Listen in on the expert panel we hosted that explored data governance within the innovative framework of data mesh. We discussed strategies for ensuring data quality, security, and compliance while fostering decentralized ownership. Learn how organizations can balance autonomy and oversight to create a scalable, efficient data governance model in a distributed environment.
Speakers:
Karin Håkansson – Data Governance/Data Mesh Lead Consultant
Kinda El Maarry – PhD Director Data Governance & Business Intelligence at Prima
Maarten Masschelein – CEO and co-founder of SODA
Watch the Replay
Read the Transcript
Speaker 1: 00:02 All right.
Speaker 2: 00:04 Okay, we are live. Thanks everybody for joining us for another data mesh learning panel. I’m Paul Head of community for data mesh learning community. Today you’re here for data governance and data mesh panel. We have a really exciting list of panelists. Before we get into that, I’m just going to go over some things that are coming up. So first, next week on Tuesday we have the data mesh learning monthly round table, and the topic of discussion will be role of humans versus technology and automation. So yeah, if you haven’t been to one of those before, these are interactive discussions. So we break out into breakout groups and we talk about a few questions related to the topic of the day. So those are really fun. Andrew will be there, Carl will be there. And then of course also we have Amy Regatta and Juan Rossier. If you can join us, you should do that next Tuesday.
01:07 Then later in the month on August 28th, Amy Regatta, she’s has a new series monthly interview series called Data Chronicles. And so we had Yulia Chocho last week or last month. And then this month we’ll have Andrew Jones, the principal engineer from GoCardless. So check that out later in the month. And then of course we post all of our talks to YouTube. So if you miss them including this one or you freely want to watch it again, you can go to YouTube and watch all of our recorded sessions. The only sessions that we don’t record are the monthly round tables because those are more interactive, but all of our sort of webinars are posted there and we will also coming soon be posting audio of our talks to Spotify. So there’ll just be another way for you to consume this material. If you want to have it on the background or don’t need the video, you can come check it out on Spotify.
02:07 Then of course, if you want to get in on the conversations, we have a pretty large Slack group, so if you scan that QR code, that’ll get you the invite to the data mesh learning Slack group. You can interact with peers, ask questions, and then also help answer questions that people have about their data mesh journey. So definitely check that out if you aren’t already part of the Slack group. And then last but not least, we have the Data Meh learning website where we have hundreds of resources. We list all of our events, lots of use cases, so it’s just a good place to get more information about data mesh. So that’s it for me and I will hand things over to Andrew, our moderator.
Speaker 3: 02:53 Thanks Paul, and good afternoon everybody. I’m really pleased that we’re now on Spotify because that means my daughter might actually start listening to these things. She starts getting very excited when things are on Spotify and she can actually stream them. So this afternoon’s all about data governance in data mesh, and we’ve got a great panel this afternoon to explore the issues that we want to surface. Probably a little bit about me and then I’ll get Karen Kinder and Martin to do the introduction. I’m Andrew Sharp. I’m a principal consultant for a data consultancy based in Leeds in the north of England. We work with a lot of different clients who are trying to deploy data mesh. I’ve done 15 or so years doing data governance, so I’m from the classically trained community of data professionals, so very exciting to explore what that means from a data mesh perspective. So perhaps if I can hand over to Karen who can introduce yourself.
Speaker 4: 04:05 Yes, sure. Hi everyone. My name is Karen Hawkinson and I’m located in Sweden, Stockholm, and I’m self-employed as a consultant. So I work as a consultant in data governance with a focus on data mesh implementations. So my background is I’ve been working a little bit of everything. I started out within business, moved into it, became a software developer and then a data engineer, and then moved on to more manager roles and then in the end ended up in data governance. So yeah, that’s me. Should I hand over to nda maybe?
Speaker 5: 04:49 Thank you Karen e. Well also from my side, I’m Al Maori and I am currently the director of data governance and business intelligence at Prima. Prior to that, I was the data governance lead at HelloFresh. And in both those roles I was lucky to get introduced to data governance as both companies were transitioning towards data mesh, and that’s how I ended up in the governance field. But that does not mean that I haven’t been in that field. So prior to that, I was focused on data quality and I got really excited when data mesh came along because that’s exactly where you want to get the quality controls in the place on that left push. But that’s another topic for another day. I’m pretty excited to be part of this group today and to delve into the questions that Andrew has in store for us. Martin, I leave it to you.
Speaker 6: 05:42 Thanks, KDA. Hi everyone, my name is Martin. I’m currently the co-founder and CEO at Soda. We’re a software company. I’ve actually always been on the software side. It’s been 12 years now, first six were at Collibra where I think Collibra is in the market for really data governance. And one of the reasons why I left there after six years I wanted to start on my own, was that we needed more tools, more engineering tools really to operationalize data governance. And then a year or two later came data mesh, which was right in our wheelhouse, which was fantastic. So yeah, with the experience that I have over the last six years is more building tools for data engineers also in open source. So we get to serve quite a big community, which is fantastic. Some of you might know. So the core is that project that also implements data contracts. I’m sure we’ll get into these things. Very exciting stuff. But yeah, super happy to be here. And Andrew, I’ll let you take it away from here.
