…is it the data problem that I get most excited about and see the biggest potential for? Or do I have a particular end consumer or business use case that I desperately want to solve, and data happens to be the way of doing? There's no right or wrong answers, but they're just different problems, and both very exciting.
An interview with Matthew Ford, a Partner at Mouro Capital, a $400M fintech venture capital arm of Santander.
Peter Zhegin:
Hello, and welcome to datafounders, a series of interviews with entrepreneurs and investors, mostly structured around data science startups and challenges associated with these startups.

My name is Peter Zhegin, and I'm talking to Matthew Ford, who is a partner at Mouro Capital, a $400 million fintech venture capital firm. It's a pleasure to have you here, Matt.
Matthew Ford:
Yeah. Thanks for having me.
Peter Zhegin:
Awesome. Today, I would say we have a special talk, kind of experimental one, it's our first talk when we focus on a particular vertical. And today it's finance and banking. We even can narrow it down slightly, based on the things that Matt is particularly excited about.

Today, we are talking about data science, how it can be applied in fintech, what kind of challenges are there? To kick it off, maybe you can tell us a bit more about yourself, Matt?
Matthew Ford:
Sure. Yeah. Sounds good.

Hi everyone, I'm Matt Ford, I'm a partner at Mouro Capital. I've actually only been with the fund about a year. So prior to that, I was on the entrepreneurial side. Immediately before Mouro, I was the Chief Product Officer at Tandem Bank, which is one of the neo-banks in the UK. We should probably talk about some of the stuff we learned there, which is hugely exciting from a data science perspective. At Tandem I ran all the products and marketing teams. That was about identifying the big customer problems that we wanted to solve, and working with the product and marketing teams to both build those products out, but also distribute them to market. I spent a couple of years there.

Before Tandem, I was actually the founder of a company called Pariti, which, if you're familiar with open banking, we were an very early mover in that space. I set that business up in 2014. And really, the problem that that business was trying to address was the vast majority of people in the UK who and globally as well, who really struggle with money, they don't want to think about it. It's the cognitive load of thinking through cash flow, and budgeting and all these types of things is, you know, is really challenging.

But actually, at the same time, a lot of financial products that exist, can be hugely enabling in life, but also actually quite dangerous. Like credit cards, if you use them in the wrong way can be a real challenge. Personal Loans can be a great thing. But actually, if you don't pay them back effectively, then they can be quite dangerous as well.

So Pariti was really trying to help people take control of their money, trying to help them navigate the ups and downs of cash flow to understand how much money they had today, but more importantly, what they needed in the future. And to help them pay down their debt.

So I ran that business for four years. And really, that was the first foray into data science. I'll hold my hands up, I'm not a data scientist by background at all. I studied history at university, which is a very British thing, I think. I think internationally everyone thinks you're going to be a teacher if you study history, but, I studied history, I worked in strategy consulting for a number of years after that, I worked for another small fintech before Pariti.

But really a big opportunity was that open banking is about basically the bank's data had historically sat in their vertical silos, all your spending data to sat in one bank, you know, all your savings data is sat within the other bank. And open banking basically enables kind of the opening up of that data and for you to use that data more efficiently and effectively for your own purposes.

And I really saw a big opportunity - Okay, once that data is available, so once you can start to use that data and pull it out of those banks, can you start to underwrite more efficiently? So can you start to be able to understand the credit risk of an individual based on their spending behaviour, not just on a credit score, which is historically the way that the banks and lenders have had assess people? Also affordability as well. Whether you're a good credit risk, but can you really afford to make those repayments? And what are those cash flow implications?

So a big opportunity around data science and the use of open banking data to to make people's lives better with money, basically. We founded that business in 2014. We were really early in the space. You know, I think we made so many mistakes, the way that we set up the data science, our data pipelines, all that type of stuff, but learned a huge amount along the way and ended up selling that business to Tandem. So that was how I ended up at Tandem Bank, which was much bigger, kind of grander setup if you like with a much more mature data science setup.

