Quant Foundry – Labs Division
Recommender Systems In Finance
Dr. Chris Cormack and David K Kelly
Introduction
Recommender systems are now commonplace in across many retail platforms outside of finance. Many companies have been applying algorithms to highlight similar purchases from other customers for quite a while. Most organisations build these systems on rating frameworks where users rank for example a film.
Within areas of finance such as private wealth advisory the clients typically do not rate the products or solutions as they vote with their feet.
For an investment advisor looking to engage a new client or provide better-guided advice to their current clients providing meaningful solutions is key to the success of their business. The intended interaction places demands on the types of model that the advisor can use require not only valuable advice but also an explanation of the drivers of the recommendation.
Current Practice
In many recommender systems within the financial industry, only positive ratings are available when the client has bought an investment product. There are no negative ratings that make traditional recommender systems unsuitable.
Current state-of-the-art approaches to generate recommendations leverage off matrix factorisation that are proving to be very difficult to interpret by the advisor and their client. This lack of interpretability makes such models unsuitable in an industry where regulations stipulate the customer has a right to an explanation for any financial product provided.
Any model needs to have the right combination of reliability and transparency is critical for any advisory function to build and maintain its client engagement.
Quant Foundry Solution
The Quant Labs research division of the Quant Foundry has developed a sophisticated, scalable generative model over clients and products that allows interpretable recommendations. The model is linear in the problem size so will scale well with new clients and products. The outcome of the model is to provide weighted probability recommendations for types of client and product
The team at Quant Labs has designed the model to assist sales and marketing teams to build an enhanced customer engagement strategies and to provide a means of ensuring sales teams can obtain rapid insight into the recommendations that the advisor can easily communicate to their clients. The model can be customised to allow business to build sophisticated categorisations of both their clients and the products on offer.
The Quant Foundry recommender system offers sales, marketing and advisory teams a way to scale up their business in terms of both clients and products by allowing them to focus on more productive engagements.
Furthermore, the transparency of the recommendation systems allows those teams to scale while ensuring consistency of client advice – a vital component of any advisory business.
Figure 1: User / Rating Grid (Black selected product, White Unselected product)
Figure 2: Recommendations based on the Algorithm across clients and product.
Quant Foundry
Dr. Chris Cormack and David K Kelly formed Quant Foundry in 2018 with the understanding that our success depended on our ability to attract and retain talent.
Through our unique control framework, each senior quant has access to our collaborative and multi-sector network of problem solvers that support and challenge solution choices as well as a quality assurance “check-in, deliver, check-out” framework that enforces consistent, high-quality delivery.
Rather than have a costly “bench”, our Quant Labs division runs several research-based and highly creative initiatives inside and outside of finance where our senior practitioners expand their knowledge base as well as supervising our junior Ph.D. talent pool. Where all members of the Quant Foundry can contribute to new applications.
Please contact us via our website: https://quantfoundry.com/contact-us/for further information.
March 2019