Patent Application
20220076887
This invention relates to personalizing financial portfolios based on investor behavior and experience.
Investment portfolios are usually haphazardly arranged and sometimes tailored to specific investor profiles but there are little, if any, highly organized methods of specifically tailoring investment financial portfolios based on the build-up of optimal product allocations based on specific answers to questionnaires provided to investors and the formulation of an investment model and computer program based thereon.
This invention comprises a method and system for providing a personalized investment portfolio for computerized implementation and based on the following method steps and criteria.
The present invention further comprises a data processing system method of establishing a personalized financial portfolio based on investor behavior and experience.
The invention comprises several subsequent steps to build the optimal product allocation for an investor or client.
A first step comprises establishing a client profile, based on a series of questions regarding his (the term “his” is not gender specific and is used herein for convenience) behavior, covering both his daily life and his investment approach and experience (behavioral profile).
A second step comprises a model to determine the optimal asset class allocation for each client profile, covering a wide range of assets, including real estate, insurance, arts and traditional financial asset classes (termed, holistic asset allocation).
A third and final step of the method comprises a model of establishing a personalized ranking of financial investment products for an investor, based on product characteristics and investor profile (termed best fit investment).
The entire method operates within a telecommunication network architecture of the client-server type which is dynamically updated based on market trends and product performance and evolution.
The invention also comprises a computer program on a non-transitory medium, which can be loaded directly into a working memory of a server processing system, to implement the steps of the methods of the invention, under the control of the server system, when the program is run on the server processing system.
The invention also comprises a computer program product comprising a substrate which can be read by a computer, on which substrate the program is stored. It also comprises a program which can be loaded directly into a working memory of a server processing system to implement the steps of the methods of the invention when the program is run on the server processing system.
The invention further comprises a data-processing system for establishing a personalized ranking of financial investment products for an investor, as well as a data-processing system for establishing a personalized composition of a portfolio of shares in financial and non-financial assets for an investor.
The following is an expanded and more detailed enumeration of intermediate steps of the present method (prefaced by a descriptor of “BestFit” as shorthand notation for the present method and system):
The following ten steps outline non-limiting procedures, mechanisms, and actionable insights for the recipients of BestFit generated client information and which includes multi-disciplinary behavioral science, neuroeconomics and psychometrics, to explore the invisible universe of the unconscious part of the brain with indirect, non-invasive questions.
BestFit selects a set of Predictors (i.e. character traits and personal preferences) from its Predictor Universe which is deemed to be most important to the specific business segment BestFit is being deployed in. Based on the priorities of each business using BestFit, a sub-set of predictors is chosen. This defines the custom-tailored basis of the ultimate user profile.
A team of behavioral science specialists and commercial experts suggest charming, non-invasive, non-offensive questions.
BestFit employs 3 distinct question types:
The basis of BestFit is the indirect approach in gaining insights of the user's personality, by creating charming, non-invasive, inoffensive and discrete questions. This creates user engagement and avoids gaming of answers (i.e. the deliberate distortion of answers based on the image the person wants to create of him-/herself rather than his/her true personality).
BestFit's algorithms are assigned to each single answer. A Predictor based user profile can therefore be established with as little as one answered question. The answering of additional questions is supplemental and improves the quality of the Predictor results.
Psychometrics determines how many answers are required to assure a reliable viability of a Predictor variable.
Sample personality traits focus on:
Psychological interpretation of a user leads to highly differentiated communication
Profiled by BestFit's technology, a company's customer base can be effectively and accurately clustered and segmented. This is especially helpful for designing effective marketing campaigns (e.g. lead lists, top 20% least fee sensitive), develop demand driven products and services, and maximize pricing efficiencies (e.g. dynamic pricing). BestFit data intelligence allows companies to create profile-based model personae on true data sets.
Profiling data can be either stored in the cloud or on A server of the client. Data can be extracted and integrated with other information sources.