So my background's entrepreneurial, my background's fintech, I've always wanted to solve big consumer problems with money. And now I've moved on to the venture capital side. My interest in the space is still huge, but it's more now helping enable entrepreneurs to fund and to build those businesses to tackle their own problems, rather than just building my business.
Finding the idea for a startup – a collision of a problem and ability to solve it via data sciecne
Peter Zhegin:
That sounds great. And that's an interesting combination, I would suggest to talk a bit about your entrepreneurial journey, and then switch to venture capital part, and to talk about the ecosystem of fintech and discuss it further.

When you started your company, how did you arrive to that point? What was your maybe journey, or your mental model? Any framework? Maybe? How did you understand that it was the right time for you to launch a company?

Matthew Ford:
Yeah, it's a good question. So I briefly mentioned it, I worked for another fintech that was called OnTrees. I was a very early employee at that business. I'm not sure if people are familiar with mint.com, it's probably quite a more well-known example, but OnTrees essentially enabled you to see one view of your finances across all your different accounts, and helps you categorise and get a clearer view of where you're spending your money. That was kind of the main thing.

And not to oversimplify it, but really, it was about - 'I'm spending too much money on on restaurants, I better cut down so that' - it was tentative steps into some of the data science problem around bank data. But we weren't a data science team, we were a development team and a product team that was quite reliant, I suppose, on one of our suppliers to do all the correct categorization and to get the accuracy of that categorization correct. I think we learned really quickly that everything was so nascent there, and if the 50% of transactions were getting categorised as generic categories, the usefulness of that tool is actually relatively small.

And I really started to understand the size of the data science problem, because... to take an example, your coffee might be an essential expenditure for you, because you know, I can't start my day, if I don't have my coffee. But then for somebody else, it could be frivolous, and something that they can definitely cut back on. So how you start to bring bank data to life, I think it was a really interesting challenge. And how you start to add meaning and insight and enable people to do with it, was something else.

But OnTrees, it was very successful, we ended up selling it to MoneySupermarket, but that was kind of what sparked my interest in, there's got to be more that you can do with this data. Like nobody's really maximising the opportunity there. And this was also kind of at the same time as in the UK, and globally, when there was an increasing rise of payday lenders.

It was a largely unregulated space at that time. And there's a rise of providers who were offering very high cost credit, in the 1000s of percent APR for people who had ultimately run short, and for whom banks often were not able to serve very well. So they might have an overdraft of 50 pounds. But that was about it. And this was becoming a really common problem. And I think in the UK in about 2014, two - two and a half million people were using payday lenders. And there was a lot of research coming out at that time about actually the danger of those types of models.

Although it might seem a small amount of money that people were about borrowing at the beginning, it was starting to rack up. So if it was 5000% APR, you can imagine, all of a sudden, it's a very slippery slope into quite a lot of problems.

So those two combined forces, I suppose, of really understanding that the technology was starting to arrive, and that there's tonne of opportunity around bank data to bring it to life, and to understand it better. But then actually, there was also this massive opportunity to help people who wish falling short of money to not fall short of money, but also to borrow it much more affordable rates.

So it's a bit of a collision, I suppose, of both problem and opportunity at that time. And I'd never started a business before I was completely naive and fresh. And although I worked in this fintech, it wasn't my business, I was an employee, not the founder.

So yeah, I threw up everything into the air, it seems a bit crazy now looking back, but I quit my job. I, you know, I just focused on this space for six months. And really, I'll come on to it in more detail, but just spent a tonne of time really looking for is this problem as big as I think it is? And more importantly, is there a better way of trying to solve it using some of these early hypotheses that I had?
Customer development – checking the major hypotheses, importance of the MVP
Peter Zhegin:
And this process of checking this hypothesis, can you recall maybe what was your approach to understanding that these hypotheses really work. What did you do?
Matthew Ford:
Yeah, I would do it very differently. But I think we followed a fairly traditional approach. So we did the classic interviewing, researching as much as possible. And I think you learn a huge amount, just looking people in the eye and asking them quite often quite emotional, difficult questions about money that's a very sensitive topic.

