MoonPay’s blueprint for cracking self-serve analytics

Tarush Aggarwal (00:00):
Welcome to episode two of People of Data, where we get to highlight the wonderful people of data. These are our data leaders from various industries, geographies, and disciplines.

Discover how companies are using data, their achievements, and the challenges they face. Today on the show, we chat with Emily, a data whiz with over 14 years of experience. She's currently leading data at MoonPay, which is going through exceptional growth. She spent time in top companies like Uber and Coinbase. Emily, welcome to the show.

‍

Emily Loh (00:33):
Thank you so much for having

‍

Tarush Aggarwal (00:35):
Thank you for being here. Let's get right into it. What does MoonPay do?

‍

Emily Loh (00:39):
Yeah, so MoonPay is a fintech company with crypto DNA. We were founded in 2019 with one core product, which is our Fiat Omramps. So for those of you who aren't familiar with crypto, Fiat is what we call normal currency, like US dollars, etc. The product basically allows buying crypto from normal currencies directly into wallets. So for those of you who have never done it before, get on MoonPay.

By the way, it's very easy, but it used to be the case, and it still is actually to a large extent, where you have to go on an exchange, which is essentially sort of a bank. So that's what Coinbase is, for example, you buy the crypto and then you transfer to the wallet of your choice, or you keep it on the platform of whatever exchange you choose. So MoonPay enables the middleman to be cut out, and you can use your traditional payment methods like credit cards, debit cards, bank transfers, and so forth to be able to buy your crypto. What happened was that there was a recognition. So our founders, Ivan and Victor, recognized that the crypto ecosystem is really hard, as exactly as I was just explaining. You know, it takes multiple steps and you don't know what it is. It's very hard to get into the industry. So it's very challenging, and MoonPay is basically trying to take away all of that abstraction.

Trying to make it as easy as breathing or similar to the way that Uber revolutionized the way you can call a car just by pressing a button. It's not as simple as just one button, obviously, because it is still a payments and finance product, but it is as simple as it possibly can be. So today we've integrated into some of the largest exchanges, marketplaces, and wallets, including, for example, Bitcoin.com, Metamask, Trust Wallet, and other big names in the industry, OpenSea for example. We also have an integration with them. Over the years, we've launched other products like Offramp, so that's selling to Fiat and Swaps, so trading from one crypto to another. We also have a minting infrastructure, authentication, identity verification, and so much more that is wrapped up into our entire platform. So today we serve over 20 million customers across 180 countries and we support over 500 partners. So yeah, we've been doing pretty well and we're very proud of where we've gone since 2019.

‍

Tarush Aggarwal (03:07):
Yeah, that's awesome. I've been a MoonPay customer and the on-ramp, off-ramp stuff, in some ways it's become like the Stripe of the bridge between FinTech and the crypto world. How does MoonPay make money?

‍

Emily Loh (03:19):
Yeah, so we make money off of each transaction. We attach a fee to all the transactions. It usually covers a whole gamut of fees that are associated with crypto payments, such as gas fees for certain kinds of crypto, network fees, transfer fees, and that sort of thing. We take a small cut from every transaction. That's basically the way we make money.

‍

Tarush Aggarwal (03:48):
Awesome. And how does data play a part in helping MoonPay with the business, making money, all of the fun things?

‍

Emily Loh (03:58):
Yeah. So MoonPay is at heart a data-driven, product-led company. Data is the foundation of pretty much everything that we do. That is actually applicable to crypto in general. Like blockchain, at the end of the day, if you really break it down to what it is, it's a series of data points that are just chained alongside each other. They tell you a lot of information about what's being purchased, when it was purchased, by whom or what wallet, etc.

So data is really in the DNA of crypto, nevermind just MoonPay. For MoonPay specifically, we do everything from integrating data into operations, to product development, to customer experience, and yeah, like I said, everything. Our vision on the data team is to empower informed decision-making.

This includes both the manual type, people literally saying, "Hey, I choose this," as well as the automated type of decision-making through machine learning, through AI, and so forth, and even through data models that are integrated into the product. At MoonPay, we consider data to be an engineering function, actually.

I know that there are different companies that treat data very differently. For us, we really do provide the bridge between engineering and the operational part of the business. On the engineering side and the technical side, we do literally everything from end to end, including data engineering, analytics engineering, analytics, advanced analytics, data science, and machine learning, just to name a few of the things that people do with data in general.

