[Music]
[Jared: 00:08.3]
Ed, great to have you on the podcast today. How are you?
[Ed: 00:11.6]
Very well, thank you. I’ve just got back from a London to New York to West Coast, back to New York trip. So, I’m just happy to be in my own space!
[Jared: 00:20.6]
That’s a lot of travel – racking up those miles when you have to go that far. Wow. Those are long trips. Well, Ed, I’m really excited to chat with you today. For folks that don’t know Bud Financial yet, I’d love to hear a quick overview before we get started. I have some questions, but as I told you, I like to keep these conversations casual. So, let’s start there.
[Ed: 00:44.3]
Yeah. So, Bud Financial is a fintech B2B company. We’ve been operating in the B2B space for about seven years. Before that, we spent two years as a consumer app in the UK. We focus on helping banks and fintechs take their core data—typically transaction data—and turn it into a range of actionable insights. These can be used internally by the bank to understand customer behaviour, or to provide the bank’s customers or credit union members with insights into their spending, so they can make better financial decisions.
[Jared: 01:18.4]
So, what kind of problems are you helping these banks and financial institutions solve? What type of information are they looking for?
[Ed: 01:25.0]
There’s this idea we all have about our bank: because they see our spending, they know everything about us. We assume they have machines and algorithms processing all this data and that they’re able to target us accurately. We hold that in our heads, but at the same time, we get mailshots from the bank we’ve been with for over 10 years, offering us products that are completely irrelevant. All our mailboxes are full of that stuff.
[Ed 00:01:54]
The reality is that at almost every single bank we’ve spoken with—which must be hundreds, and maybe by next year it will be thousands—we’ve never come across a bank with a system capable of processing its own transaction data, leveraging that across the bank, and showing customers real insight. No one’s built that. There are a couple of people who have built poor versions of it.
[Ed 02:14.8]
It’s really in this era of more advanced AI models that we’re able to do that very accurately now. But the reality is that banks don’t have that, and we bring it to them. There are many pains and solutions that stem from that. Improving the customer experience is a simple one. Another is that customers often see transactions without a merchant identified, or without a location or details on how often they’ve visited that merchant.
[Ed 02:39.4]
Customers often call the bank to say they don’t recognise a transaction. Depending on the scale, banks can spend hundreds of thousands, or even tens of millions, a year just dealing with those queries. So, that’s a really simple use case we solve. But in the world of AI, there are much more advanced use cases and problems we address.
[Jared: 02:56.2]
Yeah, not knowing about certain transactions is actually pretty frustrating. When you look back over the years, it’s surprising how little information there is for the consumer. And I always think, even from the business side of things… when you have to take a picture of your receipt to log it. I’ve always said: “Why? Why do we even need to do that?” Shouldn’t we just know everything about the transaction so our accountants have all the details? It’s always seemed crazy to have to take a picture of a receipt. Anyway, that’s a side tangent, but I’m very bad at that.
[Ed: 03:37.1]
Also a side tangent. I’m very bad at it too. Oh, yeah.
[Jared: 03:39.3]
The only reason I mention it is PTSD, because my accountant is always chasing me for those items.
[Ed: 03:46.9]
Absolutely. Yeah.
[Jared: 03:50.0]
We hear a lot about personalisation in banking. But what does good personalisation actually look like to a customer?
[Ed: 03:59.8]
When a bank does personalisation, they’re usually doing it based on—before I started at Bud, I worked at Salesforce—CRM tools. The personalisation they’re doing is just: “Where do you live, how much do we think you earn, or how much did you earn when you opened the account or took out your last credit product?” You know? Your age. That’s really it.
[Ed: 04:28.8]
In financial services, good personalisation is basically knowing the customer almost as well as, or better than, they know themselves—and then finding them an opportunity to improve their finances in some way. Whether it’s a product or just a helpful nudge. That’s what good personalisation is.
[Ed: 04:44.9]
The nudges build trust, and that makes financial offers tailored to you from a data perspective much more likely to be taken up by the individual, because they already have some trust in the bank or credit union.
[Jared: 05:00.8]
Can you give… I always like a complex example and a simple example of how transaction data can lead to a better customer experience.
[Ed: 05:14.7]
A really simple one is if you have a credit card bill due on the 17th. We notice on the 15th or 16th that you don’t have the necessary funds available, and we know when your payday is. So, there might be an immediate action we could take a week out, or whenever we decide. So, that’s a very simple one.