Step 9 Use of data for action items include
The following discussion, examples and drawing provide additional explanation for the present invention with the drawings, in which:
The method and system herein comprises steps involving several parts:
The process herein starts with collecting information about the client, in order to define his or her investment profile. The main data source is a questionnaire which comprises a set of questions regarding habits, lifestyle and personality. A sample version thereof includes 37 salient questions (based on a Eurocentric investor), as follows. Variations are possible but with the questions centering around similar themes:
A) Sample Questionnaire
A—You will read a set of sentences regarding your personality, please give a score on how much you agree with them on a scale from very high to very low (VH, H, M, L, VL):
B—You will read a set of sentences regarding your personality, please give a score if True or False:
After collecting information from the questionnaire and alternative data sources, the algorithm has to turn this information into a synthetic value, which represents the profile of the client.
There are several ways to do so, and in one embodiment a multidimensional profile approach is used. The output of the algorithm is, in fact, an array of values, including multiple different parameters. In a version, the parameters are the following:
Once an answer is chosen, the matrix is reduced by eliminating the rows of the not chosen answers. When all questions have been answered, the total sum of each column will provide a value (lower than 1) for each level of the output parameters. The algorithm picks the values with the higher value to assign the final scoring to each parameter.
The Definition of the 12 Main Profiles
Once the set of scores is defined, the algorithm picks one of twelve main profiles, which were defined statistically after testing the questionnaire on a set of 524 subjects. These profiles individuate twelve typical
The algorithm may assign different values to each of these parameters. Most of them can have a low or high value, some can have one of two values (such as Anxious/Relaxed), Risk Propensity has a wider scale from low to high and Time Horizon a scale from short to long.
c) The Scoring System
Each answer can affect several parameters. This is done through a scoring matrix, with rows corresponding to all the possible answers to the questionnaire and columns corresponding to all the values of the output parameters. Every element of the scoring matrix is either a zero or a fraction.
personas with specific behaviour characteristics. The personas can be matched with the already existing risk categories of the bank, or kept as a standalone profiling. Each profile gets access to a subset of financial instruments (selected as further explained below) which are suitable with its risk profile, time horizon and behaviour characterization.
D) The Model Portfolios
In order to estimate the strategic model portfolios a model is used which is able to combine Bayesian models and heuristic models. Specifically, it uses the Black-Litterman model to estimate the expected returns, and the intra-group boundaries for the subsequent optimization. This choice was made in order to avoid the typical limits of the Markowitz optimization, i.e. instability of portfolios, high sensitivity to input errors, unreasonable corner portfolios.
The Black-Litterman model starts from the market-neutral portfolio. Generally, it is the portfolio that replicates the market capitalization weights of the chosen asset classes. The main idea is, in fact, that an investor without any views about the future evolution of the markets should rationally replicate the market neutral portfolio.
Through a reverse optimization process, the model estimates the expected returns that would justify the composition of the market neutral portfolio. The following formula synthetizes the process:
Π=rf+(λΣ)*WMN
where:
Π is the equilibrium expected returns vector;
rf is the risk free rate (on an annual basis);
λ is the risk aversion coefficient: it indicates the extra return that the investor needs to accept one extra unit of risk;
Σ: is the variance-covariance matrix of historical returns of the benchmarks;
WMN is the vector of the market capitalization weights.
Using a risk aversion coefficient λ, in conjunction with a Correlation Matrix provides Equilibrium expected returns which leads to a Prior returns distribution:
The value of A can be estimated through the following:
where:
The expected returns just computed can then be adjusted according to a series of views given by market analysts.
The views have to be expressed on a specific time horizon and can be of various types:
The views have to be given together with a confidence interval, expressed as a measure of how much the analyst believes in the given view.