I quit my job, and I think it was March 2014, and probably spent about the next nine months, if I'm really honest, just talking to people really getting to understand that problem, understanding whether this was something that... if you provided them with the tooling, they would even want to take it, or if it's something that they just kind of shut the door and pretend isn't happening and won't engage with it.

A lot of it was face to face interviews, we didn't do a huge amount of statistically significant surveys or anything like that. I think there's definitely a place for that. But again, when you're funding it yourself at the time, you're always finding hacks, I suppose, and cheap ways to try and get a referral to a friend or whatever it may be to try and get that insight. I suppose it is very kind of face to face qualitative research to begin with.

But then I suppose the thing that really started to then click... two of my old colleagues from OnTrees, they'd also both left the business at that time. And we teamed up together to knock together a prototype because, there's only so much research you can do. And this is what I mean by 'we probably would do it even different', this time round. But there's only so much research that you can do, you just need to get a product in people's hands, you really need to see the data, you really need to see how people are behaving. So we knocked together an incredibly simple prototype, which connected to your bank accounts for a start to understand whether people would even connect their banks, if they struggled with money. So if they were in the target segment that we were looking at, would they trust the company to do that? And then more importantly, could we start to build some simple rules and categorization and approach to basically what I was mentioning before, which was this cash flow forecast to understand how much money you've got today and what you need for the future. So we we did that in our spare time. We begged, borrowed as much as we could.

A company called [unclear] was very kind at the time, they offered us to use their service for free for a number of months. Like we just hacked it together. But by the end of the year, we'd actually got something that was working, that was that we're learning a load of stuff on we've got real customers coming through. And we've got real feedback. And that was the point at which we realised that we've got something that needed properly developing at that point.
Overview of fintech: three major themes
Peter Zhegin:
This amazing journey. I guess we will return to some pieces of it a bit later. But right now, I would suggest to wear your venture capitalist's hat. So you mentioned this personal finance space, also open API theme, could you maybe give us a bit broader picture of what actually fintech is?
Matthew Ford:
So very good question. I mean, fintech can seem quite narrow in some regards, but actually, it's an incredibly broad space. And it's probably worth outlining it in the way that we think about it at Mauro, at least.

So we're quite thesis orientated as a fund and at a very high level, we try and split it down into three broad categories.

The first being infrastructure, and what I mean by that is, if you have a look at a bank, or if you have a look at an insurer, and we invest more broadly than just fintech, we bring insure tech and B2B SaaS and as well, but if you look at the traditional stack, the traditional technology stack of a banker and insurer, there are so many layers of old archaic technology, inefficiencies across that whole stack. So we see a massive opportunity. And this is kind of bread and butter fintech, really, it's 'fintech 010', of basically improving the efficiency, the accuracy, the ability, for large incumbents to modernise the technology that they're sitting on.

But then the flip side of that is obviously if you can improve the stack that you build financial products on, that brings a whole new number of challenges in, and lower barriers to entry for challenges as well. So that first bucket, as I say is the banking and insurance stack. And I would include things like KYC (know your customer), AML (anti-money laundering), core banking, collection, servicing - a suite of different things within there, within that infrastructure category.

We're also quite bullish on the infrastructure as a service space. What I mean by that is historically, there's a bank that's fully regulated and the bank has its technology partners, and they build the full vertical solution, or increasingly, we're seeing this trend and there's a lot of other VCs investing in this space of what we call embedded finance. Which means non financial brands will start to embed financial services within their product suite. And there's infrastructure that will both enable that from a technology perspective, but also from a licensing and regulation perspective. So that's bucket one, that infrastructure broadly.

