this change where they have the easy to use systems that catch so much when it comes to the bad actors.

Anyway on that note, I will sign off. I very much appreciate you listening and I’ll catch you next time. Bye.

Today’s show was sponsored by LendIt Fintech Europe 2019, Europe’s leading event for innovation and financial services. It’s happening September 26th and 27th at the Business Design Centre in London. Registration is now open as well as speaker applications. Find out more by going to lendit.com/europe.