SDS 276: Data Science in Wealth Management

Welcome back to the SuperDataScience podcast everybody. I’m super excited to have you back here on the show. Today we’re continuing our series of episodes about data science in different industries. And today we’re talking about wealth management.

So what is wealth management? Well, wealth management is an investment advisory discipline that helps people manage their financial planning, portfolios of investments, enhance income, grow their long term investments and things like that. The goal of full revenue is hard to determine, but the top six banks report a revenue, some of over $70 billion and the global volume of net investible assets is estimated around $60 trillion and expected to rise to 70 trillion. That’s trillion with a T, $70 trillion by 2021. As you can imagine, it’s quite a large space. An interesting side note here is I was listening to a podcast with Naval Ravikant, who is the founder of Angellist, a very wise and genuine person. And, this was actually on the Joe Rogan experience on Joe Rogan’s podcast, episode number 1,309 if anybody is interested. One of the best podcasts I’ve listened to ever, Naval is very wise, very genuine, and got a lot out of it. And I’ve got a lot out of it. And one of the things he mentioned, which was interesting is that wealth is not a zero sum game. It’s a positive sum game.

Everybody can be wealthy, everybody can create products and services and help make the world a better place and through that grow their assets, their resources. And on the other hand, things like status are a zero sum game. For instance, if for somebody to be number one, somebody has to be number two, somebody has to be number three for number three to go to number two and number two has to go down to number three. So status is a zero sum game. Wealth is a positive sum game. I found that quite an insightful idea. Anyway, moving back to our episode. So we talked about what the industry is all about. Let’s look at some use cases of data science and artificial intelligence in wealth management.

So use case number one, probably one of my favorite ones, automated portfolio management. So there’s a lot of data as you imagine with markets, with securities, dogs, bonds, options, futures, forex, now cryptocurrency, there’s a lot of different data sources that represent these financial instruments. And that data can be analyzed by machines that it doesn’t, a human doesn’t have to sit there all the time and analyze patterns. That’s, that’s the way it’s being. But more and more, these things are becoming automated and the portfolios that are constructed for wealth management are based on decisions made by artificial intelligence machines and they’re showing really good performance. So here’s a couple of examples. So one of the example, number one, one of the largest investment management firms in the world, the BlackRock solutions developed Aladdin, a risk platform that uses machine learning to provide investment managers with risk analytics and portfolio management. According to BlackRock Aladdin performs daily monitoring on 2000 risk related factors such as currency rates, which makes wealth management quicker, more efficient and more automatic. It is used by over 25,000 professionals and supposedly handles over $10 trillion in assets.

And example number two is one of my favorite companies, Renaissance Technologies is a company that I was fascinated by when I was younger, when I was still at university, still fascinated by them. Renaissance Technology is not a well-known firm, but check this out. So they have about 250 or 290 employees and they are one of the most successful investment firms in the world. They’re based in Long Island, New York, and they have this fund called the Medallion fund. They have several funds, but the Medallion fund is the main one that they’re famous for. It’s mostly for employees only, so it’s really hard to get into that fund. The company was founded in 1982 and from 1994 to mid-2014, they averaged… Get this 71.8% annual return, 71.8% annual return.

That’s crazy. And when you think about… So for instance what was the funds worst years, what about the GFC? And so on? Well, from 2001 to 2013, the fund’s worst year was a 21% gain after subtracting fees, not a lost the 21% gain in their worst year. Some companies don’t make that much in their best year. Medallion also reaped a 98.2% gain in 2008 the year of the financial crisis. And when, S&P 500 index lost 8.5%. But you could say that maybe they just went in the right direction in 2008 and that’s how they gained their profit. But a track record of 71.8% annual return for over 20 years and the worst year being at 21% return. That is crazy. And the most interesting part why this is relevant to this podcast, why it’s machine learning, data science, AI is because they… A third of their employees have PhDs and not in finance but in fields like physics, mathematics and statistics.

And Renaissance has been called the best physics and mathematics department in the world. So they predominantly hire scientists, physicists, mathematicians and statisticians and not people who are in finance. Just stats as an example of what data science, data analytics, machine learning, AI, statistics, all of these disciplines that we work in, in data science, what they can do in a wealth management and finance. So just an example to think about. That’s very inspiring company.

Example number two or use case number two, sentiment analysis. So looking at data on financial markets is a great start and it is a very powerful for AI to do, but that’s not the whole story. There’s also events that happen, whether it’s election, scandals, even sporting events, those things can affect the value of different assets. And it’s important to look at them.

And that’s where sentiment analysis comes into play by monitoring feeds like Twitter and what’s happening on social media, artificial intelligence and other NLP algorithms can figure out that something is happening, send a signal to the wealth managers and they can act before the rest of the market has actually realized what’s going on. And protect the funds or make appropriate decisions.

