NETFLIX FOR INVESTMENTS – AUTOMATICALLY RECOMMENDING YOUR FUTURE SAVINGS

netflix of investments

Kattintson ide a magyar verzióért.

Nowadays, we are not surprised when we receive relevant suggestions during online shopping or when looking for movies. Sometimes we are amazed at an offer, sometimes we’re not that interested, but more often than not, it may shock us how well the machine knows us. We want to highlight some of the most relevant features of the recommendation engines and show how can it be relevant in our field, in the investment market.

Recommendation systems use different information sources to make personalized suggestions. This information can be anything, which has any connection with us, such as our past activities, the devices we use, our location, our likes and dislikes as well as more personal data, such as age or gender.

The most well-known areas for recommendations are in e-commerce, social media and entertainment: online stores use product recommendation and some of them – like Amazaon and Ebay – apply AI-based solutions to make their proposals even more personalised. Facebook and Instagram generate plenty of suggestions to offer us attractive content. Furthermore, Netflix and Spotify users mostly watch movies and listen to music only based on what was offered to them. There are differences, but the goal is the same everywhere: Mass customisation. They want to provide the best personalized service because customers demand it.

If it works so well in other industries, why not use it in an area like investing?

Recommendation engines are algorithms that suggest products and services for users based on data analysis. Massive data sets are used, supported with artificial intelligence.

There are many techniques for recommendation engines to work, but the two mains are:

  • Content-based filtering: based on a single user’s interactions and preferences. Recommendations like ‘products like this’, are typical instances of this type of approach.
  • Collaborative filtering: This approach makes recommendations based on interactions by other users with similar tastes or situations. ‘Next buy’ recommendations are the typical usage.
  • + Hybrid method of mixing the previous two.

In practise this means that in the first case, the products are grouped according to their characteristics, and when someone is interested in a product, the system makes additional recommendations from the group of the product already selected. In the other case, users are grouped based on their common characteristics and properties, and then the system recommends a product favoured / purchased by one member of the user group to the other, regardless of whether she is interested in similar products.

What makes an AI based engine better than a simple recommendation system is that it connects personality traits and qualities that we cannot see or notice from the outside. The algorithms work in a “Black Box” and give a result at the end.

There are some common traits in connection with the most known examples such as e-commerce based on user feedback and ratings. It can be said that products are time-constant (t-shirt to t-shirt, pop music to pop music) and a wrong suggestion-experience does not have a profound effect on the lives of consumers. These are important to mention because these characteristics are not true in the field of investment. Clients typically do not rate their financial products explicitly; certain characteristics of stocks can change quickly, and recommendations can have a huge financial impact on the clients’ life. So, to be able to recommend a stock to buy, engines must work in a different way. Of course the basic principles of the afore mentioned methodologies can still be applied.

So, what are the most important elements of an investment recommendation? The recommendation must adhere to the clients’ risk profile and investment goals, which narrows the options considerably. Afterwards we have to do profiling tasks based on the investors’ various social and psychological characteristics. In addition to this, building on historical transaction data is particularly important.

This area also has other specialties. When buying a new shoe, we typically don’t bother to provide personal information or fill out questionnaires to get a recommendation. However, when we want to invest to increase our wealth, we devote more time to make the right decision. For example, we are much more willing to complete a test to share our preferences with the system for the best possible recommendation.

So usually, these recommendation engines operate using risk profiles, transaction histories, personal information, characteristics, and personal beliefs. It is crucial that only factors that are relevant to investment decisions are selected, and neither too little nor too much data is entered.

Comparing the selection of a shoe with making an investment decision, there are more serious preference factors for the latter, which makes the system much more complex. At the same time, in the case of investments, the user has a greater willingness to cooperate and share personal information, which in turn helps the accuracy of the recommendation engine.

These all may cause challenges, but recommender systems can have many benefits. For example, it can help for an individual investor to select the appropriate enterprise to invest infor advisors to manage their clients’ portfolios more efficiently or even in replacing advisors with an automated system which can handle portfolio management by itself.

Wondering what all this means in practice? When browsing the internet, it is easy to find working examples that are specifically used in the field of investment, but no doubt the Netflix-level breakthrough has not arrived yet. But the sector is working on it and Dorsum is also developing its own technology to get a more accurate picture of the preferences of those who want to invest, going beyond the aspects indicated by MiFID2. In addition to the development of the product recommendation engine, we also conducted a large sample questionnaire survey. As a result, we found previously unidentified parameters, allowing us to create even more accurate customer groups. The project is completed, so we look forward to revealing more details about it.

Dorsum is committed to further R&D on the future of profiling and recommendation engines as part of tender 018-1.3.1-VKE-2018-00007 „Automated and verified evaluation and categorisation of financial products and service providers for secure, personalised investment services”.

Have a question?

If you are interested in our solutions, contact us via the form below.

Dorsum is proud to benefit from the European Commission’s Brexit Adjusment Reserve.
Learn more about our project aimed at aligning our Wealth Management Suite to the UK market. (Hungarian content)