Recently MAS announces that it will allow full suite of life insurance products to be offered online without advice. Moreover, regulations will be updated to include robo-advisers (Source: MAS opens door to full suite of life products online, and much more in fintech).
Just last Sunday, the Straits Times announced that the launched of an insurance ‘robo-adviser’ which later I found out was not true. More on this later.
So, what’s the impact to consumers and (human) financial advisers?
To begin with, many consumers are already using technology when they use certain financial services. Interbank GIRO is one such service. I too have been using technologies to assist in my financial advisory practice. If there is no such technology, I will invent one.
I started learning programming when I was 13 years old. My dad bought me my first computer at that age and it was the legendary Apple IIe. Besides learning to play games like all boys that age, I taught myself programming languages. The first language I taught myself was BASIC. I taught myself because I couldn’t get into the secondary school’s Computer Club. Within the first year, I taught myself the 6502 assembly language. Apple IIe was running on the 6502 CPU. At 13, I also learned Z80 assembly language when my dad got the Z80 card. Subsequently, I also learn dBASE language. I would use the dBASE to catalog a large number of church songs for the Youth Fellowship at that time. All these was done during my secondary school days.
I opted for Digital Signal Processing as my specialization in 4th year NUS when many of my fellow friends opted for material science instead. My first industry attachment in NUS was at Creative. I was given an assignment on implementing the MP3 decoder. Had a very fun time in this project.
As everybody know, my first career was an R&D Engineer. My first official job was in development of firmware in assembly language for a real-time decoder video conferencing product.
Subsequently, I join various companies to gain my exposure in R&D but my passion was in firmware development. One of the most interesting project I did was embedded linux. Long before Android (Android is an embedded linux) took the world by storm, I was already embedding linux kernel into embedded systems. I remember I studied source codes of actual internet hacking tools so that I can ‘harden’ the firewall I was building.
In later part of my engineering career, I shifted focus on precision motion control and I was involved in development of various HP printers. PID close loop control – sounds familiar?
But everything stopped in 2006 when I discovered that my type of engineering job had no future in Singapore. So I switched to finance as I predicted that this is the future of Singapore. Majority of my classmates in NUS also switched to finance and they are doing very well now.
My familiarity with software development did not halt when I switched career. I learnt high level languages to continue development to help me in my financial planning practice.
Here are the list of software I developed to help in my financial planning practice. I also use other people’s software which I also listed it here.
The Robo-Ranker is a software I developed to rank funds against its peers according to their returns over various periods. The purpose is to identify funds which must be avoided. Based on experience, funds that are underperforming compared to its peers tend underperform for a very long period of time. The Robo-Ranker can be access HERE (login required). A cron job is done every midnight to do the number crunching.
What the Robo-Ranker cannot do: It cannot determine the reasons why the funds underperform or overperform. Reasons funds perform well or not could be due to style drift, style becoming unfavorable or simply the core team resigned. So what I usually do is to ask the asset management marketing team the reasons and make a judgement call as to whether should I advise my clients to switch or stay put. As it can be seen that the software is merely an assistant. A human is still required to make a judgement call.
I developed the SRS calculator in response to concerns that the SRS actually tax capital gains. Singapore does not impose capital gain tax for investment made using cash. Hence, many people refuse to invest using SRS for the fear of paying too much tax. Hence, I developed the SRS calculator using an algorithm called relative-value described in one of the CFA text book I studied. The calculator does its job pretty way because it is based on published mathematical model. Nevertheless, some part of the model has no close form formula (like calculating IRR). Hence, I used a binary search algorithm. As it can be seen that technology does its job pretty well when there is a well-defined algorithm. In the event there is no close form solution, technology can find the answer based on iteration provided there is a convergence.
What the Robo-SRS cannot do: It cannot tell you whether the tax savings from SRS is “better” or worse compared to other tax saving strategies such as CPF Full Retirement Sum Topping-Up scheme and CPF Voluntary Contribution Scheme. In actual cases, I used various tax saving tools in conjunction with a holistic retirement plan which also consists of rental income (property purchases would involve paying more taxes), annuities, retirement products, etc. What a Robo-adviser cannot do is when the problem cannot be defined accurately. “Retirement Planning” is an example in which the problem is rather fussy and the kind of retirement plans the clients want depends on what their state of knowledge (which can change rather rapidly once they meet a human financial planner).