Second bucket is around what we call the challengers. So take any financial product, take a mortgage, take SME lending, take personal lending, you name it, there are attacks from all angles from new challengers who are either serving a different segment that historically been ignored, or using a new data set or data approach, that means that you can underwrite or offer products to customers that that weren't possible before.

In the past, it could be a new business model, they could be completely rethinking pricing and how that works. Or it could be a technology solution that they've got, that just means you can rethink that financial product. So that for us means challengers. So we have backed Upgrade, for instance, which is a challenger bank out of North America, and similarly Klar, down in Mexico. But then in Europe, we've also just backed in the last six month, Uncapped, which is a revenue-based lender. So there are all types of businesses, which are really just rethinking what a financial product is and service and, challenging the incumbents basically. So that's category two, which is the challengers.

The third one, is super interesting, but also a little bit further out, which I call transcending fintech. But, but this is kind of really thinking about those massive societal shifts that are occurring, be it sustainability, and climate, education, mobility, lots of these things are fundamentally going to shift over the next 10-20 years. And therefore, the role of finance is going to completely change as well. And I often use the example of cars. So historically, you would want to buy a car, you'd finance the car, and then you go and insure the car. An example being that actually the service that we want is to get from A to B, and mobility is going to completely change, be it driverless cars, services that you can subscribe to cars, all those types of things. So we're really thinking beyond the constraints of a traditional financial product, and looking forward for what these kind of big societal shifts will mean for the way that we interact and use money.
Peter Zhegin:
A great overview, and which layer sounds like more exciting to you? Or maybe where are opportunities for data science startups in European right now? Across all three layers? Or maybe some of these layers are a bit more prone to disruption, or augmentation by data science?
Matthew Ford:
I mean, there's data science opportunities across all of it. You know, I think, personally, I am very interested in the infrastructure side. So I pulled out a few of those earlier of KYC and AML. And there's a tonne of opportunity there for building much better technology solutions, and data science is often the key thing in each of those different areas.

I am personally very excited in that space. It's increasingly mature, but there's a lot of new startups are continually rolling into that space as well. From that category two, challengers' perspective - I was trying to solve that problem before. I mean, underwriting is still it's still hardly been disrupted, like the credit bureaus still exist, Experian, Equifax, they are still the predominant ways that that lenders source data, and often they're not even building their own sophisticated scorecards, [but use] quite generic scorecards. That's still a passion space for me, because it means people are getting expensive credit because they can't be priced effectively, it means people are getting rejected for credit because they don't fit within a traditional frame of what a lender is looking for. Or it means that they're struggling with credit. And they're misusing credit because either affordability wasn't judged or the product was too complex and those types of things. So I'm personally incredibly interested in in both of those spaces and actively looking for startups.
Two ways of identifying a data science problem, what to look at
Peter Zhegin:
If we focus on these two spaces, if we look at them in a bit greater detail, is there any approach or a framework to identify problems within these spaces that are really well suited to be solved by data science? How a data scientist might think about it, especially those people, those engineers or data scientists who don't have the fintech background?
Matthew Ford:
Yeah, it's a really good question. I suppose there's a couple of things that springs to mind. I mean, the first is, is almost kind of manual inefficiency. If you get walk into a traditional bank, there's so many manual processes involved in that, there's people manually looking at flags that occurred on their transactional screening for potential fraud, and there's people working those cases, taking a list of cases and manually working those cases. That's crazy. It's huge opportunity to both identify the false positives and to be much more efficient in some of that screening upfront.

But also then if there is a degree of manual intervention, prioritising the workflow, the backlog, automating some of those tasks, actually kind of removing some of the manual processes that exist. I think there's a tonne of data science opportunities, wherever there's manual processes, automation opportunities. And, you know, this is spans fintech, but it's completely applicable in other sectors as well.

To use a real example. There's a company called Juno, for instance, which is in the legal space and not quite fintech, it's the conveyancing space. And they basically said - law and conveyancing (being kind of the legal service of when you buy a property) hugely manual, it's about sending a PDF here, it's about reviewing a contract there. So that just chipping away bit by bit, every single one of those manual processes and automating it and using machine learning and AI to both prioritise the backlog of the lawyers, but to just chip away at those manual processes.

