« Surveys and studies show that 86% of consumers indicate personalization plays a significant role in their purchasing decisions »
The world of finance is a conservative one and usually starts to implement new technologies long after other economic sectors, like tech or e-commerce, have mastered them. I remember going two years ago to a meetup hosted by a major financial institution to hear them talk proudly about their new Proof Of Concept using machine learning (ML), which was able to perform a classification task that all ML manuals teach in their first lesson. Needless to say I was not impressed.
Since then, Machine Learning and IA have become mainstream, and financial institutions are starting to heavily invest in the field. There are more uses cases of machine learning in finance than ever before, thanks to more accessible computing power and popular open-source deep learning tools, such as Google’s Tensorflow.
Millenials, A new kind of investors
To illustrate this techno-lag, If I turn today to my online banking services to choose a fund, I am proposed a long list of financial products with some filtering criterias. This is definitly outdated and ineffective, leading to a frustrated customer that will usually just turn away, and that’s precisely what I did. This exemple also says something about me, the client : in the past decade, clients have changed, too.
Millenial increasingly consult peers and media before acting
Nowadays clients are younger, digital, agile. To quote Deloitte they are « cautious and conservative in regard to financial matters. At the same time, Millenials (as they are named) demand and make use of technological advances. They consider technology an important aspect of financial advice. […] Although 72 % of the millenials describe themselves as self-directed, they also tend to lack financial knowloedge compared to older generations […] Millenial increasingly consult peers and media before acting. »
Recommendation Engines and collaborative filtering
Lili addresses this challenge and transforms it into a business opportunity in a number of ways, most importantly through its social features. E-retailers understood that more than a decade ago, and McKinsey estimated last year that 35 percent of consumer purchases on Amazon come from product recommendations.
Amazon and Netflix’s success rely on the quality of their recommendation engine, and the same can be done in the financial industry, recommending financial products or strategy to private investors.
Recommendations like «users who liked this product also liked this other one » are the result of a now popular algorithm called Collaborative Filtering
Part of the core of Lili’s innovative advising services, Lili’s recommendation engine will provide a personalized user experience based on each investor’s data, such as his Know Your Customer (KYC) passport, his current portfolio models and his navigation history on the website.
Recommendations like «users who liked this product also liked this other one » are the result of a now popular algorithm called Collaborative Filtering pioneered by Amazon a decade ago. At its core it looks at how often two products are present together in the browsing or purchase histories of users. This is the way most recommendation engines work today, and not suprisingly will be applied to Lili’s website.
Dealing with new customers
Personalized recommendations, however, require a great amount of user’s data, which a system does not have in the case of new visitors. This is called « the cold start problem », and can be addressed in many ways, like for instance applying a popularity based strategy, or using the new visitor’s available data (his KYC passport) to assign him to a marketing segment and start to make recommendations according to that segment, until navigational data for this new user flows in.
Fine grained Marketing Customer Segmentation
Clients are generally segmented according to their financial maturity: you don’t address ‘novices’, ‘loners’ and ‘cautious’ clients the same way. But you also have to understand their digital maturity.
While baby boomers have a preference for traditional channels like human interaction and emails, younger generations have embraced newer technologies to interact with their financial institution, requiring omni-channel (mail, chat, bot) communication over multiple devices (smartphone, laptop).
To adress this need, navigational data from the website will have a tremendous value because it will allow to derive the digital profile of each client and assess his digital maturity. This insight will supplement knowledge about financial maturity, to help engage him in the most effective way and on the right channel.
Next gen’ roboadvisers
In the past years, The passive-investing revolution has gradually seen billions flow out of active managers and into index funds or ETFs.
Rushing on the now open path of passive-investment, roboadvisers made a lot of buzz, with a promise that they would take the work done by old-fashioned human financial advisers and use algorithms to perform it instead, for a modest fee. Led in the US by Wealthfront and Betterment, they target a new type of investors: millenials investing passively.
2017 and early 2018 have seen some exhaustion in the model, and these roboadvisers fund start to move towards higher fees, even though capabilities are limited and performance is a bit disappointing
However the roboadviser technology is still a work in progress. As AI and deep learning models become more efficient, the quality of robos is improving, and the latest generation will take decisions based on “profit, risk appetite, and liquidity aspects” of the client.
This kind of roboadviser will help to calibrate Lili’s portfolio models to the goals and risk tolerance of their owners, providing increasingly personalized and valuable advice to Lili’s customers throughout the lifespan of their portfolios models.
And to stress the benefits of the social model of Lili, it is no surprise that Accenture states that even though, “competition, innovation and new technology will dramatically increase robo-advisory capabilities in the near future, personal connections will still remain essential.”, and the key to a good balance to automation is humanized, peer to peer and social features.
PHILIPPE DE CUZEY
ARTIFICIAL INTELLIGENCE Advisor for Lili
Expert noSQL, Data Intelligence, Big Data & Machine/Deep Learning (EU Commission, Amundi)