Speaker 3: 06:46 Thanks, Martin. What I would say is please, if you’ve got any questions, we encourage you to put them in the respective channels and we will receive those. And we’re obviously asked those questions as we go, or we’ll perhaps do a mop up at the end for anything that’s not been covered. So thank you panel. I suppose the first question really that we want to unpack today is what do you see as the key challenges in implementing federated governance across different domains in a data mesh environment is probably the key question. So maybe start with Karen.
Speaker 4: 07:27 Yeah, the key challenge of implementing, I would say the first challenge is to agree what does federated actually mean? Because for me, I’ve been in different companies and it means different things. So that’s definitely one of the key topics, like first degree on where are you aiming, who decides what and align on that. So that would be the first challenge I would say. But otherwise than that, I think it also depends on where are you coming from, right? So do you have a central data governance team already? And then you will have certain other types of challenges implementing federated. Then if you come from nothing and you want to start to implement federated governance from scratch and you have nothing, and then you will have other challenges. So for me, it’s all about where are you coming from. Yeah, I think that’s
Speaker 3: 08:35 Super useful. And you mentioned obviously that the idea of federated. From your perspective, what would your definition of federated be, Karen?
Speaker 4: 08:48 Well, for me it’s more about like, okay, so you assign the ownership within the different domains and then you get together all the domains and then you together decide on whatever needs to be decided. Then of course you could have a governance supporting that, but the decision actually lies within the together on things that needs to be central and then on things that shouldn’t be central. Then the domains decide themselves. And of course in federated governance, you should really aim for having the central part as small as possible and then try to delegate as much as you can out to the domains. But that of course also depends on a lot of things, what you can move to or what you can move to the domains. So I’m sure we’ll get into that,
Speaker 3: 09:45 But yeah, sure, sure. Martin, no thanks Karen. Martin, what was your thoughts on the general question and what Karen’s also been exploring?
Speaker 6: 09:58 Yeah, I think it was great because with data governance in the past, I think it’s really started more central because we were thinking about policies and things that everyone needs to do around data, what is good practice really with data, and that’s where it started and we kind of pushed down those responsibilities into the different domains, especially in larger companies. So I think that was one move that you saw. And I think what’s happening now, especially with the advent of data engineering and data mesh et cetera, is really also you see these local teams building data products and they have different levels of maturity and they need their own autonomy as well. So we kind of see this nice dichotomy in a way in which there’s a bit of probably a push towards, hey, we want less architecture and less of tools decided for us.
10:55 We want to decide ourselves, but there’s still this very strong need for almost every company to adhere to some enterprise wide things. And the question really has nowadays I think with data measures become how do we do that in a very scalable way? So I think it’s a very logical transition I think that we’re making, and it’s a really exciting time as well because there’s a lot of demand regarding to how do we create tools and systems and mechanisms, processes that are scalable for everyone that are very easy to adopt. They’re not complicated. Data governance often has this, well, it’s extra work, it slows me down, I don’t want to do it. So it’s a very interesting challenge to think about how we can do that in a non-intrusive way, in a way that it’s built into the process and sometimes people don’t even have to know they’re doing good governance.
11:56 It just becomes natural and it’s because we’ve built the great products for everyone in the organization to use. So that’s a bit my 2 cents on it. I think it’s a very good evolution. It’s definitely one that comes with challenges because I think traditionally with data governance, we’ve worked more with IT teams and you had architects for example, helping modeling data, which was very important. We also had some technical support around some of the old tools that we were using that required a lot of support, and now all of a sudden we’re shifting that towards product and engineering teams that help us and there we can really think about how do we design this in the ideal world, which is really exciting in my opinion, but comes with its challenges. We need product managers for example in order to do that. So that’s quite a transition.
Speaker 3: 12:47 Absolutely. No super interesting thoughts there. Kinder, what’s your thoughts on the key challenges for implementing federated governance?
Speaker 5: 12:59 I think Karen and Martin said a lot of hit it home basically how to define federation for instance, because even that within the same company will be different. And the challenge here is, and again that’s to Martin’s point, the different domains that you are giving them this new responsibility, we have different maturity level and that maturity level depends on the composition of their domains, the use cases that they have and a lot of different things that makes up that domain. And so you might find yourself, and this is what I’ve seen in practice, even within the same organization that you’re rolling out data mesh, that you are starting to push more and more governance standards, some domains you’re going to have to be more hands-on and other domains you’re going to find yourself that you can step back earlier in that journey. Obviously at the beginning you always need to make sure that you go ahead of time before you allow anything, make sure that any literacy gaps is bridged So at the beginning.