[Ed: 05:37.2]A more complex example is understanding your FICO score. By looking at transaction data—whether we have the primary account or even aggregate accounts—we can see almost all the different financial products you hold, because there is typically a flow of money at some point throughout the year. We can identify those and then suggest, for instance, that you might be paying off two loans and that a consolidation effort could be worthwhile, because we know the loan rate the bank is offering or the rate currently available in the market. That is a slightly more complicated one, but it is just a matter of bringing lots of data points together.
[Ed 06:18.0]If you think about a bank—say, a large bank or even just a mid-market bank in the US—they have a million customers. They have no machines processing this, and there is no person going through individual transaction statements to do it.
[Jared 06:34.9]Now, what kind of reaction do banks have when they start seeing the insights that Bud Financial can uncover? Are there any light-bulb moments there?
[Ed 06:42.6]Yeah. I mean, for many years, we were essentially saying: “Look, we have this transaction intelligence and look at all the things you can get from it!” And banks were like: “Yeah, that's quite cool.” But then, about two years ago, we started building this product we call ‘Drive’, which is a platform you can access to get data visualisation. We can segment customers based on pretty much anything in that transaction data, so what we often do with a bank is say: “Give us a small file of your customer data and we will pull out a bunch of insights for you.” We just give them a login to the dashboard and there they can see a whole host of things.
[Ed 07:21.5]We can see credit card usage—you know, which credit cards and which products the customers have. That then gives the bank the insight to say: “Oh, actually, I didn't realise that about my customer.” Most banks in the US assume that Amex is the biggest credit card, but the data might show that in your customer cohort it actually isn't, and you might be working on a partnership with Amex, for example, or perhaps buy-now-pay-later is a big thing: “What do they use, and when are those used during the month?”
[Ed 07:56.1]Then those types of insights lead the bank to say: “Actually, I could probably offer a product that is specifically tailored to my customer cohort that might be competitive with Capital One, rather than trying to think about being competitive with American Express.”
[Jared 08:09.6]When you think through personalisation efforts, have you seen any of these efforts backfire or miss the mark? What makes the difference between helpful and creepy?
[Ed 08:25.8]Yeah. I think the ‘creepy’ factor was a concern about five years ago, when everyone was worried about Facebook data and being overly personal. I think in financial services, if you are offering something helpful, you are not being creepy. If you start saying things like: “Hey, you started shopping at these specific merchants, so are you about to have a baby?” or “Did you know your partner started spending on your joint account?”, maybe that is going a bit too far, and if there is no financial outcome or directly related financial benefit, it just doesn't make sense.
[Ed 09:02.2]Where we have seen it backfire is simply through core inaccuracy. We began building this language model around 2018 that understands bank data very well; it is highly tailored to that. There are two things that can go wrong there. One is that your model is not accurate enough, which is the main reason we get brought in. Banks are using other suppliers or they have built something themselves, and perhaps they have reached a level which is pretty good by industry standards—say, 80% accurate. But if I were to know you as an individual, Jared, your financial world involves—I assume—hundreds, if not thousands, of transactions a year. So, if you are 20% inaccurate, those errors add up. In fact, if I want to give you an insight, it is more likely to be wrong than right if I have a high degree of inaccuracy.
[Ed 09:56.0]So, we have really focused for many years—and spent a lot of money, as our investors will confirm—on developing highly accurate, granular customer enrichment, reaching the mid-90s in the US. Our older, or more advanced and mature, models in the UK are 97% accurate. And then we have to build customer-specific context models that take you a little bit further.
[Ed 10:20.5]We can see, for instance, that a transaction is regular, even though it doesn't look like a subscription. It is something you spend on regularly. Therefore, we might know that when you get a receipt back from that merchant, it is a receipt and not income. That often happens because, sometimes, you might have a return that is close to your income. Or even small things, like going to a petrol station and spending £3.50; you probably didn't buy petrol, it is likely a convenience store purchase.
[Ed 10:51.2]So, all of those little things add up to being able to understand you much better. I suppose a lot of people have promised personalisation, but the failure has been that they haven't focused on the core problem, which is the accuracy of understanding the transaction from a machine perspective. They have gone for the cool ‘bells and whistles’ of pie charts and things like that first.
