Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
026 - How Monzo hyper-scales Analytics Engineering
445 Plays4 months ago
We’re back, and we’re kicking off with a BANG! 💥
Monzo is home to one of the most respected data teams in the UK—true trailblazers in Analytics Engineering, regularly sharing insights into how they scale data. This episode is no different!
I had the pleasure of sitting down with John Azzopardi, Senior Analytics Engineering Manager at Monzo, who’s been with the company for over 6 years. We dive deep into how Monzo’s data team scaled from 20 to over 170 people and how Analytics Engineering plays a crucial role in supporting their hyper-growth.
In this episode, we cover:
✅The evolution of the Analytics Engineering function at Monzo
✅How the team is structured for success and their key responsibilities
✅Challenges of scaling Monzo’s data warehouse and infrastructure
✅Why incremental modeling is a cornerstone of their data strategy
✅The future of Analytics Engineering in fintech and what’s next for Monzo
If you’re looking to learn from one of the UK’s top data teams, this episode is a must-listen! 🎧
✨ Hear firsthand how Monzo is mastering data at scale and driving innovation in their infrastructure.
We’re dropping new episodes every other week, so make sure to FOLLOW & SHARE to stay in the loop!
#Podcast #Fintech #AnalyticsEngineering #DataEngineering #Monzo #Data #Innovation #DataInfrastructure #Scaling #IncrementalModeling #TechLeadership #ModernData
Recommended
Transcript
Introduction to Stacked Podcast
00:00:02
Speaker
Hello and welcome to the Stacked podcast brought to you by Cognify, the recruitment partner for modern data teams hosted by me, Harry Golop. Stacked with incredible content from the most influential and successful data teams, interviewing industry experts who share their invaluable journeys, groundbreaking projects, and most importantly, their key learnings. So get ready to join us as we uncover the dynamic world of modern data.
Monzo's Innovative Banking Approach
00:00:34
Speaker
Hello everyone, and welcome to another episode of the Stacked Data Podcast. This week I've got another ah very special episode for you, as I'm joined by John Atsopardi, the Senior Analytics Engineering Manager at Monzo.
00:00:47
Speaker
Monzo's household name, which has been challenging the traditional banking sector, putting tech and data at the heart of its business. Widely regarded as one of the UK's top data teams, I'm delighted to have John on the podcast to share more about what makes Monzo's data team so special and how analytics engineering has become a core part of their business.
00:01:09
Speaker
John's been in the organisation for over five years now and truly opens up the door as to what makes the team so special, the biggest challenges they work on and what they're looking to do as they move into the future. John, welcome to the podcast. Thanks for your time. How are you doing today? Very well, and thanks so much for having me.
00:01:27
Speaker
Pretty happy to be here. Yes. It's been a long time in the making, but Cloud,
John Atsopardi's Journey
00:01:31
Speaker
we're here. um I suppose first off, Jonny, it'd be great to get a bit of an intro on your yourself and your backgrounds into leading up to driving analytics engineering at Monzo. Yeah, sure. My own background is one of professional services. So I was a data consultant in the past, mostly within the big four. So I've worked at a couple of the big four firms before.
00:01:55
Speaker
And I'd kind of been working in the banking space or the financial services space for a while, mostly doing risk analytics things. I was always interested in sort of fintech and the overlap between fintech and data. I was sort of an early adopter of Monzo, myself, but and sort of saw it was quite exciting, quite innovative product and potentially disrupting the the banking industry. And sort of move across really in early 2019.
00:02:25
Speaker
And for most of my time at Monzo, I was actually doing data analytics and building out our finance data team for most of that time. So I spent about four years doing that. And just due to the nature of of sort of joining Monzo at a sort of the early stages, it was really a lot of analytics engineering work we had to do in order to build out the finance data estate. So after doing that for a few years, I actually switched across by the year and a half ago, the focus on analytics engineering at Monzo.
00:02:55
Speaker
um And yeah, I'm currently leading a team called Data Orchestration and Platform. And we really own our core data assets and also our data tooling, like BigQuery and Airflow and integration store internal tooling. And so yeah, I've had a few different roles at Monzo, but I'm currently a senior analytics engineering manager. And apart from Platform, I also do quite a bit of work to sort of curate analytics engineering as a discipline.
