France 
Germany 
Spain 
English 
Belgium 
Italy 
Portugal 
Netherlands 
Published on 16/12/2024
How machine learning applies to split payments
How can you process two million financing requests per month while guaranteeing a response in principle within 150 milliseconds? This is the challenge we face daily at Floa, one of the French leaders in split payments. Between technological innovation and ethical commitment, our company pushes the boundaries of data science to simplify access to financing. Meet Sébastien Robert, Head of Data and AI at Floa, who explains how artificial intelligence is transforming the split payment experience.
As one of the leaders in split payment, how does Floa leverage machine learning to optimise its financing services?
Our position as a major player in BNPL (Buy Now Pay Later) means that we must be on top of technological innovation. Our mission is clear: approve as many financing requests as possible to effectively support our customers and partners, while also managing non-payment and fraud risks.
To achieve this, we have made the strategic choice to use Dataiku, a leading AI and data science solution. This platform allows us to develop particularly efficient machine learning algorithms that calculate, in real time, precise probabilities of different events, and thus allow us to prevent financial risks.
Our approach is based on three fundamental pillars. First, speed: our systems deliver a response in approximately 150 milliseconds, a record time in our industry. Second, simplicity: we minimize the amount of information our customers have to enter while our systems maximize automatic enrichment. Lastly, efficiency: our models have a high standard of performance and guarantee that our infrastructure is highly robust.
In concrete terms, once our models are developed, we deploy them on Dataiku, and they become available via API1 for each funding request. This streamlining allows us to achieve our objective: to offer a seamless experience to our customers and partners, while securing our financing operations.
What concrete benefits do our merchant partners gain from this expertise in Data Science?
This expertise translates into very concrete benefits. First, our increasingly performant machine learning models make it possible to significantly increase the acceptance rate of financing requests using our payment solutions while reducing or even avoiding non-payments.
The major asset for our partners lies in our ability to enrich data in real time. In concrete terms, this makes their lives easier: they only have a minimum of information to provide to us, because our technology intelligently exploits contextual data, our customer knowledge bases and open data2.
Beyond simple performance, our partners also have access to a system that’s highly available and ultra-fast. But what’s particularly important to emphasise is that by choosing Floa, our partners are choosing to work with an ethically responsible company. Our sponsorship of the IA Digne de Confiance Chair of the Bordeaux Universities and our contribution to the work of the IA Act with the France IA Hub bear witness to this. It is a commitment that makes a difference to our relationship with our partners.
And for our customers, what are the benefits?
Great question! Ultimately, all our innovation is about improving the experience of our end customers. Their benefits naturally align with those of our merchant partners.
In concrete terms, our customers have from a triple advantage. First of all, in an increasingly digitalised customer relationship, they obtain an almost instantaneous response in principle to their financing requests4. Furthermore, their journey is considerably simplified: we have reduced to a minimum the information they need to enter manually, which makes the experience much smoother and more pleasant.
But the most important aspect for me is that our customers can be sure that they are dealing with a trusted player, as I’ve just explained. It is this combination of simplicity, speed and trust that makes all the difference in the customer experience.
Faced with the explosion of data, what is the main technical challenge that Floa must overcome on a daily basis?
The performance of an algorithm essentially depends on the quality and quantity of the data that feeds it. At Floa, we process over two million funding requests every month, which represents a considerable volume of data. But the real challenge lies in their diversity: depending on our products, whether it is the duration or the type of financing, our journeys and our partners, we receive very different and often very specific data.
The major challenge is therefore to find the right balance in our algorithmic approach. If we were to limit ourselves to a few generic models, we would certainly have the advantage of exploiting enormous volumes of data, but we would lose precision in terms of our partners’ specificities, our products and our journeys. Conversely, creating a separate model for each scenario would lead us to work with insufficient data volumes, compromising the performance and robustness of our models.
This is why we opted for an intermediate solution with around fifty different models in Dataiku. This approach allows us to maintain this balance between data volume and specific requirements. However, this involves two major technical challenges: being able to instantly select the right algorithm based on the particular parameters of a request and efficiently managing the creation and regular retraining of all these models.
Floa has just won an award for its innovation in MLOps5. Can you tell us more about your technological advances?
Our main innovation is the development of SOUL, an internal tool created via our Dataiku platform, which earned us the title of “Best MLOps Use Case” at the Dataiku Front runner Awards 20243. This is a recognition we are particularly proud of, especially since Dataiku's Head of AI Strategy described our achievement as “incredibly impressive”.
Concretely, SOUL is a continuous monitoring tool that automates and standardises the various stages of transformation and combination of the data required to develop our algorithms. The result is spectacular: we have increased our productivity by 25% while significantly reducing our risk of error.
This award, which follows the Snowflake Data Driver Awards 2023, confirms our technical excellence in the field. But beyond the distinctions, it is above all the automation of the relearning of our models that allows us today to maintain a stable performance over time and to free up time for our teams, who can thus devote themselves to innovative projects with high added value for our customers and partners.
These technological advances are just a first step. Our ambition is to continue pushing the boundaries of innovation in data science, while maintaining our commitment to ethical and responsible AI. The objective remains the same: to offer our customers and partners an ever more efficient, fluid and secure financing experience. It is this vision that guides our teams on a daily basis and makes Floa a recognised leader in the field of split payments.
Do you want to increase your sales while securing your transactions? We have more than 15,000 partners who have placed their trust in us. Join industry leaders by signing up to our split payment solutions. Contact us to find out how Floa can boost your sales performance.
Request a demo
Would you like to try the FLOA offering? Contact our teams for a live demo of our service!
Request a demo
Welcome to the FLOA blog!
Welcome to our blog!
This blog has been designed to support you on a daily basis. We analyse payment trends, share best practices from all our partners, and inspire you with amazing projects! All this at FLOA!
They trust us!
Request a demo
Would you like to try the FLOA offering? Contact our teams for a live demo of our service!
Request a demo
