Show 99 – Fintech Forum 2020 – Part 1
The first of two episodes recorded in partnership with CFA UK on the subject of Fintech for 2020, where we interviewed a number of the speakers who were due to present at their Fintech Forum.
The event itself was meant to have taken place in London in March, but as a precautionary measure, to minimise the risk of the Coronavirus, CFA UK took the decision to cancel the event, however they weren’t prepared to let COVID-19 beat the podcast, and so we recorded all our interviews online.
In this episode:
- Holly Black, Senior Editor, Morning Star interviewed Richard Davies, Group Chief Operating Officer, Revolut about the growth of the company
- Nikola Tchouparov, Co-Founder and CEO, Moneyfold Ltd talked about Blockchain and cryptocurrency
- Iuliia Shpak, Quantitative Strategies Specialist, Sarasin & Partners discussed Machine learning in practice
- Geoff Horrell, Director of Innovation, Refinitiv, explained how they work with Fintech alliances
For more information from CFA UK, visit www.cfauk.org
Richard began by giving some background to Revolut. He said that the company, launched in 2015, was originally centred around offering unique multi-currency wallets, as well as a card to help people save a lot of money if they were doing a lot international travel or international payments. He went on to explain that since then, the business has developed into many different areas for example, cryptocurrency, stock trading, as well being able to be used in a number of new countries Europe-wide as well as other places like Australia and Singapore.
Holly then asked Richard what he thought the biggest challenges are in this area and he replied saying that the challenge is always having the people to realise the opportunity that may be out there. He then went on to describe what he believed were the biggest opportunities within the sector. Richard said that the vision is to be a super app, a platform for all the financial services needs of customers.
Richard said that a lot of innovation often emerges from a crisis, which then becomes a wave of maturing these innovations. He explained that this means that some of these innovative ideas will get to be really big and some will fall by the wayside or be bought, so, we’re seeing some of that happening in the last few months, both in Europe as well as the US.
Holly explained how she believed that one of the big things that came out of that financial crisis was a loss of trust of some actual banks. She said that she thinks people can be relatively cautious when it comes to their money and who they will hand over to. She then posed the question ‘as a new business, how do you get customers trust?’ to Richard. He responded by saying that it is something they take very seriously at Revolut and they gain trust with two main parts of their business. Firstly, being regulated under e-money requirements, which means that they safeguard all of their clients’ money with very large global banks to ensure its fully protected. Secondly, is fully deposit insured bank accounts under full bank regulation. Both forms give very good customer protection.
Holly described what she calls the ‘mistress phenomenon’ where people will take out an account with a new bank, but it’s their secondary account, it’s not where they’re getting their main wage paid into or where the direct debits are coming out of. She said that she believes this is a big challenge of distruptors in this space of fintech. Richard replied saying that at Revolut they really focus on how much people use their product as oppose to whether or not it’s their primary or secondary account.
Richard points out that whilst many traditional banks are trying to make their way into the fintech space, they also have their own huge benefits to being a traditional bank. For example, when it comes to things like mortgage lending, HSBC are extremely well positioned because they’ve got super low cost of funds, super low cost of capital, and therefore you can see them currently dominating the mortgage market. Therefore, it all depends which bit of the market you look at as to which bits are under threat.
Richard added that for him, the most interesting insight from the data they see and use is the fact it’s everywhere. He says that they’re trying to embed data throughout the business to a really strong working level in order to make a difference. For example Revolut has internal machine learning algorithms that look at user’s photos so that if a customer is looking to try and change their device but it could actually a social engineering attack where someone’s trying to take a person’s account. Machine learning for image recognition in comparison can decipher if it is the actual customer or an attacker.
When asked about cyber security, Richard explained that at Revolut they use some of the biggest, most globally recognised cloud hosting providers to provide security as well as an in-house information security that’s highly skilled. However, he went on to say that there can be a bit of friction with security as the more controls you put in place, you can damage the customer experience and the ease of use.
Nikola began by defining blockchain and cryptocurrency. He explained that we have DLT, Distributed Ledger Technologies, which is an umbrella term, capturing many different types of technologies and under that umbrella is blockchain. Blockchain was created about 12 years ago with Bitcoin, but now has evolved into much more. He went on to explain that the idea behind blockchain is that it is an open, permissionless system that anybody can join. However, this brings up some important questions; How do we coordinate that? Who owns what?
Nikola then told us how there are multiple competing visions for blockchain and cryptocurrency. He went on to say how the current popular decentralised finance, crypto-DeFi, vision has unregulated instruments sitting completely outside of the financial system and that people are using them to invest or to speculate, which has been quite successful as it has attracted around $200m worth of value.