00:03:22
Speaker
as a whole, I work with two other newly promoted senior analytics engineering managers as well. So it's not just me, but we also have sort of the EEM community at Monzo. Nice. And I suppose for the audience, how how many of these engineers build up this this community ah at Monzo?
Growth of Monzo's Data Team
00:03:41
Speaker
grown over the years, and we're currently the total size of data or data practitioners of Monzo is about 200 people. That covers data science, analytics, engineering, ML, and data analysts as well. So of that, AE is a quarter, so we're about 50 people today. And growing quickly, it will likely be around 60 people sort of by the spring, 2025.
00:04:09
Speaker
Amazing. Oh, that's great to hear. And obviously you mentioned what initially drew you to Monzo, John, was this sort of innovative new sort of way of looking at banking. But what's kept you here all this time and and how's it evolved in the five, six years that you've been with Monzo? Yeah, it really has, like my own role has scaled with the business. So going from, it was less than a million customers when I joined and We had sort of just got our banking license and we're just beginning to do a lending, for example, and it's really been phenomenal growth, sort of autistic growth over the past approaching six years now.
00:04:47
Speaker
And my own role has scaled with that. So starting from doing hands-on IC work, as you have to do if you're working in a small startup, to then actually doing the people management side, recruiting, growing the team.
Integrating Data with Business Strategy
00:05:01
Speaker
So it's it's remained interesting. So it's I always like to say when I get this question, it's a combination of interesting um problems and great people to work with.
00:05:10
Speaker
And those have been the great things that that sort of have kept me at Monzo is the the challenges have scaled with the growth of the business. There's not just number of customers, but also complexity of the products themselves, starting to doing blending and then different kinds of lending and all the sort of the multiplier effect and knock on impact of that. But yeah, and then I would also say that Monzo also supports sort of internal mobility. So I've been been very, very lucky to make the most of that and having a few different roles within data at Monzo as well.
00:05:39
Speaker
Brilliant. sounds like ah Sounds like a fantastic community to be a part of and lots of opportunity to work across different areas of data on varying complexities of of top projects. So obviously today we're going to be talking about analytics engineering specifically and and really how analytics engineering is scaled at Monzo and what it's like. So I think let's start at the beginning. How and why was was analytics engineering started at at Monzo?
00:06:07
Speaker
yeah so Analytics engineering is really something we call a job family within the data discipline at Monso. Maybe to start at the beginning, I should explain what we mean by discipline. so At Monso, the technical capabilities are effectively matrixed. We bring data skills to different parts of the org.
00:06:27
Speaker
and the horizontal in that is effectively that we bring analytics engineering or data science or ML skills to, for example, our core product, or we bring that to the borrowing part of the business. And so in Monzo language, we call those different departments, we call them collectives in Monzo, and we call the actual skill set that we bring to those areas to discipline.
00:06:54
Speaker
And in the early days of Monzo, we had sort of one or two people embedded in each of the key areas and each of the key collectives. And they were doing, shall we say, covering the entire data stack, like doing a little bit of data modeling, doing a bit of data analysis, sometimes doing some data science, sometimes some ML. And as we've scaled, we basically noticed we needed a little bit more specialization in those domains. And so within data as a whole, we set up sort of these job families within the data discipline.
00:07:23
Speaker
And Analytics Engineering is one of them that is focused and and our role is to ensure we have a complete and accurate data warehouse that gives us the right data assets that we need to enable the business to meet its goals and specifically the specific collectives to meet their goals.
00:07:42
Speaker
Perfect. So how has that changed over time with your with with your new growth and and how has the core structure of analytics engineering, how do you set them up for success?
Focus on Data Modeling and Expansion
00:07:54
Speaker
Yeah, so we started, I should mention, we started analytics engineering as a job family in late 2020 and early 2021. And it's kind of grown in two growth spots really. The first in towards the end of 2021 and throughout 2022. And then actually more recently over this summer and autumn from the first AE was actually embedded in my team in finance at the time.