14:03 But as you move throughout that journey, you’ll see some domains are faster and are more adaptable and they have the right mindset, they have already the people that are needed and you can already step back and go into that fantastic role, which I think is on the other hand, I always call it a golden opportunity for governance because for so long we’ve been seen as the police and because we’ve been isolated. And I think with data mesh, this offers us an opportunity because it’s all about empowering the domains. It’s all about working with them very closely and making sure that they all can be autonomous. And that means that the set of standards that you are producing from the governance side of things, it has to be a thin layer of standardization that works across the board but still gives leeway for the domains to create the standards that they need that fits their use case. And that means you naturally need to be in that conversation and co-creation process. And so you move away from that central entity where all knowledge disseminate and move towards a more collaborative approach. So I’m moving from the challenge and I’m taking us a little bit more to the positive side.
Speaker 3: 15:23 No, absolutely. And I think that you’ve sort of done a really good segue into one of the other key questions that we have in terms of how organizations balance that sort of domain autonomy with the needs for consistent data standards and practices. And I think you’ve started kinder to sort of unpack that there. Maybe you can bring that to life fetish out a little bit more as to how you actually get that balance in organizations, which I’m sure is a tricky challenge.
Speaker 5: 15:57 It is an absolutely tricky challenge in the sense that it takes a data people out of our comfort zone and that is out of our data world and out of our frameworks. And that is you need to speak the language of the business. So you need to deep dive and go to the domains, understand the use cases and co-create with them. And one of the most practical things that you can do is have a working group, have a working group and have different people from the different domains. And this is not the domain list, but the people who are actually doing the work, the product managers, the analysts, the data engineers, and start having these conversations talk about these basic standards and you will hear people then raising their hands saying, but that doesn’t work for our use case. And giving them that platform to speak up and to give feedback and seeing that you’re taking their feedback and you’re tweaking things when there is a possibility, that’s when things starts to work.
16:59 And that’s where there is a shift of we’re not the only entity that is producing obviously the oversight and ensuring that adoption is happening and ensuring that we’re empowering the domains by giving them the onboarding that they need, the trainings and all of that. But it shifts in the sense that they feel like they’re also owning that and they’re also co-creating and they are making the standards and tweaking them to fit their own use cases. And this is when you start seeing that balance between the autonomy and the consistent standard layer that you need to put, but there is obviously an art to it, which standards can you actually say, okay, this is the red line, we cannot actually change that. This is usually when it comes to security and compliance, these are the things where you have to put a hard line and where can you actually give a bit of leeway? And that always depends on what it is you’re actually looking at, but it’s not easy. But I think that’s the interesting side of it and the interesting challenge that’s at least for me is very exciting.
Speaker 3: 18:06 And just on that point about forming a working group and bringing everybody together, is that something that you’ve done in your organization or seen as a vehicle to actually get engagement that feels to me as being quite a critical step. People have to be bought into this and come along to the meetings in order to get those benefits. So I’m guessing that that is quite a key milestone for the success of these sorts of things.
Speaker 5: 18:36 Absolutely. So these working groups definitely have to be bringing value to the people who are attending and it starts bringing value when it’s interactive and it’s not in the sense that we are the governance people. We’re coming, we’re going to tell you what you need to do and what is right and what is wrong, but it is about giving them space and allowing them to tell their stories, what has worked for you, what hasn’t worked for you? We’ve pushed out this particular feature from the platform, tell us criticize it and allowing for that failures to come up to the surface so we can talk about them. So once you create a space where it’s a two way, not even a two way, but it is an X way collaboration and communication between the different domains, that’s when you start bringing value because the domains want to hear from the other domains, how did things go for you when you first onboarded? What were the lessons learned? What were your perhaps workarounds? And once you start creating that space, that’s when you start seeing people engaging more and more.
Speaker 3: 19:44 Yeah, makes sense. Martin, what are your thoughts on this sort of striking the balance and how best to go about getting that balance in an organization? Sorry, may you’re on mute.
Speaker 6: 19:59 I’m sorry. So I think about this is that in general the less kind of overheads, let’s put it this way, or the less centralization, I think I’m ProE doing as limited as possible and if you need to do it, do it with a product mindset mindsets. And what that means is really think about goat, treat the people that you’re building this for as your customer and you’re going to build an experience for them and if they’re not happy you didn’t succeed and do so in a very kind of iterative, we start small, we work on the first principle, don’t make things complicated. We really start with small increments and then what you see is if you treat that person, for example, within the domain, let’s say data platform engineer or a product manager there, you just treat them as your customer and slowly build and iterate and work together. That goes a very, very long way and I would never go in with, oh, these are all the things we now need to start doing. You’re going to lose people in the process. So it’s more my practical minds that will be my, at least my tip around how to do that is really start light on what you need to do centrally and focus on your customer being happy.