But at the same time, we also strive to be the sort of conciliaries, as I call them, to the decision-makers. So the right-hand people of the decision-makers, making sure that decisions are being based on more than just intuition and that they're really trying to drive the business forward. Our main focus is actually product analytics at the moment. I think that makes the most sense. Like I said, we're a product-led organization, which means the product team, which we are technically a part of, is the largest part of the organization. We use data to understand pretty much everything as much as we can between the customer journey, including onboarding and all the factors that go into that, including know your customer, which is very typical of most fintechs and finance institutions in general, right? Payments orchestration, routing, transaction success.

And of course, asset delivery. And this is where it's really unique, I would say, because when we talk about, for example, my experience as an Uber and Coinbase, well, Coinbase is a little different because it's also crypto, but definitely Uber, you know, as a marketplace platform, the problems you're trying to solve are very different than the ones we are on MoonPay. Part of it is crypto delivery, right? And that is actually where we differ from Coinbase because Coinbase as an exchange, you know, they hold the crypto and they can deliver it right away.

For us, we're going to optimize what is the best exchange to get the crypto farming to deliver to our customers. What's the fastest? Sometimes things happen on-chain that you don't really know about and that you need to solve for. So data is being used basically in every part of that optimization. And we're trying our hardest to make the best experience for our customers and, of course, optimize our bottom line at the end of the day.

‍

Tarush Aggarwal (07:36):
That's such a good answer. What does the data footprint look like today? How big is the company? How big is the data team? What are some of the roles over there?

‍

Emily Loh (07:46):
Yeah, so the company overall is around, the numbers are between 250 and 300 on any given day. I think that's very typical for startups, right? Like Uber is not comparable because at the time I was there, it was like 13,000 people or something along those lines. But for a company of this size, it's numbers between 250 and 300.

‍

Tarush Aggarwal (07:53):
Yeah.

‍

Emily Loh (08:07):
The centralized team, which I head up, is about 15 people. And we're a mix of data analysts with some engineers and some data scientists in the mix. That includes ML as well, MLOps too. We actually do have a few people who sit outside of the centralized data team that are in finance, pricing, think, is one of them, and compliance, just a couple of examples. But we work really closely with them. So basically a dotted line into the centralized data team. And we make sure that they have everything they need to do whatever their goals are. We also expect a very high level of analytical skills whenever we hire. So the data footprint is actually quite large as a result. And the central team, my team, is responsible for building all the infrastructure, maintaining the tech stack. So we have Airflow, GCP, as well as DBT as our transformation layer. And then on the very front end, we have Looker, which is our BI tool.

‍

Tarush Aggarwal (09:26):
Yeah. Nice. When you mentioned GCP, you more specifically using BigQuery as your warehouse? That works really well with BigQuery and Looker. When you say Looker, I'm assuming the sort of commercial version of Looker, not Looker Studio. Awesome.

‍

Emily Loh (09:35):
That's correct. Yes.

‍

Tarush Aggarwal (09:35):
Exactly.

‍

Emily Loh (09:35):
That's right. Exactly. Yeah. So we're, you know, we've done as much as we can to actually use Looker as our self-service analytics platform. Of course, the platform is the entire thing, but Looker is the face of it essentially. And we have been pretty successful at pretty much enabling everyone to self-service most of their day-to-day needs. And this is super important because, like I said, we have 15 people out of the entire company who are on the centralized team.

‍

Tarush Aggarwal (09:57):
Yes.

‍

Emily Loh (10:11):
Um, I think something that's very typical of most young data teams is that you start to field all these requests, right? I think everyone has that experience in data where, especially when you're starting out, you're just answering questions all day long from stakeholders, from whoever. So we're actually trying to expand our self-service platform to make sure that we're not the blockers as well. Cause, again, 15 people are also across multiple time zones in the world. So we stretch mostly from LA, where I guess you are right now actually. So the west coast of the US all the way to Asia. And you can imagine, you know, someone's waiting for an answer. Most people are sleeping at some time. So that can be really risky in terms of who is able to access what information when.

