Picture of Frederrick Hamann
Frederrick Hamann
CCO

How does Subaio detect recurring payments?

Now we’re going to talk about recurring payments and, more specifically, we’re going to answer the question – how does Subaio detect recurring payments?

And the way that we detect recurring payments is of course through data. So we need a lot of data to do this and we work on transactional data. And that type of data we get from the bank or we get them also straight from the aggregators.

And this means that we can work on different types of payment data as well. We can work on credit card transfers, we can work on direct debits or even account to account based transfers and all of these different types of payments we can use in the algorithm that we have, to detect these different recurring payments.

But the data has to have a certain granularity. It has to have a certain amount of details for us to work on it. And what we basically need are three things. So, first of all, we need the merchant name. Secondly, we need the amount. And then we also need the currency. We need a couple of more points as well, but these are basically what we need to be able to make the system work. And this is the type of data that we then put into our algorithm.

A good example would be how we go into a new market. Once we go into a new market this is what happens; So the data from that new market will go into our algorithm. And that means that this algorithm then looks for patterns. It looks for frequency, it looks after amounts and also merchants, and these different patterns, also using cluster analysis. These different patterns then spit out different boxes. So that means that after the algorithm has gone through all the data from that market, it spits out the parts that are recurring, it spits out those that are subscriptions, and then also the ones that are the one-offs.

And what we basically are focusing on here is the recurring payments and the subscriptions. So we don’t need the one-offs so we just throw them aside. It’s really the recurring payments and the subscriptions that we’re looking at. And the difference between those two things is basically that these subscriptions are the ones that are easy to cancel, whereas the recurring payments can be financial products, like mortgages, insurance and so on.

So all these different parts are being showed to the end user in the overview. But that’s not all! You can actually go into your transactions, and you can press on the different transactions saying that this one that we didn’t catch, is actually also a recurring payment. And that’s the way that this algorithm learns and becomes better and better, and we’re in need of help in this space, because more and more subscriptions are coming in every day.

This summer a burger chain here in Denmark just launched a subscription, whereas a coffee chain in the UK also launched a subscription. So more and more subscriptions are coming, which means that we need the help from the end users to be able to detect these different recurring payments. 

And that’s the way that we then suddenly have a lot of happy customers, not only because they’re getting the overview of all the different recurring payments, but they’re actually also helping us. So they’re part of the whole system here.

But let’s look a bit more into the numbers behind all of this. Let’s start with 5 billion. That’s the amount of transactions that we’ve gone through so far. So a lot of transactions have gone through our algorithm, which means that we are really becoming better and better at understanding the data, and also going into new markets and very quickly seeing all those different nuances that are in that market. 

That means that it’s not just Netflix and Spotify; the obvious ones that we can see, no. We’re also able to look at the local fitness chain, the local supermarket as well and that’s actually what we’re really proud of. That’s also what the users want – they want to be able to see all the different recurring payments. 

And that means the number you have to remember is 98.7%. That’s our accuracy in detecting these different recurring payments. That’s what the end users can trust when they’re looking at the overview of the different recurring payments, and we only have a 0.044 false positive number which means that 98.7% of the cases we are totally right in finding the recurring payment and showing it to the end user. 

Get a personal demo of how the Subaio solution will work for your bank​

*
*
*
*
*
*