Example here is the company Refinitiv services over 5,000 investment firms and hedge funds. It supports hundreds of billions of dollars in bond and Forex trading daily and tracks over 3 million individuals and entities potentially dangerous to the international business community. The system sends out up to 7 million price updates to financial markets every single second.

Okay. Use case number three, personalized advice. So wealth management firms need to strike a delicate balance between personalization and cost reduction. So on one hand, if the client, manager or portfolio manager is tweaking the portfolio too often, then they might get maximum performance out of it. But the fees are going to outweigh the gains because their time costs money. And usually it’s not cheap for a portfolio manager to constantly be looking at your portfolio. On the other hand, if you cut the costs by getting your portfolio manager to look at your portfolio less, then you’ll cut the cost, but also you make less profit. So there’s a balance between personalization and cost reduction. On the other hand, AI is scalable infinitely so and therefore and also doesn’t cost pretty much anything for the AI to look at the portfolio and tweak it round the clock. There’s no difference between tweaking it occasionally around the clock. It doesn’t have a cost associated like with labor on the artificial intelligence and therefore things like supervised learning can help optimize this approach. Basically supervised learning can learn from how human advisors have advised and tweaked portfolios and do that in its own way as well.

Example, here we have is Vanguard Group offers a personal advisor that gives its users investment advice based on automated algorithms combined with insight from human advisors. So they are doing like a hybrid combining AI and human advisors. The system takes personal factors such as age, risk aversion and existing investments into account when calculating the optimal investment track. And the interesting thing is like unlike most other areas where young people tend to be the early adopters of a new technology, around 50% of Vanguard’s users are over 65. Vanguard Group’s plan is to gradually phase out the human factor and allow users to interact directly with AI that handles over $65 billion worth of assets. However, as we’ve learned in previous FiveMinuteFriday episodes, AI chatbots still have quite a way to go before we humans will be fully comfortable talking with them, but we’re on the way that.

All right, use case number four, report generation. So wealth management needs to obviously be reliable and responsible and also has to prove to regulators and customers as well that they are reporting on all types of events. So being precise in your reports can be time consuming and tedious and a lot of man labor or human labor is required for that. And therefore an artificial intelligence can take over. And it’s not a simple task. Not a matter of just copying some numbers from a spreadsheet. It’s actually and looking at different portfolio investments and a lot of data needs to be identified, interpreted and analyzed and integrated into the reports and therefore it is a more, more complex task than something that are very simple, I don’t know, maybe an RPA or robotics process automation type of where I could do this a bit more complex. So here a more supervised kind of artificial intelligence can come in useful.

An example here we have is Quill, which is a program that uses natural language generation rather than NLP, natural language processing. It uses natural language generation to help communicate insights from structured data. The software can be taught using existing reports, for example, detailed sales reports for regional managers and learn to create a new reports automatically. Narrative science, the company that developed Quill claims to have helped a bank automate the generation of suspicious activity reports and important document that they must file off to detecting suspected money laundering or fraud. They report savings of over $200,000 a year and reduced time to generate one of these reports from 4 hours to 1.5 hours, one and a half hours. So that’s pretty cool. Just with artificial intelligence and using natural language generation. And we don’t often think about natural language generation. We mostly think about natural language processing.

But as you can see, there’s a massive use case and there’s plenty more massive use cases for natural language generation, which is kind of the opposite. And a very, also very interesting space to get into if you’re in data science, something to consider as well.

And use case number five, loan underwriting. A huge and often over overlooked part of wealth management is loan management. So machine learning algorithms can assess a different trends in loans and detect situations that might influence a loan. And so the questions that need to be answered are, for example, are young executives more likely to default on a loan rather than older business owners? How does marital status change the odds of default and so on? So basically understanding, assessing the risks of any specific businesses probability to default on a loan. And if a loan, a company that provides loans, has a lot of data and they usually do, if they’ve been in business for a while, then that data can be used for supervised learning for artificial intelligence and machine learning to create, to understand what patterns exist and get insights from them.

And the example we have here is Underwrite.ai uses nonlinear algorithmic modeling to analyze thousands of data points from credit bureau sources, allowing it to accurately model credit risk for an individual consumer. They claimed to have reduced the first payment, default rate of an online installment lender from 32% to under 7% within nine months, way below the industry average of 35%. Their model is built on publicly available data can even help companies that have little or no data of their own. So that can be useful to many different businesses as well. Even those that are starting out in this space.

So there we go. That’s data science in wealth management and five use cases. Hope you enjoyed this. And if you are in the wealth management space, maybe that helped you generate some ideas. Of course, there are plenty of more use cases of data science, AI in this industry. And if you are not in wealth management, then some ideas from here could help you in your industry. For instance, natural language generation. How can you use NLG into your business or in the space that you’re at?

On that note, thank you so much for being here. I look forward to seeing you back here next time. Until then, happy analyzing.