INTERVIEW: Feedzai CEO, Nuno Sebastiao

Fighting fraud


Tell us about yourself
My training background is in software engineering. I did a lot of work for seven years in Germany at the European space agency in charge of the simulations group. We experimented with simulating different levels of fidelity in a spacecraft. This was all to do with real-time, building computer models representing the spacecraft at extremely low latencies. It essentially involved ingesting a lot of data to make sense of this. After completing my MBA in London I began thinking more about real-time and the changing aspects of data processing.

How was Feedzai found?
I got together with my two co-founders Paulo Marques and Pedro Bizarro and considered building a technology that can crunch a lot of data very fast and remove the difference between real-time data processing and more patched stuff. So we built this technology that we have today which essentially allows us to process a lot of data in real time but also enriching and contextualising it with historical data. Moreover, we have machine learning models that infer knowledge about that data.

Where did you go from there?
After building the platform, we began to investigate what markets we could go in to and what kind of problems we could solve with this technology. We had traction in a number of markets from telecoms to utility but by far where we had the most traction was in the payments space. This involved looking at what’s going on in the networks of merchants, issuers and acquirers and enriching the transaction data. 

The data you see in a transaction is actually quite small when considering the various agents, which makes it hard to make an inference based on that snippet. But what if we can look at this by contextualising it and correlating it with everything we might know about that piece of data? This involves correlating with user or merchant information and what we are seeing in the network at that particular time.

We do multi-dimensional analytics down to the level of the user combined with machine learning to stop fraud in milliseconds while it’s being committed. We do not distinguish between the real-time and offline world as most organisations do, so we are enabled to retrain and update our models as information comes in.

How do you look at data?
Feedzai pulse is a technology that is able to correlate the real-time incoming data with all of the history we might have. We call it profiling. When a transaction comes in, we update thousands of metrics in order to analyse how the behaviour compares with six to nine months history we may have.

People change, therefore data that is over a year old is not necessarily helpful and indicative of future behaviour. We consider more recent activity and keep a huge number of metrics which give a greater insight and help us to understand if something is out of place and anomalous or fraudulent.

Due to the size of the data volumes we are talking about, potential clients are reluctant to put the data outside their walls – we are talking over 600GB a day. So we set up our technology at their data centres for a month or two to show them the results with their own data. Based on this we go into a contractual or commercial agreement.

How has Feedzai progressed?
Since building the technology in 2011 we went into production in 2012 with our first client. Today we are processing and looking at over EUR85bn in production on an analysed basis. On any given day we process anything between USD500m to USD1bn and will stop USD2m to USD15m in fraud. It’s a staggering amount when considering 80% fraud detection rates.

Could you tell us about Feedzai from a security and fraud prevention perspective?
We see a lot of flash attacks that come through out of nowhere. We are able to catch them by being able to see the transactions being made in real time. We see these attacks more in digital only transactions such as online subscription and ecommerce services. We aren’t talking about goods that are being physically shipped so it becomes harder as digital goods perish pretty fast. Time is therefore essential in order to prevent such attacks.

What is Feedzai’s strategic plan going forward?
We are in the process of bringing some clients on board in the US. The idea is to branch out to other geographical regions.

Additionally, we are launching a product that is targeted at merchants. It will be an API that merchants can tap into and effectively get a score on their transactions before they are sent for processing. With us they do not have to incur any chargeback or processing costs at all. This is something that we are putting out there for smaller merchants for free for the time being. They are paying us back with their data.

The more data we have the better it is! We need to collect as much data as possible from different points in the chain. The closer we are to merchants the better the information is. If you are just with the processor you are missing out.

The problem we are seeing is making sure organisations understand the importance of sharing data. For example, merchants may not want to share data with acquirers.  By making a push to share more, we can correlate data from different places to gain a better understanding of the bigger picture. 

Related reading

Leave a comment