In our latest Team Insights interview we spoke with Dr Sophie Harbisher, our Data Science Lead. We find out how Kani’s Automation Solution improves data reporting & reconciliation, what Sophie does in the background to support our clients, and how she got into fintech!
Hi! Can you tell us a bit about your background, and your role at Kani Payments?
I’m Dr. Sophie Harbisher, and I’m the Data Science Lead. I’m in charge of modelling data and looking after the reporting side of the Kani Payments platform.
What does Kani Payments do for its clients?
Generally, our clients will work with multiple payment processors. They will have data on different payments that are made by different customers. We take clients’ data and transform it into standard schemas, aggregate it, and give clients an overview of what’s happening with their data. Basically, we do all the hard work of turning that data into pretty pictures and formats they can understand! We’re all about helping our clients get the most out of their data.
Without Kani Payments, clients would need a lot of people using a lot of spreadsheets and trying to get data aggregated across lots of different sources. It’s a heavily resource-intensive process for them. With Kani, because we import all the data and normalise it into a single format, we can produce reports and reconciliations in a matter of just minutes rather than spending multiple days of people’s efforts on that. We save clients huge amounts of time and effort.
What inspired you to get into fintech?
The truth is I had no idea what fintech was! But while I was at university, one of my PhD supervisors was involved in a project with Kani and knew the CEO, Aaron Holmes, and put me in touch with him. I had a conversation with Aaron and he gave me a job. Although I had no idea about fintech, it sounded really interesting. So I thought, let’s do it!
After speaking to Aaron, and learning about finance and fintech, I realised there was such a lack of understanding about payment data. You’d think it’s easy for computers just deal with everything. But behind the scenes, it’s a lot messier. A big part of that is the lack of understanding of terminology. So often, even payment processors who have provided the data don’t necessarily use the correct terminology. And it’s an industry-wide problem. These terms are not very well documented. If you try to aggregate data across all the companies within the industry, their definitions won’t necessarily match your keywords. Part of our day-to-day job is helping clients get around that.
Do your clients request certain outcomes with their data or do you make recommendations based on the data they provide?
As part of the client’s onboarding process, we get an understanding of what data they have and what they want to achieve with that. That includes reconciling bank statements to transactions that their cardholders have made. With other clients, like a travel card issuer, they’ll be dealing with transactions made through third parties like travel agents, multiple varied hotel or flight bookings, with data coming from many sources. Then they also have the invoicing data that they need to match transactions with.
As individuals, we perform reconciliations on a day-to-day basis by looking at our bank statements against our transactions. If you saw you spent £10 in Tesco on your bank statement, but couldn’t remember if you’d shopped in Tesco that month, then you’d be concerned. You’d pore over your bank statement and be racking your brains about whether you’d been in Tesco. If you hadn’t been in Tesco, then you’d be really worried! But obviously, finance and fintech companies need to do that on a much bigger scale, with thousands or millions of transactions, and without being as messy as it normally would be.
With our clients, it’s about helping them identify their goals, and how their data can help them achieve those goals, for instance making sure that their invoices match what Mastercard or Visa is telling them that they owe. Or for a retailer, like a big supermarket chain, we can look at cardholder spending and compare it to previous years, and we’ll be able to predict how well they’ll perform at Christmas. At the moment we’re focused on getting the clients to the point where they’re happy and their data is doing what they need it to do. From that will come lots of opportunities for further data explorations, to uncover more insights and get more value.
How does Kani Payments’ automation solution improve data reporting and reconciliation?
For companies trying to reconcile transactions, they won’t necessarily have a reference that is consistent across two data sources. They might have to look at different fields and features of the data to try and match across them. That includes financial data for reconciling, but also with applications related to fraud and watchlists. A client may have some cardholder customer data, with names and addresses, and they might want to look that up against a watchlist published by the FCA or another regulator – but how do they go about this? Kani Payments provides different tools that clients depend on, such as reconciliation, record matching, and fraud detection, so they can identify behaviour that isn’t quite normal for particular customers.
In order to deliver consistent reporting and insights, it is beneficial for the data to have a standardised format and structure. Right now, we’re involved in quite a large project where we’re changing the database structure behind the scenes. Our clients don’t necessarily see that, but it involves a lot of work for us in terms of writing queries on the data, getting the data out in the correct format, and making sure that we’ve got the structure correct, so that we can build new reports and data insights. Getting that correct allows us the flexibility to produce better reports, discover insights from data, and perform statistical inferences.
What’s next for Kani Payments, and what new solutions can clients look forward to?
We’ve developed record matching for our clients. In an ideal world, if a client had a reference from one data set, and another data set had a similar reference, they could just match them up. But we don’t live in an ideal world! Often, data sets have missing or incomplete references, which produces reconciliation discrepancies. What we are developing is a new solution, probabilistic matching, which looks at all these different references from different data sets, and looks at how similar things are and assesses the likelihood of transactions being a match. Probabilistic record matching allows us to do that with accuracy, even when transactions don’t directly reference each other.
Probabilistic record matching says: “this thing looks very similar to this other thing and overall that means the records are likely to refer to the same thing”. My name is a good example – Harbisher isn’t a very common name! So if you’ve got two Harbishers, and one’s an S and one is a Sophie, they’re likely to be a match. It’s an interesting thing, it gets very messy and geeky, which is my ballpark!
We’re also automating more workflow actions, with a new business intelligence tool, which essentially allows us to make client reporting more interactive. For example, where a report shows an aggregate figure, like a count or a sum, there is drill-down functionality so the client can click through to see all of the records that make up that figure.
We’ve got that automated alert system so the user can go in and check any report where it’s applicable. What the alert will do is at a set time it’ll check the conditions that you’ve set. And if that happens, an email or a Slack message or some other form of communication is sent off. It’s a fully automated process giving clients full control over when that happens and how often they get alerted. It’s important that data flows into our platform so that reconciliation and reporting are accurate and up to date. A great use case for alerts is on file imports – clients can be alerted to any file failures which allows them to investigate any issues.
How do you see the fintech industry evolving?
I’ve been here just over two years now and I feel like I’m fairly new to the industry, in my first adult job! From coming outside of fintech to inside it, I’m working with clients that are more agile and faster-moving than traditional high street banks. Fintech is fast moving in terms of what services are being created and what people are offering.
I think there’s also this fantastic social aspect to fintech. Some fintechs are allowing users to look at the carbon footprints of the transactions they are making whilst others are making it easier for carers to make purchases on behalf of vulnerable people, or providing financial services to those who may have difficulty accessing traditional banking services. That’s a fantastic example of getting data from those payments and seeing what companies can do with it, for social good, environmental actions or financial inclusion.