00:08:21
Speaker
with the remit of of so having somebody focus just on data modeling. The idea there was to have a dedicated person to really curate a data estate and be so focused on that. Sort of um a more modern version of a BI engineer, if you like. And that then frees up and enables data analysts and data scientists to be more impactful and and have good data to work with to make an impact. Now, again, historically data analysts and scientists were doing their own data modeling.
00:08:50
Speaker
and And they still do at Monzo. But fundamentally, we set up the team for success by linking the AE strategy and the analytics engineering estate to the goals of the wider data group and to to the goals of the area of the business at the time. So what that really means is we have a strategy for AE in each part of the business. In other words, taking stock of what are the key data assets and the key tables, the key data sources for that area, and come up effectively with a plan for where are we today and where do we need to get to to enable further growth. Typically, it's always linked to some kind of business growth, like launching new products. Like a few years ago, we launched, for example, Monsoflex, which incurs a lot of knock-on changes to the data warehouse.
00:09:40
Speaker
It's also no secret now that we've been quite open about it, that we're looking to grow internationally as well and expand to other other jurisdictions as a big impact on the data modeling approach that we take. So that all translates to ah refactoring and remodeling and making changes to the data models in various parts of the business. So typically the AE strategy and the AE work we need to do is very linked to the business strategy.
00:10:08
Speaker
That's brilliant. When you're doing this remodeling and this refactoring because of these expansion projects, is an analytics engineer at Monzo guided on what to do or do you give them freedom to come up with their own sort of
Empowering Engineers with Autonomy
00:10:20
Speaker
solutions? what What's the process for implementing new models? Yeah, so typically the overall business direction comes from the top or comes from the overall strategy business strategy. ah But then how that translates to what happens on the ground is very much a key part of performance and key parts of each.
00:10:40
Speaker
analytics engineer's role. So no, it's not it's not completely prescriptive. Each of the teams and each of the squads will take an overall business direction and come up with their own plan. So it's kind of like a top-down meets bottom-up kind of approach to but to to planning. And then that trickles down to everybody's personal of plans as well as engineers. Every engineer will have a ah personal development plan, for example, for the next typically six months. It's very much linked to our financial halves, H1 and H2.
00:11:10
Speaker
And it's really important actually that engineers work closely with the managers and team leads to have a really clear plan for what they're working on and the projects they're contributing to. Then ultimately, we have quite a lot of autonomy for analytics engineers on the ground. They'll be working typically within a larger project team, particularly closely with backend engineering, which is a really really important part of analytics engineering engineer's role, as well as with product and maybe in some cases, also sort of domain specialists like finance analysts or borrowing analysts. And so, and so yes, analytics engineers kind of bridge the gap between, between those worlds and a really important voice in proposing what are the right models to build? What are the events we need from our backend systems? This is all part of what it means to be an analytics engineer. So I would say the way that we work and what
00:12:02
Speaker
What a good analytics engineer at Monzo looks like is someone who's able to really traverse those worlds. So it's not particularly prescriptive. Nobody's going to tell you exactly what to build and please pick up these tickets. It's rather being able to navigate ambiguity. That's a really important marker for success at Monzo. And sort of our best engineers are those who can take a somewhat vague requirement and break it down and figure out exactly what needs to be built and then build, actually technically build a good reliable and scalable solution.
00:12:32
Speaker
That's brilliant to hear because I know some of the reservations we're moving into larger data teams and organizations in general, you know, I'm just going to be siloed and could be very prescriptive prescriptive with decisions made further up top. It sounds like Monso has that autonomy where they empower their engineers to be excellent technically, but also able to navigate the organization and and really hunt out them the problems and architect the right solution to to solve them.
00:12:56
Speaker
Yeah, one thing I would say is, I mean, we've been, I guess, lucky the business has done well and has grown so quickly, you know, we're now at 10 and a half million customers and across quite a diverse product set as well. And I think that's evidence. In fact, we've kept shipping, actually, lots of new products that we've shipped and are are are yet planned. You know, things like I mentioned, Montsoflex earlier, we also have savings products.