I developed a software to track the prices of bonds (not bond funds). I figured that it is cheaper for me to spend a few hours coding than to pay Bloomberg a hefty recurring amount just to track bond prices. Tracking bond prices is important to me because I use bonds as part of my clients’ asset allocation. The software I did is eventually integrated into the Robo-Portfolios Aggregator (to be discussed later).
Here is a chart showing the dirty bond prices with and without coupons reinvested using the data collected by my software:
What the Robo-Bond-Charter cannot do: The software cannot advice the client whether the bond is suitable in terms of risk and expected returns in the context of the client’s financial situation and objectives which changes frequently like all humans.
Some of my clients requested that they want to use ETF as part of their portfolio and Regular Saving Plan. What I did for them is to write a software that on a monthly basis will calculate the exact number of ETF shares to buy based on the ETF share price and forex rate. This information is emailed to them by my robot. My clients would than use the information received in the email to purchase the ETF concerned. It must be noted that a number of brokerage house and banks have created products that buy into ETF/bluechip local shares automatically and regularly to implement RSP. While my software appears to be primitive, it has one superior feature: It is able to support any ETFs (and theoretically shares of companies) listed in any part of the world. By the way, most foreign exchange allows minimum of 1 share. Singapore is still living in a stone age in which the minimum lot size is 100 shares creating problems for small purchases like RSP.
What the Robo-ETF-RSP cannot do: It cannot determine whether the ETF is suitable or not. Hence, a human decision and judgement call in the ETF selection is required. Moreover, the software cannot determine whether the RSP should be $1000, $10000 or $100,000 a month. The amount has to be determine as part of the overall plan in retirement planning and any other objectives.
Robo-fund updates to wordpress
I received tons of market updates in my email every day. Most advisers would just delete those emails because they are just too many. I wrote a software to extract these fund / market updates to upload to this website (which is running wordpress). Once it is on wordpress, it is very easy for me to browse through the contents. I will delete what are nonsense and categorised the remaining according to whether it is ‘useful’ information and ‘very useful information’. The market updates are than accessible to subscribers. For example, the ‘useful’ information is available for free to anyone registered on this website while the ‘very useful’ information is only available for existing clients or those who is willing to pay a subscription fee. .
What the Robo-fund updates cannot do: It cannot read the materials. A human person need to read to separate what is useful and what is useless.
Robo-Insurance Needs Calculator
I developed the Insurance Needs Calculator for Promiseland and is now part of the needs analysis template that is used for the advisers' Balance Scorecard. The calculator incorporates multiple dependents, multiple debts into its analysis using time value for money approach. The Robo-Insurance Needs Calculator double up as a sales tool because it produces nice charts that can shown to prospects.
What the Robo-Insurance Needs Calculator cannot do: It relies on someone keying in sensible data. It cannot read and understand client’s existing insurance policies legal jargons, fine prints and exclusions. It cannot recommend whether those policies remain suitable. Obviously, it cannot recommend any products. Only a human financial adviser can make provide suitable recommendations.
Recently the Straits Times reported that a FA company has created a ‘Robo-adviser’ for insurance. When I tested it, I discovered that it was a disappointment. The algorithm appear to be primitive. The worst part is that the so called ‘Robo-Adviser’ made specific product recommendations without looking into existing insurance policies and it actually recommended an integrated shield plan which could end up to be a policy replacement if the customer has an IP belonging to a different insurer. Under the existing Balance Scorecard regime, if the ‘Robo-Adviser’ is a human adviser, this will be considered as a major infraction. In fact, I may ask MAS whether would this constitute a major infraction.
Robo-Passive Income Analysis
The Passive Income analysis tool took me a few years to conceptualized. Once I got the concept right, it took me a few days to develop an Excel version and an online version. The concept is simple: given a target retirement income, how would various sources of passive income meet that need? This tool has also become part of Promiseland’s needs analysis when the adviser want to sell retirement products.