Speaker 3: 21:28 Makes perfect sense. This idea of filtering out and working out what the priorities are, is that something that this group needs to actually agree upon or is it something that you actually give them some options? Because I’ve been in meetings where it is a challenge, everybody has a particular perspective, so it’s always challenging to get consensus.
Speaker 6 21:53 Yeah, I think ultimately it’s about what’s important to the business. Is it really about, and I think it could go back to fundamentals of a business case building. Is it about revenue? Is it about cost is about risk. That’s something that executives can answer or should be able to answer or give you as inputs. And then it’s of course going to be dependent if you have a domain team that works on the customer data versus domain team that works on something completely different, of course you’re going to have different priorities and I think your leadership should be able to guide you in the right direction as to what is most important, whether it’s through an OPR framework that they’re using to communicate strategy and objectives or something else. But
Speaker 3: 22:40 That makes sense. We have a question in from Catherine sale who’s, and I’ll pose it to you Martin, but we then involve Karen and Kinder and she’s saying, what is your advice for those setting up a governance function for the first time? What would you do first?
Speaker 6: 23:01 That’s a very good question. I think, so I’ll go back to the analogy, to the tradition to how I’ve seen it been happening and then maybe reflect on that for a second. I think traditionally it’s been about the data governance policy, it’s been about ownership and it’s been about definition. I think those three areas and the policy kind of constitutes how we should be treating data, et cetera. It’s more of a visionary type documents where we want to get to and what good data management practice looks like. Then ownership, because traditionally where we focused, so we had so many differences in even what was a customer, what constitutes a customer, there are so many people who are misaligned, so we didn’t have the basics down, we didn’t even know that there was a different way of defining something in another part of the organization. So for that, you need to have or identify people who can bring this towards such a kind of decision in which you either adjust the terminology or at least you’re just aware of it now and you’ve defined it in both areas. And from that point on you can move forward. I think this is the traditional way of doing it and I’ll probably let Karen and Kindda, you guys are more proficient
Speaker 3: 24:30 Perhaps hand over the conversation to Karen who’s been waiting patiently. Karen, your thoughts.
Speaker 4: 24:38 Yeah, I would say mean these things that you mentioned that constitutes the traditional governance, I still think that it kind of holds, right? You need to do these things as well. So actually if I would set up a new data governance function, I would probably start there. But then again, it depends on what you want to achieve. Also, what’s the point of starting a governance? I mean, what’s the pain point? Because I would actually, usually that is the pain point, so that’s why you start it up, but maybe it’s something else. So I would really first of all, okay, what is the problem that we are trying to fix by having a governance function? Because governance means so many things and they can do so many different things and focus on different things like compliance or as you say, business glossary or whatever it is, but they all solve different problems. So make sure that the problem that you’re trying to solve before setting up a governance function I would say, and then have a priority on that. Yeah,
Speaker 3: 25:51 That’s really good advice. Kinder your thoughts.
Speaker 5: 25:55 I agree with that. The challenge that I see a lot of people in the governance field do is that they go at it in a very, let’s say, not pragmatic way, but it almost feels like there’s a standard roadmap that everyone starts with and it’s completely isolated and disconnected with what the business is trying to do. And so eventually there is this isolation and there’s no way for the governance people to explain why getting a data catalog is of value. And so my advice and what I’ve always done in the past is understand what are the goals of the company, what are the use cases, what are the P ones? You look at one P one and you start connecting that with the data that needs to unlock that. So for instance, one thing could be if you’re in a supply chain company, you want to improve delivery times, you’re looking at supply chain optimization, start thinking, which data do I need for that?
26:54 You probably need the supplier data, you need the shipping data, inventory data. Then dig into that, what are the problems there? And all of those foundational things that you need to bring in from the governance side, whether it’s assigning ownership, whether it is discoverability, that is where you can sneak them in kind of gria tactics there and just focus on that use case. Not rolling out ownership everywhere, not rolling out the data catalog everywhere. Only on that particular use case, show the value and that success story and then take it from there and roll step by step. So I’m not a fan of big bang initiatives. I’m really targeting use cases and that was worked for me in the past.
Speaker 3: 27:42 Yeah, I think that makes a lot of sense. And we have a sort of statement that we use in our organization about this idea of data governance almost by stealth, which is partly under the radar, it’s doing it incrementally, it’s as you say, way marking a use case as a vehicle to demonstrate what you can achieve. So rather than doing a big bang, that’s an approach that’s worked very well and it also plays to the realization that there’s not a limitless pot of money for some of these things, so you have to invest where you’re going to make the most impact. That’s really useful. There’s another question coming from Sean McClintock who’s asking how does organizational culture come into play with establishing data governance? Do you think change management and a focus on culture is important? So perhaps we’re start with yourself kinder and go back the way?