00:13:20
Speaker
pensions that we've announced as well. so The investment I've seen on on my app as well? Yes, exactly. so The product mist is ever-expanding, and that gives us a very natural stream of work as well to for us as analytics engineers to to to support on. so All of those products need good data. like We need good metrics about what's the engagement of those products, of the profitability of those products.
00:13:45
Speaker
What are the risks, especially with things like lending? You know, we have to beat the regulations as well and do good reporting and be really on top of our controls. That's all data driven at Bonzo. And so the role of analytics engineering is really to power all of that by having good data assets in our warehouse and give us reliable data about the business.
00:14:04
Speaker
Brilliant. Well, it's been fascinating to hear, up I suppose, how it's set up, how what you look for in analytics engineers and and who succeeds, and clearly the type of exciting projects, which I know from my time in speaking to people that building, innovating, problem solving is is definitely the key things that I know people are looking out for and on their job pattern. What are some of the biggest challenges that you face though at Monzo as analytics engineers?
Challenges in Scaling Analytics Engineering
00:14:31
Speaker
I mean, one challenge is literally what we're just talking about, keeping up with the pace of change, keeping up with the ambitions of the business, and that that has included the people side as well, actually scaling the discipline itself, and then ensuring we have, you know,
00:14:49
Speaker
really good doable plan for each areas of the business. So the first challenge is getting the right people in the door and getting the team staffed. It also links to our data strategy as well, being really concrete about what are the data products we're supporting. That's actually a non-trivial exercise to actually prioritize, but especially when maybe a bit earlier in our life When we were headcount constrained, we needed to support the growth of the business with a relatively small AE team. So we had to be really concrete about what we're taking on.
00:15:22
Speaker
So defining the data products and really prioritizing sometimes ruthlessly on that. And that's still important. We just have a new data strategy really that we worked on over the summer, which was all about moving away from sort of reactive work to being really focused on what are the key data products in each part of the business, defining those and then building a planner on those. That's the first challenge with scaling is is being really deliberate about what you're working on. But it's very easy again to be overwhelmed by abundant and reactive work.
00:15:52
Speaker
I suppose that links into your ability, why you need your analytics engineers to be sort of good, have good commercial and business acumen to be able to prioritize their own workloads and spot where it's going to be the highest value projects versus is the the time put in and what's really going to move the needle. Yes, no, it it always comes back to being connected and having a really good finger on the pulse of priorities. Any person who's worked at the fast moving scale up will probably resonate with them.
00:16:18
Speaker
of of really being needing to manage yourself, manage your time as an and and as an in any engineer. The other big part of a challenge of scaling analytics engineering is scaling the warehouse itself. These are the actual data models. I'm talking about not just the volumes of transactions and the volumes of the data models themselves. um Again, approaching 10.5 million customers, you start to encounter large data volume problems, the transactional tables that run into several billions. And so doing things like know sort of doing a full refresh of a model doesn't become so straightforward anymore. You need to have tooling in place or you start to encounter sharp edges with tools like Airflow and dbt. And so the tooling part of the equation becomes increasingly important.
00:17:07
Speaker
But also the way of data modeling um becomes becomes more and more important that you know you're incremented by default, for example, when you start to encounter those kinds of data volumes.
Developing Internal Tools for Data Models
00:17:17
Speaker
So data volumes is one of them and the modeling architecture that ah sort of those sort of volumes imply.
00:17:23
Speaker
um The second is the complexity of the business itself. It kind of comes back to i guess the first scale of of being really clear what you're working on. Banking is hard. There's no way around that. Just the complexity of data points you need to track and need to be on top of. Everything from financial risks, liquidity in capital, to understanding product uptake and product engagement. It's ah it's a complex complex space and that means We've got a lot of data models, effectively. Today, it's close to 8,000 data models in our nightly run. That presents all sorts of challenges just on its own of maintaining those data polls, scheduling those data polls, as well as the performance and within a cloud context of getting the data reliably refreshed every night.