What the Robo-Passive Income Analysis cannot do: It cannot recommend retirement products and neither advice on the complicated CPF Life and its associated CPF schemes which changes almost every month. Based from what I observed from the online version, most users feed nonsensical inputs because they don't know what inputs to feed the robot with. Hence, a human financial adviser still need to perform the appropriate fact finding and research into the suitable products and government rules to advice the client.
Robo-Cash Flow Projector
The Robo-Cash Flow Projector is a generalized version of the Robo-Passive Income Analysis. This cash flow projector attempts to answer this questions: Given multiple sources of expenditure and multiple sources of income, how would one do up a budget? This software is to be used for both working personnel and retirees. This is the most complex software (since leaving my engineering job 10 years ago) that I ever did and I do not think I will share it with Promiseland nor online. Sorry! Obviously, I am using this tool in some of my financial planning cases.
What the Robo-Cash Flow Projector cannot do: The software relies significantly on the user’s goals and the ability to provide accurate breakdown in expenses and income. Hence, it cannot automatically find out how much a user is spending and obviously cannot read his mind as to what kind of financial goals he wants to obtain. Thus, a human still need to feed the software with relevant inputs otherwise the outputs will be nonsense.
Robo-Template Model Portfolios
I did not develop this software. But it is available to any financial advisers in Singapore who are not tied agents. The idea is this: the FA company will construct a few template model portfolios based on various risk levels. The company’s customers would have the investment accounts ‘tag’ to these template model portfolios. Once tagged, the customer’s money will be invested in a portfolio that will mirror closely with the model portfolios. If the FA company changes the composition of the portfolio, the customer’s portfolio will follow accordingly either based on non-discretionary basis or discretionary basis. Some stock brokerage house also offer similar services for stocks.
Interestingly, this is actually the hype in the Robo-Adviser which have created a storm in the US. End users are able to invests in portfolios managed by robots. But in reality, these are not robots. These are software using algorithms to determine the composition of the portfolios. These algorithms are developed by human beings. Using mathematical models for investment is nothing new. What is important is whether are the algorithms sensible. Check out this article on how this robot did in terms of its performance: The Super Underperforming IQS Performance Fund.
By the way, one of the common way to determine the composition of the portfolio by Robo-Advisers is to use the Efficient Frontier. I have a lot of problem with this algorithm which brings me to the next point…
Robo-Efficient Frontier Calculator
I have a software that is able to calculate the efficient frontier. I did not develop this software though. But I have never used it on actual cases because it is not useful. A portfolio is said to be on the efficient frontier when it has the highest expected return given a specific risk. Or, it has the lowest risk given the desired expected return.
The software uses historical performance to determine the correlations and expected returns. The problem with the efficient frontier is the tendency to churn out portfolios that are highly concentrated. The software effectively overweight the best performing asset classes. In other words, it is telling investors to chase after the highest performing assets. Sounds very human to me ! No pun intended.
What the Robo-Efficient Frontier Calculator cannot do: It cannot determine whether the statistical properties remain stationary and hence relying on historical data is like the blind leading the blind. The software lacks common sense when it churn out concentrated portfolios.
I subscribe to a software that aggregates my clients’ portfolios from various sources like: ILPs bought from other agents, bonds purchased from private banks, unit trusts from me and ETFs purchased from my Robo-ETF-RSP. The software cannot support stocks and bonds. But I’ve overcome this by writing a bridging software (see Robo-Bond-Charter). So these limitations are of no concern. By using the software, I have a helicopter view of clients’ entire portfolios and whether are they over exposed to certain assets. I am also able to determine what are their top 10 bonds/stocks holdings. Obviously, the software can plot time weighted average return charts which are important to demonstrate whether I did a good or bad job!
What the software cannot do: It cannot determine whether the existing portfolio is suitable since it does not understand the end clients as well as I do.