Speaker 5: 28:48 Yes, it’s absolutely vital. The reason for that is you need the people on your side. You’ll either have people on your side or you’re going to have pushback. That all has to do with that change management specifically if we are talking about data mesh where there is a lot of change, there is a new responsibilities that are coming people’s way, there are a lot of question marks. What does it mean for me to own the data? What does it mean to produce a data product? And all of that has to be addressed. And the fundamental thing is why are we actually doing this? That’s one thing that we need to address. The second thing is understanding where everyone is on their journey and how are they ready and how we can empower them to reach that level where they feel comfortable and empowered to start doing the things that we’re asked to do.
29:42 And this is from the literacy side of things, from the budget that a particular domain has to get new people coming in perhaps who weren’t there before. It’s about providing the technology, the data platform and the tooling that allows them to do the things as in the lightweight as possible. So the whole culture and the change management is a huge part of it and there’s a lot of strategies that you can put into play to make sure that you have the right mindset there. And that’s whether with through data literacy program, whether it is by creating some kind of events that brings people on board with what it is that you’re pushing. So lots of things that you can do and it’s extremely important and it has to be a focus point for at least one of the data teams that are pushing to for That
Speaker 3: 30:37 Makes sense. Karen, your thoughts on culture and what role it plays?
Speaker 4: 30:45 Yeah, a lot. I would say I completely agree with kind there definitely. I mean the culture is everything. Culture eats strategy for breakfast or whatever the saying is, but governance, it’s all about people communication and change management. I mean that’s like 90% of the work is there and then the 10% of the work is somewhere else because when you have people with you, you have the culture is working for you instead of against you, then you can actually make things happen. So I completely agree with Kendall here.
Speaker 3: 31:26 Perfect. Perfect. Martin, have we got three violent agreements or have you a different point of view?
Speaker 6: 31:35 I was more thinking about giving a tip really than anything else because of course it is extremely important. I think the tip would probably be to think about your organization as it is today and how that’s different from others. What are some cultural traits of the organization that you’re in and how will that impact your change management rollout and what do you need to do actually to make sure that you do it in a culturally compliant way? And so having just in the beginning, think a little bit about the culture that you’re in, try to compare and contrast that and then take that with you on your journey.
Speaker 3: 32:15 Makes a lot of sense. P Smits has made an interesting point here that they’re saying that data governance is also a company-wide sort of case. So in the instance of data mesh, should this be a sort of central instance of the company plus the entities in each domain, if that makes sense. So it’s almost bolting on governance on top of what’s already there I think. And any thoughts on that one?
Speaker 4: 32:48 So the question is really like, okay, so do we do governance for the rest as well or
Speaker 3: 32:54 Yes, should you do a central instance plus the entities in each domain? Is that an approach? Have you seen that work?
Speaker 6: 33:06 Isn’t that the federated model? In a way federation is like you have some sort a little bit central, but then a lot of freedom and you try to push down to everyone else. That’s how that was my definition maybe of federated works. Well,
Speaker 4: 33:21 No, it what it’s,
Speaker 3: 33:22 Yeah, and I think that goes full circle to what Karen was saying, being clear on it. But yeah, arguably in some organizations we’ve worked in, you have a central core and then the entities flex that. So is that not federation? So interesting point. Kinder, any thoughts from yourself?
Speaker 5: 33:45 I was holding back because I wasn’t sure I understood the question because in fact it is what I interpret as federation. So you definitely need some kind of a central entity, but it is not probably as big as it used to be because it’s only there to act as a guidance, but it’s not the entity that’s doing the heavy lifting, so they heavy lifting and the extended arms is in fact in the domains.
Speaker 3: 34:09 No, that makes sense. Slightly pivoting away from what we’ve been talking about there, Robin. Adam is asking in his experience, a lot of effort is required to execute governance within the domains and how far are we with computational federated governance and he is wondering whether there’s any good examples or stories people might have having deployed that to overcome some of these issues?
Speaker 4: 34:44 Yeah, there are definitely examples out there. I would say there are. I mean it’s moving there. I think in mean of course it depends of the company, but there are good examples of where you can actually see computational governance happening. Yeah, go
Speaker 3: 35:07 Ahead. Yeah, sorry Martin, are you seeing it in your area?
Speaker 6: 35:12 Oh sure. I think so. Computational, when I first read, I was like, oh, that’s such a complicated way of saying what we’re actually going to do. What we’re doing is we’re building some form of automations or ways to do this at scale. So we’re kind of building products for data management. And I think the best possible example, and that’s why there’s a lot of FUS about it, is data contracts. I think that’s a very good example because in a data contract you specify what the data looks like, maybe some SLOs, maybe some you extended it to who can have access to that or their access requests are granted by default. If you’re in this, this and that group or so that is computational at work really contract is always executable and it automates something. If you change the configuration like, oh, this group now has access to, we make sure that isn’t enforced on your data source. You don’t have to think about that anymore. Or if you say, well, you add a data quality check on it, well it’s done, it’s enforced. You can even stop your pipelines when it’s not as expected. So I think in that realm, data contracts is probably very good example of how that is happening in the fields. And I think both of you, Karen, and can have experience doing this I think as well.