00:18:09
Speaker
Hell of a lot of data models. you John, you mentioned there about scheduling your your models. I do remember you when we spoke previously, you mentioned that you've been building some quite innovative tool in-house based off of sort of scheduling and being able to orchestrate and schedule your your models and your pipelines by business priority. Is that a project you're able to share a bit more on? This is something we're working on right now, and to be totally clear, it hasn't yet concluded. But what we're effectively exploring is really making the most of of Airflow, but also having some in-house logic around it. And you know we're we're using DBT Core and not DBT Cloud and integrating it with our own guess internal internal backend systems. What we've done is effectively over the the summer really we've tagged really important business use cases and sort of tracking needed delivery times for key models. So we call these exposures.
00:19:08
Speaker
internally and some in-house tooling that we've built to to actually track and monitor the health and state of our of our pipelines for those key use cases and key exposures. What we're currently exploring, and I guess as I should say it's not to so not completed yet, is using some of that metadata on sort of those use cases and when we need models to inform prioritization in Airflow and have sort of custom prioritization methods, which is based on business priority and not just on sort of standard airflow parameters like like naively, the number of downstreams you have. That allows us to do sort of more intelligent orchestration and more intelligent prioritization of our platform. So again, we're currently running some experiments on that, but yeah, it's been quite exciting. I think the direction in which we're kind of heading is having a bit more intelligent orchestration.
00:19:58
Speaker
I think it just shows the the innovation that's constantly going on in the team. And you know it's a a project which I definitely haven't haven't heard of elsewhere. And I think yeah it shows that you're constantly trying to find ways to to improve, but improve that's going to benefit the business, which I think is obviously, you've been really clear that the whole way is to, we always focus in on on what the business strategy is, which is is great to hear.
Balancing Speed and Quality in Scaling
00:20:21
Speaker
I would also emphasize though a really important part of scaling is also doing the basics really well at scale and consistently. I think this is also to some extent as harder than it sounds and has been a challenge, especially when you scale an AE discipline very rapidly, you know first going from 1 to 20, 20 to 40, 40 to 50 people as ensuring we're following sort of best practice everywhere and we're consistent in all and our architecture in different parts of the business, especially across 8,000 models.
00:20:54
Speaker
So that's also, I mean, obviously we have, you know, training documentation, good internal reference documentation is really important, but that's also something we're currently addressing and kind of having a more opinionated platform and moving beyond the standard transformation steps to be more explicit about what are the different design patterns that we follow architecture put in analytics engineering. That's also something that's in flight at the moment.
00:21:19
Speaker
And what it looks like is you know we currently have three broad stages of transformation so kind of have your resources we call currently something called base models which is what we get from the back end we then have ah sort of a middle transformation which we call staging.
00:21:34
Speaker
and the final presentation layer, which we call the entity layer. And those three layers we've found um have over the years, we've got different examples of each. So just three, and you know, three key layers isn't quite enough to capture all the different use cases. And so what we're currently working on is defining the different, like having a, and basically a larger and more opinionated different types of of layers so we can have a really consistent way of of having consistent patterns, because there are different types of entities you can have, whether it's supporting a regulatory model, for example.
00:22:07
Speaker
and therefore fairly, fairly normalized, whether it's supporting a visualization tool and is doing a lot of denormalization at that like at that level. So yeah, again, this is also a work in progress, but I just want to call out that's also been sort of a bit of a challenge of scaling is consistency in your approach and your methodology of doing data modeling at scale.
00:22:27
Speaker
I suppose it sounds like a huge problem and it's also one that takes up a lot of engineers' time when they're looking at redesign and re-architecting and following that that change. How do you guys manage that sort of trade-off between speed of delivery and and I suppose scalability and quality of the work that you're building because that's often start-up, scale-up life, let's go, go, go. but you need to make sure obviously everything's going to be there for the the foreseeable with cost optimization, performance optimization. so how How do you guys tackle that as analytics engineers? It's a good question. i mean An element of autonomy in all of the teams is always present.
00:23:05
Speaker
But then married to that autonomy is the the way that we've tried to solve this is provide those teams the best tooling that we can to allow them to manage their own estate as well as possible. So I mentioned exposures, for example, earlier. It's a relatively, relatively recent tool, for example, that we've shipped specifically for that purpose.
