Thought Leadership November 26th 2016

Will Artificial Intelligence over-perform asset managers?

Asset Management is a profession of expertise and knowledge, with many qualitative factors to take into account in a complex decision process.

So, the dream of a fully automated asset manager that outperforms the best experts might just remain a dream for years (or more!). However, the day-to-day job of an asset manager surely contains some tasks that can be made easier with machine learning.

The first step to take in this evolution is the pre-analysis of a massive set of unstructured data, from various documents and sources, like financial reports, official documents from regulators, Reuters news and so on. Many technologies are emerging around this topic, called “cognitive computing” or “insight search”. They allow not only to search within a massive dataset but, well parameterized, they enable enrichment of that data by creating links between documents or entities and by selecting the most interesting ones with machine learning approaches.

To go a step further, a lot of financial companies are working on automatic extraction of relevant information on recurring sources. For an asset manager, the focus could be on automatic extraction of data within a prospectus, circulars of regulators, and financial reports to automatically identify relevant data or changes vs. previous document releases.

After this partially automatized process of data selection from documents, the second step is to create automatized synthesis or summaries. Synthesis algorithms have evolved quickly over the recent years and have become more and more relevant each day.

All these machine learning approaches will definitely ease the day-to-day job of an asset manager and help funds create new sub-funds or take decisions on their investment strategies. This really is the first step of cooperation between humans and machines in this profession, and clearly the most achievable step for asset management. And it will help asset managers focus on what has the most added-value, taking decisions based on information that is already filtered and prepared.

The second step, where we see FinTech firms focusing, is to try to replace asset managers with data scientists. Although many of these companies are more marketing spiel than fully disruptive approaches. One route that seems relevant has been developed by Numer.ai, which chose to build their own investment funds with data scientists that are not financial experts. Their technical approach is very similar to the one that we currently use in all our own machine learning projects; they create many small models, very efficient in a specific small task, and they merge all these small models into a metamodel that benefits from all these optimized outputs. This validates their technical relevance and pragmatism, and differentiates them from a lot of firms that trumpet large neural network modeling approaches that apparently answer everything, which is clearly a pure marketing position and should be considered with highly suspicion.

As everyone can see, the profession is already changing and it’s not going to stop.

We face it at Ekimetrics, as more and more actors in the investment fund ecosystem (asset management companies, administrative agents, auditors…) are looking at pure data science companies like ours to accompany them on those topics, even if our business focus is not only on finance.

We would recommend different actions for asset managers to cope with this evolution:

·         hire talented data scientists to help them identify the difference between right and wrong within FinTech firms, or help them adapt some models to their own businesses

·         remain open-minded to technological evolution and invest in Artificial Intelligence

·         think about their own competitive advantage in a more data-driven world; what proprietary data do they have? How could it be enriched by others? What should their focus of expertise be to remain at the forefront of the industry? How do you access and treat your data, manage risks, and detect emerging trends?

A critical and clear mind on this topic is definitely the best way to clear away the hype and hyperbole, allowing you and your business to take the best steps and ride the wave of data.

Thought Leadership July 10th 2019

McDonald's France : business strategy, behavioural analysis & big data

Whether I’m in class or talking to friends, it is often difficult to explain to them what the meta-data in their daily lives can do. It is easy to imagine how big data is used in scientific calculations of meteorological models or for managing transport, but it is harder to see how it can be used in our daily lives.
Thought Leadership June 19th 2019

How to combine business approaches, advanced statistics, and technology?

It is sometimes difficult to create a link between highly sophisticated statistical approaches and business reality. The transformation of data into value is an art that needs to combine three different pillars: business understanding, advanced statistics and technology. This triangular approach needs to be adopted by all machine learning projects.