Speaker 3: 36:41 Yeah, perhaps hand it over to kinder in terms of her experience of computation or federated governance and what you are seeing.
Speaker 5: 36:51 I would say I feel the pain Robin, and actually I’m a little bit not pessimistic, but I’ve been feeling that it’s been a bit slower than what I personally would want to see when it comes to automating. So data contracts, I’m extremely happy over the past two years we’re seeing a huge push in many successful implementations from Andrew Jones at HelloFresh Gable with at Sanderson Soda now with the data content, extremely happy. Something that we’ve been working on since 2018 I think. So I’m happy that now we’re seeing some automation there. I think we are definitely heading in the right direction and I’m seeing more and more people in the industry pushing towards having the right infrastructure that enables companies to move quicker in their journey. And I’m still seeing a lot of companies struggle trying to build their own data platform just because there is this gap in the market. And so I am optimistic that we’re going to head there, but I understand the pain and I think a lot of people are going through that pain. Some areas we are much more advanced like with the data quality and with data contracts, but in other areas I think we still have room for improvement and I’m seeing a lot of improvements and talking with the people of the tooling. I’m seeing that in the roadmaps, but I feel the pain.
Speaker 3: 38:21 Yeah, that’s a really good segue probably to one of the other questions that we’ve got today, which is I think many organizations have a centralized view of data. So how does data governance evolve when you’re transitioning away from a very centralized data architecture to a more data mesh approach? What are the things that you need to think about as you move from maybe that more centralized view of data to a federated one that we’ve been talking about today? Karen, I didn’t know whether you’ve got any thoughts on that.
Speaker 4: 38:59 Yeah, sure. I think one of the things, I mean obviously before you had a central place, you have a central team and then you can always talk to them. When it comes to at least the analytical layer, it’s like, okay, but now everyone can do data products. So lots of the interactions. So of course you want to make data or compliance, security and ethical policies easy to understand. I mean it has to be super easy. So it should always be almost be governance as a product. So that’s the switch I think that really needs to happen when you move to a data mesh. Otherwise it will become a mess. So it’s really, I think that’s the biggest change, but it’s also back to Ken’s point, it’s also an opportunity because now you probably have some kind of platform where you can work with and make, you can actually automate things. So it’s really an opportunity to be more compliant and more secure and more ethical because it can actually automate it. I think that’s, for me, that’s the biggest change going from writing policies to actually making sure that it’s super easy to use.
Speaker 3: 40:27 And I think that’s absolutely a super important point today that if you can sell that benefit, then far more people will buy into and see the value of data governance and what you’re seeking to achieve. So that’s a really key takeout. Martin, your thoughts in regard to the evolution from a center to a more federated approach?
Speaker 6: 40:54 I think so you’re not doing it anymore yourselves, you have to have others do it. So it needs to be very easy to do and very well documented. And I think that the challenge really is that within data governance, it was not necessarily a skill that we were looking for when building data governance teams to build tools for other people in the organization, it’s new. So that is a key, very big challenge to get that knowledge on board and then to also release or bring these tools in the right way to your end users. I think it’s a big shift. It’s a very exciting one. We go from indeed writing policies to help build stuff, which is fantastic. So I see it in a very positive light in the end, but a challenging change nonetheless.
Speaker 3: 41:46 Yeah. Cool. We have an interesting very specific point here from P Smith. So again, they’re saying is a data catalog a must have when you are moving to data mesh? Tinder? There’s a few shaking of heads. Martin shaking his head.
Speaker 5: 42:09 Yeah, I would say yes. I would say yes because of the discoverability side of things. However, for instance now at Primo we still don’t have a data catalog and that is not opposing our journey towards a data mesh. It does support because it brings that discoverability ability. But there are other ways where in places where you can show ownership or you can show the S scheme or you can show the lineage. So I don’t have a data catalog, I can’t start the journey and that I see in a lot of different areas as well. Martin will hate me for this, but also data quality. If you have a data quality tool, fantastic. It boosts efficiency, collaboration, the automation, the scalability, you need it eventually, especially if you want to move towards data contracts, you need that automation to place, but that does not mean you cannot start yet and already bring a lot of value into place. So it’s a yes, but with a disclaimer.
Speaker 3: 43:11 Yeah. Okay. Martin, do you want to respond to that before I hand over to Karen for her thoughts?