00:23:27
Speaker
But there's also other observability tooling that we've provided to each of those areas. um These could be as simple as you know dashboards that give give a sense of quality for their estate. And we have some some tooling that we call Data Standards at Monzo. And these are effectively, it's a combination of CI checks, a really good CI tooling, plus metrics that also give you insight into what is the general quality of your of your estate.
00:23:57
Speaker
And these are things like, you know how many of your base models are incremental? Have we got the basic amount of testing, like primary key tests? you know And what's the adoption of those? It's those kinds of metrics that that give you a sense of of measurement, like to measure the data quality in a particular area. And we call this data standards a Monzo. And so that's an example of the kind of tooling that we provide to different parts of the business to allow them to actually measure and stay on top of their areas. So that's a really key part of sort of getting that balance right. So so we don't sort of force teams to sort of do specific things, rather give them the tools to proactively manage their own their own estate. Then there's the, I guess, the foundational tooling itself and the developer experience itself. For for example, dbt at Monzo happens in Docker. So we standardize the experience. So this is a Docker shell, basically, which contains all of the tool you need. That gives us an amount of standardization that so that the developer experience, you know, we can debug easily, for example. So again, there's a bit of a trade-off there.
00:24:56
Speaker
and you know setting up that tooling and and giving engineers to write the right tools that they need to do the job. I mentioned earlier we had some challenges with scaling DPT. DPT itself at Monzo is ah kind of an extended version of standard DPT core. At some point, we actually forked DPT, but now we have sort of an extension platform that we've built where we you know can insert our own code.
00:25:21
Speaker
Into the dbt model itself to some extent coupled to some of dbt internals but that sort of runs in that documents that i mentioned earlier so again there's a trade off there of choosing to do that choosing to go down that road, but that's not really a general approach i mentioned some of the the other things in flight like exposures to have good data,
00:25:38
Speaker
and architecture modeling to to define our design patterns is all set in the space of of making the right trade-offs between scalability and correctness versus speed of delivery. That's amazing. It sounds like you you really give the the team and the yeah all the analytics engineers every tall bits of of tooling and every bits of of opportunity to really help them manage their own site and set them up for for success, which is great to hear because it's often you know these challenges compound if if not kept on on top of and it's not something that I think everyone is always aware of when they're they're working day to day and they're focused in on on hitting targets and hitting deadlines. So great to hear that they have that opportunity and have the tooling to help guide them guide them on that.
00:26:26
Speaker
Just to say, I would also say it's also linked to the performance measurement and the performance plans that we have for engineers as well. like So we give we give engineers the autonomy to determine these things for themselves, and then we also recognize and reward those engineers who really are proactive in managing their estate and working closely with the managers for those areas to stay on top of their estate.
00:26:48
Speaker
So it sounded like they got that constant sort of growth and development and helping enabling them to become the the best engineers, the best data professionals that they can be. That's right, yeah. But they have to be very proactive in that. thats That's great to
Keys to Success at Monzo
00:27:01
Speaker
hear. I suppose, look, we're coming towards the end, Jon, but Monzo's AE team is, um you know, highly respected across the the industry and it's probably one of the the most well-known AE teams in in the UK. You've touched upon a few of them, but you know what are your team's non-negotiables when it comes to best practices and I suppose the the skills that you look for in your in your analytics engineers? Best way I can frame it is there are certain so are certain things that are table stakes and then there are other things that really take you to the next level. So the table stakes here are
00:27:37
Speaker
good data modeling, SQL as your bread and butter and things you do day in, day out, not just writing good SQL or scripting, but actually architecting data models that can scale. But the the real key enabler for growth is analytics engineers who can do that and do that reliably, but at the same time understand the business, work closely with, especially in those in parts of AEs who are embedded in specific areas that have domain context and are able to take business problems and break them down into milestones and break them down into things that we have to do in order to support that growth. In other words, navigating ambiguity is one of the really, really key skills and then translating that into reliable and good data modeling and technical best practice. So I would say the technical part is kind of the inner core
00:28:32
Speaker
And the non-negotiable part, the bit that really separates great candidates from okay ones are those who have the communication and collaboration skills to work really well with our key stakeholders. And for EE, we effectively navigate two really important key stakeholders. um I guess on the left-hand side, on the technical side, it's engineering.