[Jared 11:15.9]Yeah, I would say, going back to the earlier part about it being ‘creepy’, I think you are absolutely right. I think that was more of a conversation five years ago. I think it is more creepy now if you are so wrong with what you are offering me.
[Jared: 11:32.6]Yeah, exactly. This information is not information I would care about, right? Even spending habits and stuff like that. So, if I could have that more personalised to have a better understanding of who I am and what I want, that is brilliant.
[Jared: 11:48.5]I would love to shift the conversation to a newer segment I am trying out, which basically allows you to put your storytelling hat on. You have been building this company; talk me through something you can laugh about now, but which probably made you want to cry a bit when it happened.
[Ed: 12:06.2]I still cry a little bit about some of these things. I mean, there are many things we have got wrong. You see all these things on LinkedIn about how fast Cursor AI scaled to $100 million in revenue in less than a year, and then you look at how long it took this company to find product-market fit—and we are on that curve. We are at 10 years now and finding product-market fit is only just happening, and you can feel that pull, so you kind of cry or laugh.
[Ed: 12:39.36]Some of the missteps between 2017 and 2019 in the UK… We were helping three big banks build challenger apps. We were running the go-to-market strategy, we were doing customer communication for them, we ran the whole thing, and some of those banks actually invested in us at the time. So, from the top of the bank, it was very much something they really wanted to do. Then, within the space of six months, all of those customers just decided: “Actually, we are not going to let a third party run an app for us!” So, we were unceremoniously dropped by pretty much all of our customers at once. The not-so-funny part is having to let go of a large amount of the team. You get a headline saying you are basically a big failure, and you have to sit face-to-face and let go of a bunch of people. Looking back, it feels obvious that that model would never have worked, but in the unknowing world of doing it, we thought it would.
[Ed: 13:39.1]Another one is that in 2017/2018 we made this big announcement that we had flattened the company hierarchy and everyone was going to have autonomy. The reality is that a lot of even great employees don't want the level of autonomy we were giving them. It just seemed like a good idea as young founders. We had worked at big corporates before, we didn't get much autonomy, and we just gave people too much; it was very inefficient and actually a very difficult place to work because of that. So yeah, there are a few things we have tried.
[Ed: 13:39.1]I think you shouldn't try to innovate on those core business models that are tried and tested, as much as they might seem like a nice, utopian idea. Even employees will talk about full autonomy, but when they finally get it, they realise it is almost like they are running the business themselves. And the reason some people don't start these businesses is that it is very, very difficult. There is a lot of work involved in the whole process.
[Jared 14:38.9]It is interesting with what happened; you talk about some of those platforms you see every single day on social media. They hit these milestones with a certain number of employees, and I think it is… what is his name? Greg Isenberg? Do you ever see him post? I think that is his name. He always calls out these companies, and I am always positive and offer congratulations to all these companies for what they are doing, but a lot of that is just ‘vibe’.
[Jared: 15:08.0] Anyone who has tried a lot of those products knows they are cheap. You can immediately put your credit card details in, and it’s a bit like painting a picture for the people who have been building companies for years. It doesn't paint the full picture. I always look back to the really exciting example of UiPath and how they scaled from essentially just doing consulting and not making much revenue for a long time. Then, over a 10-year period, it was as if they flicked a switch after all those years of learning and just created this massive company. I think that should be more of the norm we point to than these vibe-based revenue success stories. I don't know. That’s just my two penn'orth.
[Ed: 15:52.3] No, definitely. My co-founder and I have a couple of posts we’ll probably put out at the end of the year that show this. It also highlights the difference between regions; we had some really great products in the UK, but the way the market works there makes it very hard to scale. We have a handful of customers here now that have completely outscaled the entire UK business in the last three months. It’s great in one way, but also a little bit painful. That’s the other thing: being a UK-based company for many years (with some customers in Australia as well, which is a similar market to the UK) and then coming to the US, we would always see these SaaS businesses hit this crazy scale in a short amount of time, and I just sat there thinking: “We’re terrible at business and we’re idiots!” Both of those may be true, but not to the extent we believed them to be.
[Ed: 16:47.8] Yeah. And then you go to pitch a US investor and they point to that as the example. It’s like: “I don’t know what we’re doing here.”