Speaker 6: 43:18 I agree. Look, but the reason why I was not shaking my head no is because you might be pulled into doing things that you actually shouldn’t be focused on because a catalog comes from a place where we just index everything and in data mesh, we want to build components that are reusable and we want to make those available, which means if you’re just starting with data mesh, you would have nothing in your data catalog. And if you have a whole bunch of stuff there, well, it’s not a data product. So all I’m saying is you might be pulled in the wrong direction, you might start pulling in a direction of indexing everything you have, building this form of repository of metadata and information. And that’s not necessarily contributing to what you set out to do if you’re on that data mes journey.
Speaker 3: 44:06 Interesting, really interesting. Karen, your thoughts?
Speaker 4: 44:10 Yeah, I agree. Sorry,
Speaker 3: 44:15 Just going to say, is a data catalog a must have or?
Speaker 4: 44:19 Yeah, I agree. I mean maybe you don’t need a data catalog, but you need data discoverability, right? You need that capability somehow. Is it a data catalog? Maybe, but could it be something else as well? Yes. Maybe it’s not the traditional data catalog that you need, but a lot of the data catalog vendors today fill that need. So I am, yeah, but the discoverability part is key in data mesh, I would say in order to really make it scale. But I completely agree you don’t need it in the beginning.
Speaker 3: 44:57 Interesting. Well hopefully that’s some insight for those listening. I’m conscious that we’re getting very close to the end. So the question from Catherine Seal is quite timely because she’s asking the panel, what have you found really didn’t work when introducing or maturing data mesh governance? And she’s asking, what was your biggest mistake and what did you learn from that? So there’s a set of questions. So Karen, do you want to start with yourself, your thoughts on what you’ve learned and what didn’t work?
Speaker 4: 45:38 Yeah, don’t run too fast. I mean it’s likes one of those things for me at least, I mean this is my personal experience I’m sharing here on the things that I did wrong. I read the book, I got all excited. I was like, oh my God, this is so exciting. And then the company has decided, okay, we’re going to do data mesh, but getting people on board getting, and you need to have this, I mean it’s change management, so you need to start reading books like this leading change. That’s what you should do because it’s actually really about change management and it’s about, so don’t run too fast, don’t ask for too much in the beginning. So I did that wrong. Completely wrong in the beginning, so I am learning.
Speaker 3: 46:35 No, that’s spot on. So yeah, all about change management and walk before you can run Martin
Speaker 6: 46:45 Small iterations that create value. If you don’t create value and you’ve been at it for months and months and months, there’s something wrong I think. So the more you can shorten it down and focus on outcomes, I think the better.
Speaker 3: 47:01 Perfect. And any mistakes that you’ve learned from anything that you’ve gone, I wouldn’t do that. Again,
Speaker 6: 47:10 I will say too long iteration, literally the opposite of what I recommended was the mistake
Speaker 5: 47:18 I could add to that. So it’s also small iterations, but also because that I got my hands burned on. That is especially on those first data products where you’re looking at the governance side of things, pick a data product that is simple and by simple I mean try and avoid sensitive data and PI data because you’re already starting with something new. The team that is going on that journey with you is also going out of their comfort zone and you need to put already a lot of complicated things. So that adds a whole layer of complexity that you don’t necessarily need in that first 1, 2, 3 data products. Soon as you have that going, then you can add that layer of complexity. So that’s kind of one of the lessons learned that I had.
Speaker 3: 48:05 No, that’s really good shout and I think there’s some great takeaways there for people who are going through this journey. So that’s spot on. Shout out to those that are listening. If you have any more questions, please fire them on. Now there’s a couple that have come through we’ve not covered, and one of them is from Richard Atkins and he’s talking around data governance is far more than just tooling and we’ve just been talking about cataloging. However, has anybody had any good or bad experience of GCP data plex? So that’s a very specific bit of technology, so I dunno whether any of the panelists have actually used that bit of kit.
Speaker 4: 48:52 Yeah, I have used it.
Speaker 3: 48:57 Good. Bad leading question.
Speaker 4: 49:04 Yeah, I would say it depends on what you want to achieve and also depends on your tech stack. And I mean are you just staying in GCP or just what’s the overall strategy for your implementation? If you’re looking for something, no, DataFlex is not the thing to go to, I would say. But if you’re having just GCP just that and it’s a small implementation, maybe that’s the thing to go to, but otherwise I would go for something
Speaker 3: 49:43 Else. Yeah, I mean it’s difficult. We can’t obviously quiz Richard, but it sounds like if he wants to unpack that some more, Karen, I’m sure you’d be happy to. Yeah, sure. Yeah, unpack that with him after the session today. Martin Kinder, any thoughts of GCP data? Plex
Speaker 5: 50:05 Didn’t have any firsthand experience, so I leave that to Karen to unpack it. Practical knowledge is always better. Okay,
Speaker 3: 50:12 That’s cool. Further question come in from Marina Pelu. She’s saying in a federated context, visibility across domains is key, but how you ensure that that happens in an automated way. So it comes back to the computational bit, how you can have visibility across all your domains and is it achievable in an automated way, I suppose? I
Speaker 6: 50:46 Think you can just, maybe I’ll just give the first tab I think you can build
Speaker 3: 50:50 Into,
Speaker 6: 50:51 So you can, very simple example, you’re doing data quality checks on data in a certain domain. Make ’em available to everyone, make ’em available in your catalog in any way that you use to share your data, et cetera. Just make that visible to the rest for many reasons important. I think the most important reason is ultimately because it creates some level of competition as well internally. The more you start seeing others do better than you or your part of the organization, it’s a good way, it’s a good carrot. Let’s say governance is very often the stick, so we need more carrots.