The second vision that Nikola mentioned is that of the traditional financial providers. He said that financial services providers like Intercontinental Exchange (ICE), the parent company of the New York Stock Exchange, have launched a new market for the future where the underlying asset is bitcoin that is physically settled. Similarly, he said that we have seen the CME launch futures and options that are cash settled where the underlying asset is also bitcoin. So, this is another vision of decentralised finance where traditional institutions are getting involved in investing, trading and making markets in some of these new digital coins. The final vision for the future that Nikola mentioned is that how we take these public permissionless systems and use them so that we can build value added financial services on top of them. One example he gave of this is fiat-backed stable coins, where instead of on paper or instead of a traditional digital ledger, they’re on top of a public permissionless ledger. Nikola believes that this is where the true potential lies to realise cost savings on building an operating financial service.
Nikola speculated that most people are not really looking into this space and are not aware of the potential of it. He highlighted a study that the FCA commissioned last year, which found that only about 3% people in the United Kingdom are involved in this space. Meaning that 97% of the general population doesn’t know about cryptocurrency and are not involved in this area.
Nikola then went on to explain to us that if you want to transact with Bitcoin, ether or other cryptocurrencies, that’s already possible, but if you want to use the same technology and get the same benefits, limiting yourself to pounds, euros and other national currencies, it’s a very tricky proposition. He said that at Moneyfold, they have come up with a way to do this safely and securely so that you can have national currencies on a public permissionless blockchain without the need any financial intermediaries.
When asked about the future of blockchain and cryptocurrency, Nikola said that he believes that the vision of DeFi will continue to go on for a while because it’s exciting and there’s a lot of new talent involved. When it comes to the legal and regulatory questions, there is some uncertainty about whether this will be allowed to continue or not. Nikola said that he thinks we will see more and more stable coins launched throughout the world and that we will continue to see those stable coins being used to build new Value-Added platforms. He also explained that his company is helping many others with the use of their stable coins, saying that this is an area that the Bank of England is directly looking into.
For more information;
- University courses – beginner level courses all the way to advanced software development
- Academic research
- Coinscrum and theory meetup events
Iuliia began by explaining that machine learning methods have been remarkably successful for a wide range of applications. AI can provide significant opportunities for the industry and for asset management firms. She mentioned that we can already see many firms successfully applying some of these technologies, especially in areas such as execution and trading operations, client reporting, data analysis for ESG performance assessment, among many other use cases. The strength of machine learning is in its ability to analyse complex relationships. Machine learning can help identify complex relationships in high dimensional spaces and identify input output relationship in hierarchical threshold on continuous linkages. Because of this, machine learning has a strong advantage over traditional sophisticated models. However, Iuliia also made us aware of some caveats that we have to consider when applying machine learning and investment strategies.
Iuliia then went on to further explain these challenges. Firstly, many machine learning methods had developed originally in datasets such as image or pattern recognition, which are quite unlike financial times serial data and in usual machine learning modeling assumptions, the sample contains independent and identically distributed data (IID assumption), which normally doesn’t really hold in financial data, which leads to methodological problems. As a result, it’s not difficult at all to find a model with excellent predictive power in sample, but the model completely fails out of sample. In other words, overfitting is a real problem. Secondly, one also needs to think whether a machine learning algorithm is actually the right tool to interpret the data at hand.
Iuliia explained that within the financial industry, data is released quarterly, so even in 10 years, that would still yield 40 observations, which is by no means enough to train on past machine learning algorithms.
Finally, another important issue can be explainability of AI models. Researchers therefore should make use of all available techniques such as SHAP values, relative feature importance, partial dependencies to explain and be able to understand which variables are responsible for predictions that the model makes. However, there are also many positives that come out of machine learning in the financial sector. Iuliia highlighted that there has been a lot of excitement about machine learning in the industry over the last several years, and we have observed machine learning communities growing and becoming more open source, which means offering new techniques and increasing learning efficiencies.
As described by Iuliia, overfitting occurs when a model is too closely or exactly fitting a particular data set, so an overfitted model normally contains more parameters than the data actually justified. Financial data has a very low signal to noise ratio, noise is random, so you cannot forecast random data. So, if you build a model that produces forecasts out of noise, you have a problem. Therefore, researchers should take special care when building machine learning models in financial markets. Iuliia also said that it is important to consider the main expertise, she continued by saying that you can’t just throw data science skills without taking a context of financial markets so, the practices which researchers in the industry should use are regularisation, cross-validation and sample learning.
Iuliia explained that by exploiting the information in the forecast, a researcher can affect the future price. So, reflexivity of financial markets makes the whole problem of learning about investing fundamentally different from machine learning application. For example, a model is having a task to distinguish faces of cats and dogs, which would not change by the fact that they recognized and metacognition problem. Therefore, machine learning techniques are best to be applied in markets or in time serious data sets which are not subject to reflexivity dynamics.
