Good morning, evening, or afternoon. My name is Malcolm Hawker. I’m your host for the CDO Matters podcast. I’m thrilled that you could join us today on any of our podcast channels on YouTube, on LinkedIn, anywhere where you’re consuming your data related content. I’m thrilled today to be joined by my fellow Canadian, my fellow Edmontonian, my fellow golden bear. We’ll we’ll talk about that a little bit more.

Bethany Lyons.

Bethany is the chief product officer of KAWA.ai.

We’ll learn a little bit more about what KAWA does as well in that conversation. And if you couldn’t already tell, Bethany and I have a lot of things in common. We’ll we’ll talk about that a little bit more. But, Bethany, thank you. So thrilled to have you here today.

Thank you so much. I’m really excited to be on the podcast.

Awesome. Awesome. So, we came to figure out so so I came to learn of Bethany through a podcast that she had done with my friends, Tim and Juan at data dot world and their catalogs and cocktails podcast.

And I was watching Bethany, and I was like, this is just such a breath of fresh air, Thought processes, speakers, thinkers, leaders like Bethany, I think, are what we need more of in the data space.

And and after seeing her on that podcast, I immediately reached out to Juan to get some contact information, then I reached out to Bethany. And it comes we we came to figure out that we’re both from the same hometown, Edmonton, Alberta, Canada.

Although, technically, you’re from Saint Albert.

I I mean No.

I mean, it’s like a fifteen minute drive away. Yeah. But you can call us neighbors.

Yeah. Edmonton is, like, overrun Saint Albert. Right? I mean, it’s it’s it’s a it’s a suburb of Edmonton. It’s okay to say you’re from Edmonton.

So the same the same city.

The same undergrad, University of Alberta. And we were also both involved in the sorority fraternity system, which is actually kind of rare because in Canada, that’s not a big thing. Like, that’s that’s a big year in the US. Bethany’s over in London right now, but it’s not really much of a thing in Canada. So that so so that was really, really cool.

But well, before we dive into it, you’re in London, and you’re from Edmonton.

How did you get to London from Edmonton? I’d love to hear the story.

Okay. Well, I’ll tell I’ll tell you the honest, unfiltered story.

Yeah.

I fell in love.

Oh, okay. Well, that works. That’s that’s that’s as good a reason to know you.

So I met him on a beach in New Zealand, when I was nineteen, and we’ve been together ever since.

So what that’s a love story. That’s a night that’s a nice story.

And and you found your way all the way to London. So one of re one of the reasons that of many that that I wanted to talk to Bethany today is we’ll we’ll just kinda really dive right right into it, and we can save everybody the goring details of of talking about our, our our our salad days, at at the University of Alberta, doing our our undergrads.

One of the things that really struck me in your conversation with with Juan and Tim was you had shared something that you were calling self-service data governance. Yes.

And my first reaction was, I I wouldn’t say scorn, but it was certainly skepticism. And then I heard you talking. I was like, okay. That sounds cool. Tell tell me what you think that means.

Yeah. Sure. So I’ll also say that was my podcast with Juan was a few weeks ago. And one of the things I tend to do is throw, you know, crazy ideas out there to to try to disprove them. So I’ve been on the journey of being like, is this actually a terrible idea? So so I I might speak with it with less conviction this time than I had on podcast because I’ve been faced with the harsh reality of financial services institutions.

So that that was part of the point of of of putting it out there. But, essentially, the the core idea is the vast majority of companies are run off of Excel today. Like, that’s where all data lands. And if the data is in Excel, there is absolutely no security in place. You are trusting your employees to be stewards of that data and to not email it outside of the organization.

And there’s something special about Excel where people know that if they share an Excel file with somebody who isn’t authorized to see it, they will they will be fired, and so they don’t do it. And so I I propose that Excel is some of the highest security data out there because there’s such a strong sense of accountability of the the owner of the Excel file to not let it get into the wrong hands.

So slightly slightly controversial.

And and so, yeah, in that sense, like, people are already governing their data today through ownership and accountability being devolved to the end user of the data. Whereas there’s this funny thing where as soon as the data is in an IT managed system, like, something like Tableau or Power BI.

Suddenly, the onus is on IT, the responsibility and the accountability for the data.

And I’ve never really understood why. Like, why is it that the software that you use to consume the data determines who’s accountable for the data. That just doesn’t make sense to me. And so that’s why I kinda proposed this idea of self-service governance. Like, if if people know, not to share Excel data, they should also know not to share, you know, data in a Tableau workbook that’s managed on their laptop.