I think that's a really nice example within law and within conveyancing, but that can be applied in so many different sectors and sub verticals of fintech. So I think that would be probably one of the few things out there, which is that manual work.

And I suppose it's kind of another really interesting space. And again, this kind of comes back to the Pariti experience, really, which is those areas where there's historically been just an either captive data set, or a very narrow data set, it's just the way of always doing things.

And a credit bureau is a perfect example of that, the credit bureau, the big bureaus exist, because they did deals with the banks, they get direct data feeds from the banks, and they have a huge amount of data clearly. But it's, it's a very known data set. That's grown over the last 15-20-30 years. And it just excludes a huge number of customers out there. So I think those areas where you go - 'Okay, well, you know, I don't fit into the traditional bucket of the way that a lender might look at me', or - 'you know, like, there's no data on me that identity is another area'. So where are those areas where the datasets just don't currently work, or it just doesn't fit for you. And emerging markets are fascinating in that regard, because they don't necessarily have established identity providers, they don't have established credit bureaus. And there's a tonne of opportunity there for using alternative data sources. There's been loads of examples of people using social data, for instance, where credit data doesn't exist.

So lots of opportunities to look beyond the existing established way of doing things, by new data sources, find new approaches, and really rethink how you can identify, score, assess, personalise, all these different things, using new data sources. So those would probably be my two that just jumped off the top of the head.
An approach how to choose what to build – a software/tools or and end-to-end startup?
Peter Zhegin:
I guess both of them resonate really, with the data science community and with the things that people do in some other industries. Maybe you can look a bit more at the second one regarding the data sources. Correct me if I'm wrong, there are at least two business models here, right? Somebody may build a tool for a bank to collect this data, for instance, or maybe to produce it synthetically, artificially, or somebody may decide and to build a credit bureau of 21st century or 22nd century. What do you think about these approaches? Maybe there are pros and cons to both of them? What would be better suited for a data science team?
Matthew Ford:
Yeah, it's fascinating because it kind of fits into our first two categories, a little bit as well as with the opportunities on both fronts. Because if you can become the infrastructure for many companies, you can build a hugely valuable company. And obviously, there's great examples out there, of Stripe and Twilio, and Plaid and those types of companies where they've just became the approach that everybody uses and the infrastructure that everybody uses.

I think there's massive opportunities to do that. Not necessarily build the end product yourself just for your own customers, but to provide the technology and the infrastructure for others to offer that. And I personally really like that space. We invested in a company called DriveWealth, which I wouldn't really call it a data science company, but it's a brokerage as a service, like brokerage infrastructure company, and they're doing phenomenally well, because they're doing some really hard stuff. They become the specialists, doing the hardest stuff, and then enable everyone else to implement them, and integrate them. Really big fan of that business model.

I mean, it has its pros and cons. Because you are reliant on other customers buying and using your solutions. So sometimes when you have a look at the growth of those types of companies, it can take quite a long time, because you've got to be an absolute specialist, you've got to do the really hard, difficult things. Otherwise, why wouldn't the client do it themselves, and then often requires a huge amount of build upfront, or a huge amount of kind of richness of features before you can then start to get the scale of customers that sit on top.

So you traditionally see, I think there's massive outcomes that are possible with those types of businesses. But often there's a slower start a slower burn. And then once they've really found the two or three big growth clients that start to use those, those types of services, that's when those types of businesses really pop.

And I suppose to draw it back a little bit to data science, and I suppose to underwriting to some extent, the guys at Credit Kudos do really well, and they're a really interesting example, because they're basically taking that hypothesis, which is, bank transaction data is an amazing way of assessing affordability more accurately. And eventually, you can start to use a lot of that data to underwrite more efficiently. But they haven't built, a lender, what they've done is they've built the platform for lenders to integrate. And I think businesses like that are really fascinating, because then you can get a lot of the market, 90% of the way by using one single provider rather than every single one of them building their own.