[Jared: 17:00.0] Have you tried it? I’ve tried all of these vibe revenue platforms, by the way, and they’re good.
[Ed: 17:03.8] Yeah. But they aren’t worth $100 million. It doesn’t even make sense. You can switch them out tomorrow, or something else better comes along; it’s very easy to switch. Now, some of these I’ve used. Cursor and Replit – we use both of them in the company, and we love them, for the record. But, you know, who knows?
[Jared: 17:30.5] By the way, I am a big fan of Cursor and Replit out of all the ones out there, and I think some of these companies will start to create more models. Which will be interesting.
[Jared: 17:40.7] As we wrap up here, Ed, what’s next for Bud Financial that you want to share with us today?
[Ed: 17:46.4] Yeah, for us, I suppose we’re in full execution mode right now. We’re tracking towards our profitability point at the end of this year, and we want to scale off our own backs for a little while. Margins are moving in the right direction, so that’s great.
[Ed: 18:01.3] The next thing we’re thinking about is that we’re implementing a bunch of AI products into banks. Obviously, banks and financial services are a little more risk-averse with technology than other industries. So, naturally, the next wave is agentic models. We’ve been developing some of our own, picking up tools to see how accurate we can get with them, and then putting them down again as the scale laws and the costs get really high. But I think that’s something we’re going to start reinvesting in at the end of the year, and it’s super exciting.
[Ed: 18:31.3] That definitely has a lot of potential in financial services, from internally running the bank to externally helping people manage their own money. Or, perhaps, not even needing to manage their own money in the same way they do today. So that’s pretty exciting, and we’ve got all the foundations to make that work.
[Jared: 18:50.7] Love it. And any events on the horizon?
[Ed: 18:53.2]
Well, I’ve just done two back-to-back. I did Fintech Meetup, which was great, and then I went to the Nvidia GTC, and we had some customer stories we’re working on. We announced that we’re using Nvidia and things like that. So that was pretty cool.
[Ed: 19:08.9] Personally, I spent 180 days in hotels last year, so I’ll be taking a bit of a break from events for the next two or three months to focus on managing the business a little better.
[Jared: 19:19.9] Do you have any advice for me? I was going crazy. I’m going to have 70 days this year. I just started planning all my trips and, I mean, to do 100… that’s crazy. I’m even looking at 70 and thinking: “That’s a lot!” Any advice?
[Ed: 19:37.1] I don’t really know. Do you have a family?
[Jared: 19:40.5] No. It’s the only way to make it possible. You wouldn’t have one for very long if that became an every-year thing, would you?
[Ed: 19:48.5] Well, I’ve done that for the last two years and it’s just horrendous. I don’t really know. I think the main thing is to try and eat as healthily as possible – that’s the biggest challenge.
[Jared: 19:57.0] Oh, that’s impossible. Especially when every event is in Vegas.
[Ed: 20:02.4] Yeah. There are no salad bars in Vegas! As much as I’d rather have a burger, chips and tacos every night, it adds up.
[Jared: 20:12.9]
Do you know one of my favourite quick spots in Vegas? There’s a place in Vegas that’s basically like this walk, you know, like the Chinese walk?
[Ed 20:23.8]
Oh, right. Yeah.
[Jared: 20:26.2]
But it’s a healthier version. It’s like Chipotle but for a walk.
[Ed: 20:30.9]
Oh, wow. Where’s that?
[Jared: 20:33.2]
It’s in the mall, the fashion mall.
[Ed: 20:33.2]
Oh, right.
[Jared: 20:36.2]
Yeah. And I went there far more times than I care to admit because I was in Vegas for two weeks.
[Jared: 20:44.5]
I’m sorry I missed you at Fintech Meetup. We’ll catch up next time. Are you going to be at Money20/20? I will be in October. Right. I’ll see you there then.
[Ed: 20:51.9]
Every year on my birthday for the last 10 years!
[Jared: 20:56.3]
Right. Fair enough. That’s a good way to celebrate, though, in a way.
[Ed: 20:58.3]
Sort of! Yeah, it’s a good event.
[Jared: 21:02.3]
Well, Ed, it’s been a real pleasure having you on the show today. I look forward to continuing the conversation. Please keep us up to date on any announcements and I wish you and the team all the best.
[Ed: 21:09.7]
Thank you so much for your time, Jared.
[Music]
%201.webp)