Speaker 3: 51:34 That’s a really valid point. Kinder, Karen, have you used this sort of internal competition as a mechanism to move things along? Kinder’s smiling. So perhaps turn to you.
Speaker 5: 51:47 I’m smiling a lot because Martin and I are partners in crime. We did that before at HelloFresh at Data Quality Hackathon and it was bringing people into that mindset of competition on data quality. So it was about onboarding them, what do we mean by data quality? What is our process there? And it was specifically about bringing visibility on the data quality on their data products within the different domains. And we have them design their dashboards of, okay, what do you want to see? What are the metrics that you want to see? So there’s a lot of things that you can do and definitely I encourage to bring that visibility because that is what actually, it’s not about shaming people, but it is about that transparency and having that fighting competition. And I’m a big fan of that.
Speaker 3: 52:41 No, I’ve used that in a number of different clients. Well, it’s a bit of carrot and a bit of stick, but it goes a long way to get people to get stuff done. Karen, have you used that technique?
Speaker 4: 52:56 Yeah, a bit. But I am wondering, maybe I misunderstood the question, but wasn’t the question, how do you make sure that everything is discoverable or all the data products or the data or what was the question?
Speaker 3: 53:15 Yeah, I think in its broadest sense it was about visibility and ensuring you make it visible to everybody. But can you do that in an way
Speaker 4: 53:27 Yeah. Visible that you have a data product or so you don’t have dark data happening or that you share data? Yeah, it depends on what the question is.
Speaker 3: 53:37 Yeah, yeah, absolutely. Yeah.
Speaker 4: 53:41 But I guess it’s what you also talked about is how do you incentivize people to do the right things. I think that’s something that, I mean, whether we are talking about making sure that you make your data discoverable or that you have the right data quality and you make that visible or that’s really, and I think that’s kind of a skillset that we as a governance, working with governance really need to think about. And here, I think what I have used a lot when it comes to that is systems thinking. And that’s something that I learned recently that I think it’s something that’s super important. How do we make sure that things are happening the way they should in this whole system? Not like this single data product, but how do we make sure that people actually share the data? How do we make sure that they care about data quality? And I think that’s lots of different things that you can do there, but it’s one of those new things that you need to learn, I guess.
Speaker 3: 54:49 Absolutely. And if that feels like a whole new webinar in itself to actually talk about how you actually work out in a very complex environment, everybody’s doing the right thing and systems are doing the right thing. So I think that’s a really good shout out in terms of that. But maybe that’s one for Paul for another day. I’m mindful we’re a few minutes away. I’m guessing there’s no more questions come in. I think we’ve exhausted all of those on the chat. So many thanks to people. Yeah, I’ve just had confirmation that’s everything at the moment, so please, if there is any burning questions in the 11th hour, then let us know. Now. I’m really keen that obviously we bring this to a close. I think it’s been a really lively discussion and we’ve unpacked an awful lot as we’ve gone along. So a massive thanks to Kinder Martin and Karen for taking the time out and unpacking their insights on data governance in data mesh and back yourself, Paul.
Speaker 2: 56:01 Yeah, thank you so much. That was a really engaging conversation. Lots of great questions. So thank you to our audience for asking those questions. Thanks for all of our panelists and thank you Andrew, very much for moderating. I would say yeah, I really love this model having this panel, so if people have other ideas for what they want to hear, let me know. Reach out in Slack or any other way. You can on LinkedIn. You can find me there, Paul au last name au. Or you can reach out to any of our panelists and let them know what you want to hear about. Yeah, we’ll try and line up some more panels. Yeah, I think the next one that we want to tackle is data products, so stay tuned for that. Yeah, thank you everybody so much. Really appreciate your time. Really great response and again, great engagement from our audience. Yeah, have a nice day and we’ll see you on the next one. Alright, thanks.
Speaker 7: 56:58 Bye bye. Bye. Cheers.
Data Mesh Learning Community Resources
- Engage with us on Slack
- Organize a local meetup
- Attend an upcoming event
- Join an end-user roundtable
- Help us showcase data mesh end-user journeys
- Sign up for our newsletter
- Become a community sponsor