The view Q in conjunction with the views uncertainty Ω provides a views distribution:
The two distributions so calculated have to be combined to achieve the posterior returns distribution. This can be done through the formula:
R
BL=[(τΣ)−1+PrΩ−1P]−1[(τΣ)−1Π+PrΩ−1Q]
where:
Combining the Prior Returns distribution
with the Views
Distribution
provides a Posterior returns distribution
Once we have the expected returns and the variance-covariance matrix, it is possible to proceed to the optimization of the efficient frontier, following the standard Markowitz model, but with the adding of intra-groups limitations. Specifically, the model uses:
E) The Fund Selector
The mutual funds can be ranked according to the main valuation parameters especially:
According to the client profile, the algorithm changes the weights of the product parameters, which are used to build a product ranking.
For example, for a more risk tolerant client, the performance parameters will have a greater weight than the risk parameters, while for a short time horizon-client the maximum drawdown and the recovery time will be taken into account with a greater weight.
The outcome of the algorithm is not only a synthesis of whole questionnaire, but each question has its own single role in the calculation. For example, the answer to the question “I always check and compare prices before buying something” gives information about whether the fees of a financial product are a very relevant element for the client or not and their weighting in the ranking calculation can be modified accordingly.
F) Personalized Reporting
The output of the algorithm can be used not only to optimize the client's financial portfolio, but also to deeply personalize the overall commercial service that he is getting.
The present algorithm is made to give to the client what he really wants; even with the periodic reporting of the portfolio performance, not all clients are the same.
In fact, according to the client's profile, the algorithm sets the periodicity of the report (weekly, monthly, quarterly) and which content the report should focus on. A customer with a greater level of Autonomy could appreciate a less frequent update, while a client with a smaller level of Autonomy might want to be ensured every week of the performance of his portfolio. A client who likes to take decisions according on a wide data check base, will get a more detailed report.
G) Alert Triggering Through Bank Divisions (as Applicable)
The present algorithm is moving in three directions, one towards the optimization of the client portfolio, the second through the personalization of the service, and the third through notifications to different bank divisions.
In fact, at every step, the algorithm can trigger some “opportunity alerts” to different offices of the bank to let them interact with the customer. For example, if the questionnaire underlines that a client is very risk oriented and entrepreneurial, the private equity division will be alerted to contact him and present more investments.
The questionnaire contains several questions related to shopping behaviour. Determined answers trigger the alert of offering a credit card service, others to offer a loan, a sports insurance, etc.
In the above sample, 500 interviews representative of an Italian population with a bank account and owning financial investments or wishing to do so in the future.
In evaluating the results, results on bases smaller than 80 cases were considered to be less reliable and results on bases small tha 25 were considered only as a qualitative.
The tables in the Figures which are shown as being circled highlight results with such circling which are statistically significant (95% level probability). All other results are considered in line with the total sample ones.
Based on cultural considerations with respect to an Italian mind set, a mark of 7 is a good result and a mark of 8 is an extraordinary test result.
With respect to the drawings,
As seen from the tabulated results in
The tabulated results in
It is understood that the above is illustrative of the present invention and that changes in questions, formulations, analysis, applications other than for financial matters and the like may be made without departing from the scope of the invention as defined in the following claims.
This application is a continuation of U.S. Ser. No. 15/938,739, filed Mar. 28, 2018 which is a continuation in part of U.S. Ser. No. 15/708,395, filed Sep. 19, 2017 which is a continuation-in part of non-provisional patent application U.S. Ser. No. 15/260,690, filed Sep. 9, 2016 with priority to provisional patent application U.S. Ser. No. 62/216,315, filed on Sep. 9, 2015, the disclosures of which are incorporated herein in their entirety by reference thereto.
| Number | Date | Country | |
|---|---|---|---|
| 62216315 | Sep 2015 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15938730 | Mar 2018 | US |
| Child | 17315935 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 15708395 | Sep 2017 | US |
| Child | 15938730 | US | |
| Parent | 15260690 | Sep 2016 | US |
| Child | 15708395 | US |