00:28:56
Speaker
So we're looking for analytics engineers who are capable of speaking the same language and have the technical understanding to navigate the backend. to have like really good conversations with engineers on what's the source data that we get over a product or feature. Are those complete? Do they work for engineering, but not not so well for for data? Can you advocate for that and then back change? so On the right-hand side, we then have the business itself. like These would be data scientists, perhaps running experiments. They could be elements of the business who want to answer a business question. But the ultimate answer, if you're able to leverage your technical skills with good communication,
00:29:32
Speaker
and good collaboration skills, then you will deliver business impact. That's also, at Monzo, the key differentiator in how we measure performance, specifically what impact has a person had on the business. That can sometimes sound ambiguous, but what we mean what I mean by impact is what difference have you made.
00:29:48
Speaker
really for your area or for your business. How have you moved the needle? how few you know What dial is that? Is that been cost saving? Is that been you know enabling a data scientist to do their work X times quicker or saved X amount of hours? I think you as ah as an analytics engineer, as any data professional, you do need to be able to articulate how you have had impact, what has been your your value. Exactly. And it's really hard to have a one. There's no one size fits all either. So for example, if we talk technical skills,
00:30:18
Speaker
Like we have um different really high performing AEs of Monso who have spikes in different areas. Some are really strong at the data modeling, dbt side of things. Others are stronger maybe in Python or have sort of data engineering background and can also navigate the backend really well. And success is when they can leverage those skills to have an impact on the ground. Sometimes that means joining the right team as well. Like someone can really thrive in a certain context and find another area more challenging. That's also something really important with a good sort of squad to engineer fit as well. But really good engineers do have the sort of intellectual flexibility to pick up new tooling as well, pick up new skills, and then the communication skills to to work well with with but the different disciplines. of onee
00:31:02
Speaker
Yeah, I think we touched you touched on it there, obviously, it's really core data modeling skills and DBT is something that I've seen, you know especially newer analytics engineers to the industry, they but all all they know is DBT and although they think analytics engineering is DBT without maybe some of that foundational knowledge on on data modeling. how How do you guys approach data modeling and what's your sort of stance on that, I suppose, foundational knowledge versus DBT?
00:31:31
Speaker
usage and and knowledge. It's less married as as someone might think. So yes, we use DPT quite extensively and it's our fundamental ETL and transformation tool we use. But my perspective on it is the tool we've chosen at the moment. There's no guarantee and there was a time when we we didn't actually use it. And there might be a day when we move on to whatever is whatever is the right tool for the job in the future. um I mentioned our architecture modeling project earlier.
00:31:59
Speaker
We were actively exploring like internal tooling to support early stages of the transformation. Again, this is work in progress, but in that sort of world, it won't be DBT doing everything as part of the transformation lifecycle. and So the important thing for engineer for analytics engineers is that we really understand the the foundations and fundamentally the principles of the data modeling, whether it's Kimball or whether it's it's something else. The fact that you've got the right sensitivities and ultimately born a little bit by experience, but also by being actually trying trying out and being familiar with different tools is the principles that are far more important than the tools themselves, because the landscape does change.
00:32:39
Speaker
um That's that one thing is for certain is that the tools become more commodified. And I think as analytics engineering, as a general discipline, approaches more and more some of the best practices of software engineering, which I think is an overall trend that will continue to observe, I think the tools will continue to evolve. And so having that an intellectual flexibility, as I mentioned, and also the curiosity to try new things, is far more important than knowing a single tool really, really well.
00:33:07
Speaker
The same is true for orchestration. you know We use Airflow internally today. It does the job admirably, but we're totally open-minded about having alternative tools or in-house tools or whatever it might be in the future.
00:33:19
Speaker
No, so I suppose it's underlying that you know foundational knowledge of understanding the mechanisms of data modeling or orchestration of whatever area of data you're doing is far outweighs the the experience with ah with a given tool.