Speaking of self-service though, I think it's somewhat controversial because we are rethinking, if I'd be completely honest, aspects of the self-service platform. I think this is also typical in the journey of data and small data teams because, yeah, you need it, right? It's absolutely necessary to enable the team to do more high-impact, like more needle-moving work. At the same time, especially, you know, we're not a huge company, 250 to 300 people, but we are large enough where we don't have eyes everywhere. It's actually impossible. So there have been cases, again, which is not unusual and it's expected where, you know, the level of analytical ability amongst different stakeholders is not created equal.

‍

Tarush Aggarwal (11:36):
Yeah, of course.

‍

Emily Loh (11:51):
Our metrics layer is always changing and the definitions are always changing, which makes it very hard for end-users to really know what they're looking at. And then, of course, you know, go for

‍

Tarush Aggarwal (12:00):
Is that for you? Sorry, was just, you know, in the metrics layer, you know, use Looker, which has got LookML, which is awesome. Has that really become the metric and semantic layer?

‍

Emily Loh (12:09):
Yeah, for us it has been mostly because it's easier to maintain like a singular layer within LookML. We also are very strict about our governance in terms of LookML and what we have in our transformation layer. So DBT and LookML are almost exactly the same. There are a few nuances, but we try to keep that the same just because, yeah, otherwise someone changes something in DBT.

‍

Tarush Aggarwal (12:28):
Yeah, got

‍

Emily Loh (12:37):
And then, you know, someone looking at Looker and it's not the same. We're very, very strict about those elements of our transformation process because of that reason, definitely.

‍

Tarush Aggarwal (12:42):
Got

Yeah, you have. What you're saying is essentially, you know, you don't use LookML just from a modeling perspective. There's no modeling, in fact, in it. You're really using it more in the semantics of like, you know, what you want customers to be able to do with that data, not so much on what the definition is.

‍

Emily Loh (13:06):
Exactly. Yeah. And I think there's also some tension here, right, between the way analysts and scientists use that same semantic layer, which we do, versus a more layman, let's just put it that way, user. And you might need more data and more granular data for sure in Looker for the purposes of analytics when we're driving deep, when we're doing model building, where we're really doing exploratory work.

And that doesn't always translate really well to other types of end-users. So that's why I'm saying that we're trying to rethink our approach to self-service because, at the heart of it, is, it is a great, you know, kind of initiative is definitely a necessity to have, but we have to have a lot more intent, I think, in terms of, you know, evangelizing what self-service actually is. Think the initial instance is just like, okay, just blow it.

Make it available, semantic layer is there, it's easy peasy, you know, and because of our governance, have faith that it is showing accuracy at least. But interpretability really differs between different stakeholders and that can be really problematic if you don't really control over it, have control over.

‍

Tarush Aggarwal (14:24):
Yeah, you know, we usually kind of get into like what's working well and like, you know, what are you proud of? You know, since we're on the topic, you know, one of the questions we ask on the show is, you know, what is one of the challenges you're facing and, you know, what are you doing to solve it? Is this the big challenge? And, you know, the evolution which you sort of spoke about from self-service and great everyone can use it. But that's a stage which many companies never even get to, right? Like they

All these desperate data. So they have all of these separate data sources. They have multiple sources of truth. So, you know, just the concept of self-service for many of these organizations has been a myth. So the fact that you're already there is, you know, already super interesting. But sort of going into one of the sort of big sort of challenges, what would you say that is today?

‍

Emily Loh (15:04):
That's true.

Yeah, I think it's one of the challenges, although I think it's a challenge that's under a larger umbrella about what the data team or actually, I won't even say the team. I would say the data function is supposed to add to an organization.

Having worked in a variety of organizations, it can differ so wildly. And you really have to decide, you know, like, where do we want to go? What kind of impact do we want to have?

There's also, I would say, an identity problem, right? Because when you're an analyst, and this is going back to, think, I don't know if we touched upon it, but, you know, nowadays there's usually a debate between, what is a scientist versus what is an analyst? What is an engineer versus what is an analytics engineer? And so on and so forth. I don't think it really matters that much at the end of the day. Where it does matter is where we have struggled with this at MoonPay actually.

And it is wrapped up somewhat into this self-service analytics question, if you will. And it is, who do we want to be? And what role do we want to play? And how do we actually end up achieving that? That's the biggest challenge, I would say. Because when you're in a company, so I've been in with MoonPay for two and a half years, it is super easy to...