And that that was kinda that was kinda the core idea is just distribute accountability more broadly across, like, all of the applications that you might have data in.

But yep. Yeah. So that that’s it in a nutshell.

So, obviously, this this is a a seed of an idea as it were. Yes. We’re we’re we’re seeing if if the the seed is gonna grow and if it’s gonna grow into a a nice tree of of self-service data governance or or or pick your pick your metaphor.

And and I love that, and I don’t think that all ideas need to be completely thought out or well formed before they’re shared. As a matter of fact, one one would one would argue the essence of innovation is is collaboration and iteration. And getting into the idea of kind of, like, what you do for a living, build products, is is that is that kinda how you approach product management? And is is that is that kind of the the iteration and the ideation?

Is that do do you feel like you come at that honestly because of your background on the product side of the house?

Yeah. That’s actually a really good question. I think so. Like, my approach to product management is sell the product before you build it.

If no one’s buying your idea, you should probably not build it. Not don’t actually sell it. Don’t take money, but, like, throw the idea out there and be like, are people receptive to it? And if, you know, if they’re all like, I need that yesterday, then it’s like, okay.

Yeah. We should go build that. But if everyone’s like, no. I don’t want that. I don’t need that.

That’s that’s crazy. Then then don’t build it. So it’s it’s, throwing ideas out into the world to get feedback quickly is a very good way of avoiding building regrettable products.

So, yes, I would say I definitely come by that, honestly.

So I think that’s I think that’s interesting. Right? I think there’s something in in the world of data governance, maybe less so kind of writ large data management, but I think that there’s something in the world of data governance that ideas need to be completely and totally fully formed and well thought out. And and not that I’m arguing against diligence of thought and not that I’m arguing against, you know, crossing all the t’s and donning all the i’s.

But I I think in the world of data governance at the very least, that there is this kind of this binary. It is or it isn’t. It’s centralized or it’s decentralized. It’s governed or it’s not governed.

And and and what you’re saying is is that there is a very much a pragmatic vein of governance that’s happening naturally in organizations.

And in the course of doing their jobs, people are generally doing the right thing, or at least there’s an intention to do the right thing. Do you agree with with what I just said? Yes?

Yes. I I I do agree with that. Yeah. Definitely.

Well okay. Well, so this is what when I was thinking of kind of noodling about what you had thought, when I kind of advanced the tape as it were sorry. I’m showing my age tape. When when when I kind of fast forwarded in time, what I was thinking about was, okay. Wait a minute. If people really, really are focused on doing their job, and if they’re well trained, and if they have the right tools, and if they have the right incentives, then in theory in theory, the use of the data to do the job would be appropriate.

Mhmm. What do you think about that that that thought process? I mean, I guess I guess what I’m saying here is that if if you’re doing your job, then your use of the data is necessary and appropriate.

But what’s your response to that?

So can I answer a different question?

The one that I’m about to That means I need to come up with one.

Or or you wanna No. No. No. Answer the previous one. Oh, okay.

So so so my thinking has evolved on self-service governance because as I’ve gone out and spoken to quite a few financial institutions, what’s becoming a like, very clear is that they don’t want their users to be running their business on Excel.

And they’re they’re pulling back this ability to download data, and they’re saying, like, you can only access data that’s hosted in, like, an on premise server application, and you can never get the data out of it. And that’s the that’s the dominant policy that we’re that I think is is emerging, is like the the withdrawal of, actual physical hosting of data on laptops.

And so this this creates another really big problem, which is the incompatibility of centralized security and decentralized data modeling.

I actually think this is an unsolved technology problem.

So what happens today is IT goes and puts, like, a role level security policy on the the data source and then it throws it over to the business, and the business can do whatever analysis they want on that one table. And there’s always a guarantee that they’re only gonna see the data that they’re authorized to see.

The problem is that the format in which IT presents data to the business is almost never fit for purpose.

I, as an analyst, need to take that IT generated dataset and put it through, like, a hundred transformations and generate a whole bunch of intermediate tables until it’s in the form that I need for my analysis.

What happens is that your security policy breaks. The minute you copy the data out of its original source table into a new table. And at that point, the so so as soon as the ownership of data model is transitioned from IT to the business, the ownership of the security model also becomes transitioned into the hands of the business. So there’s, like, this incompatibility of distributed ownership between security and data modeling. And that’s, like I think that’s actually the problem we need to solve is, like, distributed ownership where one person can own the policy and the other person owns the format of the data.

So, I’m thinking.

Yeah. That’s alright.