But I suppose it depends on how much of a competitive advantage as well, doing that thing yourself is. Because... Tandem is a really interesting example. So at Tandem we actually built all our data science capabilities sat in house, and that was because it was absolutely core to the strategy. So similarly, we used open banking data and we saw the potential of that type of stuff. But actually, more importantly, we didn't want to just do 'the hygiene' 60% or 80% of [what was possible], actually saw that kind of the data opportunity was the thing that could completely differentiate Tandem. It would mean that they could price so much better than than anybody else, it means that we could offer products to more people. And so really saw it as a core competitive advantage.

And at that point, you then want to build that in house, you want to build the right culture, you want to build the right processes, you want to build the right team, you want to start to use that capability as a differentiator. So I think it depends whether it's hygiene or differentiation. And then as a founder, if I'm setting up one of those businesses it depends, is it the data problem that I get most excited about and see the biggest potential for? Or do I have a particular end consumer or business use case that I desperately want to solve, and data happens to be the way of doing? There's no right or wrong answers, but they're just different problems, and both very exciting.
A team composition for a fintech startup – you can't outsource regulation experteise
Peter Zhegin:
It leads us nicely to the concept known as the product-founder fit. And there is a kind of a generic view on the ideal startup founding team like hustler, hipster, and hacker. How do you think that is applied or not really, to fintech? How do you think an ideal team for fintech startup does look like?
Matthew Ford:
Yeah, I mean, fintech has its quirks compared to other sectors. And it's kind of what draws me to it really, which is when there's regulation involved, things are always 10 times harder than when there's no regulation. But although it is different to a normal sector, there's definitely similarities to any starting team. I mean, this is a personal opinion, I think that the classic model is right for so many reasons. But for me, the most important thing is a collection of people who really care and will work incredibly hard to try and solve a specific problem.

I mean, that's all a startups about, I mean, there's this big organisations, big incumbents, they have all the money they need to innovate. They could innovate if they wanted to, but they have the barriers and challenges and I think the big competitive advantage that startup has, is often this narrow focus and this cutting out all the inefficiency, cutting out all the overhead and just getting a team of people together who are just singularly focused on solving a problem. And there's no politics, there's nothing like that.

So irrelevant of the specific roles that you have in the team. I think getting the right people that are bought into solving that problem is the most important thing. And it changes over time, obviously, as you get to 100 people, people are motivated for different reasons. But really, it's just fun finding that trust, finding that team who just really want to solve the problem. I think, obviously, naturally, then there are different roles that people have to play, but you're wearing so many different hats.

When I said, at Pariti, there were there was three or four of us to begin with. But we're all just doing a bit of everything. I mean, my CTO, he's also at the same time was looking at wireframes, and at the same time coming to investor pitches with me. So I think naturally, in the early days, you've all got to mock in, and you've all got to do a bit of everything. And you've all just got to get completely out of your comfort zone. Our designer at the time is doing research is doing design, she's doing product management, she's doing testing everything.

So I think there's that flexibility of roles. But you of course, you need a developer there, of course you do. Somebody who can move fast, who you know, who wants to ship product and get data and feedback as early as possible. It's great to have a commercial person, they're really thinking through pricing, and really thinking through how you're going to sell this. We had a designer very early in the team, because Pariti was a consumer proposition. And we felt that user experience and user research, all that type of stuff was going to be incredibly important.