To finish off Iuliia then shared the other areas where machine learning could make a difference with investing. She highlighted three key pillars. The investment management, security selection, portfolio construction and slash optimisation and execution which in quant investing are normally integrated with risk management as well. Machine learning is already very successfully applied by some brokers and some investment firms in execution and trading. There are significant opportunities in portfolio construction and optimisation, so this area still remains under-researched and provides a lot of exciting opportunities for future research. She added that in the industry two very strong trends, significant growth of alternative data and ESG investing, have been witnessed where machine learning was part of their strengths, which provides very, very strong opportunities for making use of alternative data and for better assessment of companies in terms of that ESG performance.
Geoff began with a short introduction to Refinitiv and Refinitiv Labs. He said that Refinitv is a global provider of data, of workflow tools for financial professionals and also of trading infrastructure operating in over 100 countries around the world. They have a group of full stack engineers, data scientists and experienced product designers, working very closely with businesses in order to find new opportunities or areas that see potential growth.
To explain how Refinitiv partners with fintech companies, Geoff gave an example of how they have tried various different approaches to look for interesting technology firms or emerging technology, helping Refinitiv to stay close to some of that technology and focusing on getting a strategic alignment with business. Refinitiv focuses very specifically and proactively on looking in particular areas to de-risk in order to find partners rather than receiving lots of inbound requests – quite a shift from the technology run approach to a strategic top down approach.
According to Geoff, the number one reason everybody comes to Refinitiv is for their data, from traditional market data, investment data to financial crime data sets for financial crime screenings as well as providing an advisory perspective. One challenge that Geoff highlights is that some Fintechs are concerned that they will be slowed down due to the scale of Refinitiv, however, he states that that’s where the Labs come in, helping with agility to help work with Fintechs.
Geoff shared some interesting examples of alliances with other companies and how they are successful. The first example was one that he was really excited about, he explained that Refinitiv now partner with Trulioo for digital identity, which helped them bring their strengths, they will check data for PEPs and financial crime screening with Trulioo’s technology in terms of individual customer onboarding and verification of documentation. Another example that Geoff gave is their partnership with BattleFfin. Refinitive invested a minority investment into BattleFin meaning that their respective successes are aligned. Battlefin is a marketplace for alternative data providers, bringing their data in and making it available for testing and validation so that hedge fund investment managers can get a sense of what data is out there. Refinitive provide their own data alongside the alternative data so that people can compare and contrast the old data approach vs. the traditional data approach and link those two together to get the best of both worlds.
The next example that Geoff shared was Refinitiv’s partnership Finbourne, a London based firm, explaining that Finbourne actually built a core part of their infrastructure. The final example given is Refinitiv’s investment in a company called ModuleQ, who provide an intelligent agent feeding on Refinitiv’s data. They’re plugged into a mix of teams and provide an intelligence service for business professionals in the business intelligence and strategy functions across Refinitiv’s customer base. Whilst ModuleQ are a good AI company, they need the data that Refinitiv provide to power their solutions. Highlighting here that with each of these examples Refinitiv have taken a very different approach for partnership.
Geoff explains that they made a lot of investments in the early stage of blockchain and DLT type companies and it really wasn’t a clear enough business driver to form successful partnerships. He stated that they were relatively immature and hadn’t been screened or didn’t have existing banking backers, which was a challenge. Another challenge that was highlighted was that Refinitiv made too many investments towards the beginning for what, at the time, was a very small team to really monitor and make a difference with those firms. Geoff also said that another struggle for them was small companies coming to Refinitiv to drive scale or simply as a distribution partner, which doesn’t always work for niche companies.
Looking at the future, Geoff said that they are looking at what’s on the horizon, coming out of academia, big tech such as the cloud providers and other technology firms. He says that there is a focus on natural language processing, with some of the biggest innovations coming out of companies such as Google and Facebook. Geoff believes that you have to look at what a fintech’s competitive differentiation is going to be against some of the big technology firms who have the resources and the research teams to really do some cutting-edge work, highlighting that the benefit is not in some of the core technology, but in the domain expertise. It is also interesting to look at the speed of acceleration in machine learning and machine learning platforms. Geoff also added that, in terms of business areas, Refintiv are really interested in the area of sustainability and climate, looking for firms that are working in that area, as they believe that it’s going to be one of the big trends for the next while.
Geoff’s said that for more information, look at what is being worked on in Refinitiv Labs, where they also have a series of blogs on everything from machine learning to sustainability and their partnerships, highlighted in the the Perspectives section of their blog.