‘If you realize it’s a hard problem and you treat it like it is, then you might get somewhere’: 5 questions with Bud’s CEO Ed Maslaveckas

Data is a valuable resource for financial institutions but not all firms are able to put it to work. Often FIs lack the wherewithal needed to turn rows of data into actionable insights. This is where companies like Bud Financial step in. Starting out from the UK, the company has expanded into geographies like Australia and New Zealand and has recently announced an expansion into the US. 

The firm's data intelligence models can be applied to wealth management, lending, BNPL as well as open banking, helping FIs grow their loan books and improve onboarding processes. The company is funded by HSBC, TransUnion and Goldman Sachs. 

Tearsheet recently sat down with Bud’s CEO Ed Maslaveckas to discuss the firm's expansion into the US, the differences in the Open Banking environment between the UK and US, and how the recent banking crisis is impacting their work. 

1. Why did you decide to expand into the US and how is open banking and alternative data different in the US as compared to the UK?

Looking at the market within the Open Banking world, our focus has been on turning bank data into customer insights. Because of open banking, we've had a lot of access to data and in the UK, it allowed banks and other companies to feel really comfortable sharing their bank data. Due to this, we were able to build really good AI tools to understand customers off of that data. We were able to train our models in new countries quite quickly.

When we look at the US and other markets we're in, like New Zealand and Australia, where we're not relying on open banking, we rely on banks to share the data with us that they already have, rather than aggregating new data.

We help FIs make better lending decisions or understand their current portfolio risk. Open Banking has helped people in the UK to feel comfortable changing their lending models. Learning models typically are just done from credit bureau data. In the UK, most big organizations that we're aware of are transforming to this model of using both datasets – credit bureau and alternative data – together.

In the US, there's less of that. But that will happen in time. And obviously, we want to be here when that happens.

1033 – that's the regulation that is going to come out. By speaking to the regulators over here, I gather that it's not a question of if, it's a question of when. I think at least two years, but for us, we're winning customers here just on the data intelligence.

The US is a super competitive market, right? What we're seeing now is the kind of medium-sized and small bank sector over here. They're very keen to acquire new technologies or anything that can give them some advantage in lending.

They're also very progressive. In fact, the way they use data is in some ways ahead of the UK and Europe, but in other ways behind. Here, no bank uses the actual bank data they already have, which is kind of odd. But the challenge is that unlocking bank data takes a lot of money and a lot of time to develop models that are good enough.

Credit referencing data is sort of primary data. I see bank data as primary data for lending. And then there's all the secondary data like education, your social media and digital footprint -- I would say that is secondary, maybe tertiary, data. But that's the stuff that's been used because it's been easier to access that in comparison to bank data.

2. How is risk assessment different for BNPL especially when it comes to painting an accurate picture of a client's current liquidity?

BNPL is very relevant to transaction data, given the real-time nature of it. I could go out tomorrow and stack 12 BNPL products, and the credit data wouldn't know that. It will eventually get reported, but there will still be a 60-day lag.

We do see some alternative lenders now using just the bank data, and not even using credit data. I think that's interesting for certain niches but not for everything. I don't think it makes sense to not use bank data.

In the US, we focus on getting your bank data, and making it usable for marketing, risk analysis, or for internal use cases. You may be able to integrate all these AI products that are coming out all these days, like AutoGPT, but it's really useful if your data is enriched and usable – right now it's not.

3. How have the recent market tribulations affected your operations in the US or how you are approaching companies right now?

Some UK businesses started to say to us when the interest rates changed that they would like to understand what their real time risk looks like across their entire portfolio. Given the change in interest rates, plus the flip in the inflationary pressure on customers, clients began to ask what if our mortgages aren't affordable now?

So, we started to build a portfolio analysis that gives them the ability to view all their customers in real time, see what their incoming/outgoings are like, and look for any issues within that. For example, we typically expect this person to get paid on these dates and that has not come in, or there's some variance in income and expenditure that that flags as an issue.

4. Given the richness of data, is it getting easier to personalize financial products?

I guess there's two sides to it. First, there's the personalization of financial experiences, like getting your insight into what your money is doing. I think we're there.

There have been a lot of companies that have tried to run before they could walk. We can tell you everything! But in reality, this is kind of a boring point: they hadn't spent the years perfecting data categorization. So, any insight they gave was wildly inaccurate.

A lot of banks implemented personalization, but it was just built on crap, to be honest. People were selling the dream without the ability to do anything underneath. I think we're there now. In the UK, our models are now 98% accurate.

The second wave of personalization is taking information and putting it into a sophisticated model – so a customer's interest rate is based on a myriad of factors resulting in a product that is actually personalized.

That's one of my beliefs. I think there's a lot of value to be created by lenders, if they can create sophisticated models to give personalized financial products, whether it's an investment product, a loan, or a mortgage. That's a really big opportunity. We're way off that because someone always has to go out and do it first, and I am still determining who that is going to be.

They would make more money, and they would get less defaults – it would be better for everyone. We spent six years building models that are actually good. With our latest model, we went from 96% to 98%, because we introduced consensus labeling, and that got us another 2% inaccuracy. You're never going to get to 100%.

If you realize it's a hard problem and you treat it like it is, then you might get somewhere, whereas I think people have treated it like it's just a tick-box exercise.

5. What part of partnering with banks takes the longest?

The longest part of the relationship is the sales cycle. You get a demo if you were able to get in front of someone and then move through different people in the bank. That can take anywhere from six months to a year.

Obviously, it's different with fintechs. We've had fintechs go from conversation to deployment in like three months. But banks manage risk. When you're partnering with a company to basically perform data science on bank data, that's a risk. So, certain management things need to be put in place. It can take a year to two years in a sales cycle. But deployment can be anywhere between three to six months, depending on the challenge of deployment and things like how many developers they are throwing at it.