We recently released a mesmerizing white paper on AI in the banking sector. Inside, you'll find use cases, methodology, and more. We also took it a bit further and asked a few experts to share their key takeaways on financial services and analytics.
.png?width=188&name=LOGO-Blue-DME-PNG-3%20(1).png "Blue-DME-logo.png"){kind=logo}
Today, we want to give the floor to Julien Cabot, CEO at BlueDME, a data consulting company that happens to work with us for banking and insurance customers. The below insights come directly from Julien.
Since 2011, I have had the opportunity to participate in a number of data lab projects in banking and insurance sectors. These years of experience have taught me that, in order to succeed, it is necessary to address three major challenges that go beyond the simple preparation of a data lab project.
Which challenges am I talking about?
1. Focusing on Where the Added Value Lies
Accessing and qualifying data are key factors of success since, without data, no approach is possible, however brilliant the idea may be. Moreover, 60% to 70% of the effort that goes into a project is linked to the qualification and preparation of data, necessary yet arduous tasks. Given the hourly cost of data scientists, effectiveness and productivity in the processes of collecting, researching, and preparing data are fundamental. Any tool and approach that helps to accelerate this phase quickly proves worthwhile.
2. Deploying Data Projects in the Real World
The capacity to integrate results from data science work into operational processes, whether in the form of pre-calculated indicators or real-time prediction models, is essential. Analytical study should lead to new transformations of data and “mathematical objects” — in the form of prediction models — that operational information systems are able to integrate. However, it is often also necessary to look to the study’s generation of real value.
3. Generating Value From Data Products
Measuring the return on investment (ROI) of data science projects is important. But due to the exploratory nature of some data projects, the exact estimation of the ROI unit of each project is particularly complex. But looking at the ROI of projects together as a whole enables those projects that generate concrete ROI to finance experiments that procure a more limited result, as is the case with private equity investment strategies.
Problem Solved?
Luckily, in banking and insurance, innovative approaches are already being implemented in order to address these three challenges.
Idea 1. A Search Engine of Data Dedicated to Professional Teams
A search engine for data dedicated to professional teams (from those responsible for statistical studies to risk teams working with a data lake as well as alongside a data lab) has many benefits. It facilitates the realization of a catalog of internal and external data, allowing for the exploration of this data adapted to different tasks and the sharing of the most useful data within the banking-insurance field. The underlying concept of Blue DME’s Data Exchange platform is the development of a collaborative approach to using data lake data, enriched by a data lab, between two professional domains.
Idea 2. Representation Through Web Services From New Generations of Predictive Models
Representation through web services from new generations of predictive models — random forest, gradient boosting and more — thanks to modern solutions in data science (such as Dataiku DSS) which are boosting efficiency and results. The days of manually configuring logistic regression are coming to an end. The direct representation of binary prediction models enables the development of these advanced models to be simplified and, above all, enhancing their performance.
Idea 3. Logics of Successful Internal and External Monetization
The calculation of ROI produced by the results of the predictive model in relation to the previous situation is limited to internal optimizations. The idea of developing new services based on data is becoming more and more important, making new high-margin revenues possible.
In banking and insurance, data lab projects are often the first step in a larger process, transforming the organization into a data-driven business model that goes much further than client awareness or risks.