I subscribed to this software one year ago. It is an amazing software that is capable of handling complex credit management which I usually spent hours on it. It is able to take into consideration of the customer’s cash, CPF, investments, valuation of the property and residency. Built into the software has the latest MAS endless rules on curbing property speculation. The software is also able to provide loan packages from all banks in Singapore.
What the Robo-Credit Management cannot do: It is not able to determine whether the customer can really afford the property purchase as it uses the TDSR and MSR as the measurement of affordability. Moreover, it churns out a report that is poorly presented, conflicting and confusing because the developer is a software developer who is not prepared to learn about good presentation. The software has full of bugs too. I reported most of the bugs and it has been fixed. But the main problem is the presentation which the developer refuses to improve. Nevertheless, I use its output and created my own presentation.
Both IGP and Navigator are capable of allowing its customers to drawdown a fixed amount regularly from their portfolios. I have used these capabilities and it is briefly mentioned in my article: An example of converting capital gains to an income stream.
What the Robo-Portfolio-Drawdown cannot do: The tool cannot determine the suitability of the portfolio neither can it decide how much to drawdown. A human is still required for that kind of decision.
So what’s next?
I have a Robo-Personal Assistance to help me with my paper work. No, this is not Virtual PA but really 100% software. Some of these software is purchased but most are developed by me. I also have a Robo-Firewall that is protecting this website against from hackers. As I typed, there are probably 10 hackers trying to hack this website. All these hackers are bots itself. I use a combination of commercial grade firewall and my own software to prevent my website from being hacked.
In terms of technology, I am very comfortable and familiar. I am really excited about the development of FinTech. MAS is 10 years late. But it is better late than never.
If you have read until here, you will have realized that technology and software are pretty good in algorithmic functions and the ability to handle large data set. It is also good in problem solving provided the problem can be well defined. But technology and software hates fussy-ness. In many financial planning matters I handle, there are lots of fussy-ness. This is not unexpected. If the problem is simple and well defined, clients would have DIY themselves instead of approaching me.
A Robo-Adviser cannot read in between the lines. When a customer says “I am interested in financial planning”, she could be saying “How can I protect myself if my husband divorce me.” A skilled financial practitioner will immediately pick this one by the tone of the voice and body language. This happens a lot of times when my clients tell me something which actually means another. It been said that 80% of the human communication is by the body language.
A Robo-Adviser cannot interpret silence. But any human financial planner will know that when a client remains silent, he is either objecting or confused.
A Robo-Adviser cannot follow-up and persuade a client to implement the recommended solutions. This is important – unless the recommendation is implemented, the client’s situation did not change. I have a CRM which tells me when to follow-up and how many times to follow-up. In this manner, I don’t sound too irritating but the follow-up is essential. A robot cannot follow-up like a human can. By the way, I understand that the Direct Purchase Insurance (DPI) was a failure. Actually, I was surprised by its failure. In the first half of 2016, only 435 DPI policies were sold. But later on I realized why the DPI was a failure. In the real world, nobody gets up to buy insurance. A human insurance adviser is required to persuade a client to take the necessary action to purchase. Similarly, nobody gets up to have a financial plan. However, most property purchases are made ‘suddenly’ without much persuasion. That is why, DIY portals related to property purchases are highly successful.
A Robo-Adviser cannot on its own determine suitable products to recommend. It still requires a team of human developers to feed it with data and products. The human developers have to perform due diligent on the products first before allowing the Robo-Adviser to recommend. The question is how trustworthy are these humans behind the scene?
A Robo-Adviser will find difficulty if the clients have competing needs. Its algorithm need to factor in allowance for compromise. The way it is done is usually by assigning weights to the competing goals so that the algorithm can optimize its recommendations. The problem is this: who will determine these weights? Obviously, these weights are to be assigned by a human. So we are back to the square one – who is this human and why should we trust him?
Financial planning itself cannot be handle by a robot. The financial planner need to understand human nature which is inherently fussy. But parts of the financial plan and some of the solutions in the plan can be handle by robots. I have been using such an approach for the last 10 years and moving ahead, it will continue to be so.
Like this article? Subscribe to my newsletter below for more.