Believe that, okay, people are coming to me with requests, that is my job. I answer those requests. It takes a very strong person to say, no, that is not our job. This is our job instead. So now we're moving a lot more towards understanding the role that data plays isn't necessarily to provide information just broadly.

Right? It is, in some sense, of course, but there has to be, like I said, intent behind it. Because talking about the digital, like the data footprint, our data footprint has gotten so big.

That I'm kind of like, we need to reel that in because it's basically a state of entropy where you're just like, okay, it's going to chaos really soon. And it's better than I've seen in a lot of other companies, which is great. But we're also not in the business of being like, okay, we're better than others. So we're good. It's not the case. We still have a lot to go. And so I think that data we're now operating on this model of data is meant to give operational clarity.

‍

Tarush Aggarwal (17:17):
Yeah. Yeah.

‍

Emily Loh (17:42):
To the entire business. And that's from every step. And first of all, to ourselves, actually, as the leader of the data team and leader of the data function, there are some times, admittedly, when I'm like, have no idea what's going on over there or why we decided to do that thing. And if I don't know as a leader, how do I tell the people to know? How do I know? How do I expect?

‍

Tarush Aggarwal (18:02):
Yeah.

‍

Emily Loh (18:06):
People on the ground to know either, you know, these engineers working on these really cool things to one end, you know. And so that's, think where data plays a very critical role in helping just understanding, which is different than information, I think, right? Because information is information and it can be taken at face value and you can think that it means what it means. But then when you speak to someone else in the same organization, you might even sit next to them.

‍

Tarush Aggarwal (18:11):
Yeah.

‍

Emily Loh (18:35):
Yeah. You realize like, no, they don't think the same thing as I do at all. So, you know, as a data function, you really need to kind of align everybody around those things and make sure that everyone is on the same train going forward because at a startup, especially, you can think you're on the same train, but actually you're on a different planet. And that can be detrimental to the business.

‍

Tarush Aggarwal (18:59):
I love the honesty and transparency and highlighting some of the practical day-to-day challenges. What is one achievement which you're particularly proud of where data has made an impact in the business sense, helping MoonPay do what they want to do, which is be that bridge between crypto and typical

‍

Emily Loh (19:20):
Yeah, for sure. So I think it's coming off of the same challenge that we're trying to solve, right? Exactly, how do you, what do we think about operational clarity? What does that even mean? One thing that we did on the team, one of my analysts created an opportunity sizing framework, and that has really helped rally the business in all sorts of ways around these really cool things we all want to do.

And, but we don't often have a mindset of, it's cool. But is it good now? Is the market ready for it? Do we have enough resources to put behind it?

‍

Tarush Aggarwal (19:53):
Yeah.

‍

Emily Loh (19:58):
What does success look like? And again, all these things we all presume because we're part of the same culture and part of the same company that we all think the same, but the nuance really makes a difference. So opportunity sizing framework, along with the prioritization framework that we've also put in place, really helps everybody understand the why of what we're doing. For example, when we're implementing a new payment method,

This has happened quite a lot, which is important for us because we're a payments company. So we want to make sure that we're diversifying our offerings to all of our customers. But the way that we go about integrating something like that can differ wildly depending on what we're trying to achieve. Are we trying to achieve more engagement? Are we trying to open growth of new users? Are we trying to improve transaction success? We're trying, you know, all of those things are factors and there cannot be a presumption, which we have done in the past, you know, that it is all understood because it's not. And so we're trying now to really understand whether these are initiatives that are worth doing today or, you know, what the scope could be

Depending on what we're actually trying to achieve in an explicit manner.

‍

Tarush Aggarwal (21:18):
Yeah, wow, that is a really thoughtful answer because it again goes into the why. And I think one of the challenges data teams have had, and I speak about this quite publicly, is that it's becoming difficult to quantify the ROI. If the answer is just decision support, we're helping you make better decisions, that's great. But

As a CFO writing the check at the end of the day, it's hard to like, what happens if we do 20% less of this or like 30% more of this. So I love the framework which you have around, you know, the impact framework. That is incredible. Emily, this has been incredible. We're super, super excited about what MoonPay has to do and thank you for being on the show.

‍

Emily Loh (22:04):
Thank you so much for having me. Yeah, it's been a great conversation. So I've had a great fun.