But then where fintech differs... I think in heavily regulated environments, having people who really know the ins and outs of regulation early on, if you are playing in a regulated spaces, is critical, because lawyers are hugely expensive. And if you're a startup, that doesn't have any money or it's raised a little bit of money, the last thing you want to do is spend 50 grand of that or half your money on lawyers. So the ability to understand the regulated environment is probably the biggest difference, I would say. And most of that falls on the founders counting team shoulders, and you know, they have to get down in the detail. It [soould not] be something that's outsourced.
What do you build - a data science organisation, or an organisation that has a data science team
Peter Zhegin:
And when a startup grows beyond this initial team of maybe four or six people, how may an integration and interconnection between a data science team and the other teams may look like? You mentioned that you had some learnings through your own company, I assume you also have some generalisations as an investor because you see lots of different startups with different teams composition, etc.?
Matthew Ford:
Yeah, so I learned a huge amount at Tandem. So towards the end of the time that I was there, they brought in a chief data officer who's actually now the CTO, who's phenomenally impressive, if you ever want a guest on the show, I think you'd be the ideal person for it. But he told me a lot actually, about almost kind of the cultural change and the organisational change that you need to properly embed data science into an organisation. So again, not being a data scientist, myself, and also having seen great data science problems being solved, but they're not effectively in production, or they're being like a massive gap between the people building the really smart models and doing really clever stuff. And then there's this massive kind of air gap between them and being in production.

I just saw it not working very well, at Pariti, I suppose to some extent of my own business. But Tandem was really different. Noam, who is now the CTO, he came in, and, you know, from the ground up, data science was kind of embedded into the organization. He thought in a very, very different way, about the way that completely disparate problems that different people within an organisation were looking at. That could be the team looking at fraud, that could be the marketing team who are looking at ideal customer personas, and targeting all that type of stuff. Could be the underwriting team. Like they all seem quite disparate, different problems that can equally be solved by data science in isolation, but actually, it's about pulling all that together, because a model that the underwriting team might build, there could be some learning that could be applied for the fraud team, that could be applied for the marketing team. So it was that interconnectedness and pulling together and really thinking through that, the whole organisation is the data science organisation, everybody is an input and although you need coordination and you need a structure to pull it all together. That has to be that kind of that organisational-wise way, a structure to enable that. So yeah, he taught me a huge amount.

You know, again, don't pretend to be a data scientist, but I've kind of taken a lot of that learnings on now being an investor. Because it's very easy... a bit of a joke to some extent that you just got to say to a VC 'Oh, yeah, you know, it's AI, machine learning', and all of a sudden, they're really excited. That is a really hard challenge, because without being a specialist, unpicking how much of that is sales story versus how much that is reality and differentiation and really credible. So I think I used it quite a lot of that learning from Tandem in my new job. You can just see straight away whether an organisation is what I call it, like a data science organisation, or whether an organisation is one that has a data science team. And you can see straightaway, and I think that's been hugely helpful, because one scales really well and one doesn't.
Peter Zhegin:
Yeah, and you mentioned this interconnectedness. Are there any other flags that may signal to a VC investor, and to a founder, that help to understand if a company is really data science driven, or it is a company with a data science unit?
Matthew Ford:
Yeah, I mean, it's a funny thing about being a venture capitalist. Unfortunately, there isn't a formula, you know, and I think that's one of the things that so much of it is intuition, and so much of it is a people job. A lot of my time is spent on just getting to know that person, getting to understand how they think, I'm never going to understand say science, as good as a scientist ever.

And you got to really trust in, you got to really believe in the people that you're backing. That comes from data science, from pricing from marketing, you've got to believe that they will make the right decisions and do the right things. Because you're not in control. You just got to find people who will make really great decisions. So it was probably not a satisfying answer. But I find it a lot more intuitive, if I'm completely honest. Just does this person really believe in doing things differently? And are they setting up their organisation in a different way thinking about things in a different way? Or are they showing the fact that they've got two data scientists and you know, I'm putting their qualifications on the pitch deck, rather than really thinking through an explaining why that differs from from the norm.