Continuous Improvement and Adaptation
00:33:34
Speaker
And then when you combine that with a yeah with an individual that is able to identify, prioritize, and work with stakeholders, you get a really really potent mix. And I imagine that's the type of people that you you want in your team.
00:33:47
Speaker
Exactly that. I mean i mentioned the performance earlier. so We fundamentally, as I mentioned, we measure impact. What difference has a person made? As you mentioned, have they moved a needle in their squad or their part of the business? and Then we measure that ah that impact on two axes.
00:34:03
Speaker
The first axis is delivery. What projects have they supported and what work have they delivered? And the second one is equally important, which is the behaviors, which is what behaviors have they demonstrated. You can have a brilliant engineer, but it's all equally important that the colleagues really enjoy working with them. And they work collaboratively in a supportive manner, for example, to more junior engineers as well. So it's all about the combination of technical skills and behaviors.
00:34:31
Speaker
As mentioned earlier, they're really important to make up a well-rounded engineer. Candidates often ask about sort of technical specifics on some of Monsieur's tooling. and we are generally quite transparent about that on our blog, which is available on our media. And so if you just search for Monzo data blog, you'll find a reference to that. And in that blog, we go a fair bit deeper in some of the specifics of some of the things I've been talking about and on our tooling and how we've actually set up and scaled the data and analytics engineering team at Monzo.
00:35:05
Speaker
and includes things like our approach to incrementalization, for example, um and how we scale DPT itself, topics just like that. Brilliant. Well, and we we can put a link in the in the podcast notes as well for anyone that wants to to go over and check out anything that John's spoken about in in a bit more depth.
00:35:23
Speaker
I suppose closing thoughts and comments, Jon, what is next for Monzo and Analytics Engineering?
Future Projects and Opportunities
00:35:31
Speaker
Looking overhead over the next year, what are going to be some of your the most sort of excited projects that you're keen to work on and and where are you going? Yeah, again, it always links back to the overall business strategy. I mentioned earlier, we've just gone through our second growth spurt in the discipline and we're approaching now a total number of 60 by the spring analytics engineers.
00:35:53
Speaker
So the first thing in mind is the people who have already accepted and yet to join one. So, you know, we need to onboard them well. It is a significant change, a significant number of people will be growing by. So it's embedding all of those new engineers successfully into the business, especially early in the year, I think it's going to be really important. And then there are the business priorities, new products that are coming, potential internationalization, expansion, it's going to be a really important business theme for us. And then I already mentioned some of the other projects we have in architecture modeling and design patterns that we have.
00:36:27
Speaker
and embedding that change across the entire discipline. Those are some really, really important themes that we're going to be working on in 2025. Exciting, exciting times ahead. Well, look, John, thank you so much for your time. It's it's been a real lesson, I suppose, on how analytics engineering has evolved for Monzo. And I suppose to hear some of the insights of why the team's so high performing and so well regarded in the UK. And I suppose best of luck moving into the new year. Thanks so much. And thank you for
Wrap-up and Call to Action
00:36:59
Speaker
having me.
00:36:59
Speaker
No worries at all. um Feel free to connect with with John and yeah, I'm sure by the time this goes live, I'm sure Monzo will still be hiring the the last stages of their of their team. So feel free to to reach out to to any of their talent team. I'm sure they'll be keen to keen to speak to you. Thank you, Harry. Cheers, John.
00:37:19
Speaker
Well, that's it for this week. Thank you so, so much for tuning in. I really hope you've learned something. I know I have. The Stack Podcast aims to share real journeys and lessons that empower you and the entire community. Together, we aim to unlock new perspectives and overcome challenges in the ever evolving landscape of modern data.
00:37:41
Speaker
Today's episode was brought to you by Cognify, the recruitment partner for modern data teams. If you've enjoyed today's episode, hit that follow button to stay updated with our latest releases. More importantly, if you believe this episode could benefit someone you know, please share it with them. We're always on the lookout for new guests who have inspiring stories and valuable lessons to share with our community.
00:38:03
Speaker
If you or someone you know fits that pill, please don't hesitate to reach out. I've been Harry Gollop from Cognify, your host and guide on this data-driven journey. Until next time, over and out.