So probably not an overly satisfying answer, but a lot of it is intuition. And a lot of it's just getting to know the individuals and speaking to them and getting to understand the organisatio.
Peter Zhegin:
I guess you pointed to a very interesting direction, further away from credentialism into more of understanding of what really people believe in and how they behave. And probably, as you're saying, a founder or a VC should look beyond these credentials at the deeper level.
Prioritisation: finding product market fit and investing heavily into data science
Peter Zhegin:
Just to be cautious of your time, probably the last question. We've talked a lot about different things. And we could talk more about others. But is there any question or theme or topic that we didn't talk about, but you believe it deserves to be covered?
Matthew Ford:
Yeah, I suppose we have covered quite a bit of ground. Drawing, again, back on the Pariti experience a little bit - what is the right time? So going back to that, you need to set up the organisation, right, culturally, from a mentality perspective, from an organisational structure. There's also kind of the pragmatic approach, which is when you've got like three or four people in a team, and everybody's wearing various hats, you can't spend a tonne of time on organisational structure, you're in the middle of the fire, and you're just trying to find something.

So I think a really interesting question for me is that bit at which you start to mature the organisation and now at what point do you bring in a data science specialist versus somebody who is good enough in the early days, who may not be a data specialist at all, but we'll get you so far.

I kind of use the example a little bit of Pariti because I think we did exactly the right thing at the beginning, but then we completely under invested towards the end. So when we had 10,000 - 20,000 customers that had connected their bank accounts... Not a huge amount of customers, obviously, it's quite a lot of data because you get typically about six months of historic data as well. But this is not billions of transactions and years of data to understand cycles and all that type of stuff. So should we have jumped straight to machine learning approach, should we have hired all the top talent and built out a data science team and completely changed, like the way that we structured the organisation?

At that point, it was probably probably not the right thing. You know, there was so few of us in the team. And we, it sounds cringy now, but we had a very simple rule based algorithm, that got us where we needed to get to, to learn what we needed to do to get to the next stage of business.

A mistake we then made was, we then probably ran for about two more years of refining that approach and refining it, staying in a rule based world and not then thinking about ourselves as a data science organisation. So we validated some product market fit, we validated the customer problem, we validate we were able to raise some money. And then I think we went too long. So I think a really interesting question is, what time do you hire the right data scientists into the organisation? Who were the right people to lead that? At what point? Do they change the structure of the organisation? And I think that's probably been my biggest learning.

And I don't have the magical answer to it. But there's definitely a gulf between the 200 person startup that's got it right, the five person startup that's just trying to survive, and there's that journey in between, and you've got to balance like over investment versus under investment, getting the right long term structure versus the short term structure. And, you know, I don't have the answers there. But that's what great founders can do. And that's where the great businesses have been able to navigate that chasm pretty well.
Peter Zhegin:
Do you think a part of the answer about timing might be the volumes of data that someone has, and maybe the stage of the company? Maybe someone can pay a bit less attention to data science, before product market fit is here? And then switch gears and to double down on data science?
Matthew Ford:
Um, possibly, possibly. And because, again, I think it depends on the type of organisation. Because if the organization's one differentiating factor is that they can offer the product to customer segments that previously weren't possible because of the use of data, if you don't invest early into that, then you're probably going to fall short.

If your business model is we can do this more efficiently with fewer people. And an incumbent is really expensive and slow. [Any you are] better by being digital, by being smart, by using data science, we can we can do things at a 10th of the cost. And it's more about efficiency. I think that's a slightly different question. Because inevitably, you're also going to be really inefficient at the beginning. And it's been a journey, and it's probably a slower journey, but it's a journey to then automating and doing more overtime. I think it depends on the business. I really do.

But as you say if you got no customers and you got no data, it's probably the biggest clue. But then don't leave it too late would be my advice if it is a core differentiator.
Peter Zhegin:
Thank you for raising this question about timing. I guess it's one of the cornerstones really. And it looks like our time for today is probably ending. So thank you for chatting with us today, Matt. Really appreciate that. And I'm sure we'll hear a lot of good news about Mouro and startups that you guys are backing.
Matthew Ford:
Great. Well, thank you very much for having me.
Peter Zhegin:
Thank you.