Which is probably not a good thing to be doing as a podcast host. Just like, wait a minute. I need silence while I think, which is not always true. But okay. So that’s interesting. Metaphorically, what you’re saying is that users want a meal, for example. They they’re they’re hungry, and they want a meal.

And what you’re saying is is that governance is is happening via access and permissions at an ingredient level.

Yes. Exactly.

So so a really good way to exactly.

So so once something goes from an ingredient to an actually cooked meal, the the the the access and permissions have completely changed. The maybe even the complexion of the data has changed. Maybe you’ve transformed it. You’ve normalized it somehow. Who who knows? Right? It’s it’s not what it used to be.

It maybe it started as a as a customer record or a address field or something, but now it’s represented slightly different.

The governance says nest is probably still the same.

So so so describe kind of a real world impact of that.

A real I can give you yeah.

Let me give you a real I have access to a table, and I can get the date fixed table.

To a reservations table. I’m a hotel manager. Yep. I have access to a reservations table, and it has, like, the created date of the booking, the canceled date, the check-in date, the checkout date, all as columns, and I can see all the reservations from my hotel.

What I wanna do now is calculate the occupancy rate.

You can’t calculate occupancy on a reservations table. You need a room nights table. Right. So what I need to do is transform the data so that, you know, if a reservation was in house for fourteen days, I have fourteen records for that reservation, one for every single night.

And then I can sum up and say that, you know, there was, like that room was occupied on each of these nights. So you need to actually transform I need to take the data out of the reservations table and put it into a room nights table. The minute I do that, I break the security model. It’s it’s no longer, like, use like, IT secure a row and then I change the meaning of a row.

Yep. And so it’s it doesn’t follow through to my transformed dataset.

And so, like, yes, it’s still true that I only have data for my hotel, but it’s unsecured data for my hotel. And so if I share that by mistake, another hotel can see it. That’s one problem. The other problem is you can’t have one user create calculations on behalf of other users.

So maybe you have, like, one hotel that’s figured out how to calculate occupancy in the best way.

They then wanna, like, push back that calculation centrally and say, this is how we do it. Like, other hotels go and apply the same calculation to your data. And so I wanna be able to share that transformation and that calculation without necessarily having access to, like, all of the hotels in the chain.

But but I I can’t do that. How do how do I share how do I share that back? Because it’s I I’ve disconnected my data from the security model and the centralized model and just made a copy.

Right.

So it’s like, how do you maintain the lineage so that you can push stuff back upstream?

So so it’s it’s really yeah. It’s about allowing like, ideally, what would happen is the security policy, like, automatically is inherited into the new transformed data source.

And then I can I can share that transformation with other people, and apply it to all of the records in the underlying data source, not just the ones that I had access to at the time of generating the transformation?

So in theory in theory, knock knock, there would be governance policies at every step of that workflow.

Right? In in theory, there’d be a there there’d be governance policies that would create the rules of the road at at every transformation, at every use of that data into any sort of kind of analytics output or an operational output for that matter matter.

Mhmm.

But but you were at Tableau for a while. Right?

Yeah.

Did so did you see that actually happening?

You so what what I’m hearing you say is that what what you saw was governance being applied at a field level or an attribute level, you know, at the intersection of a row and column somewhere in a table. But after that, it really wasn’t getting applied any anywhere that you were seeing day in and day out.

Is that is that accurate?

Yeah. So at Tableau, like, this was a huge it was a huge problem, and it was one that, like, people didn’t really look into the details of it enough.

And so, like, Tableau could tick all the boxes on an RFP, but it doesn’t actually work, on in a scale deployment because what happens is IT secures a table, and then the first thing a business user does is take Tableau prep and suck the data out to create new data and that data that new data is now ungoverned.

It’s like it’s in the hands of the business user and they can share it with whoever they want.

So that is self-service governance to an extent. You’re relying on them to not share it with Right. People who should be unauthorized.

But but it but it means yeah. It’s it’s it’s a twofold problem. One is, like, as soon as you make a copy of the data, it’s no longer secure. But, also, as soon as you make a copy of the data, you can apply your business logic to another another security policy.

So it’s like in the hotel example, hotel a can’t do a bunch of stuff and then share it with hotel b, and then hotel b sees the same business logic applied to their hotel’s data.

Well, this is something what what we’re talking about here is something that I see day in and day out just in the kind of the governance and data and then let’s community writ large, which is the idea that kind of governance ends at the at at the data warehouse.

Right. Right? Exactly. Yes.

Right.

Like, YT’s delivered the data and governance ends.

Yeah. That’s right. And and box checked, governance delivered, it it ends at the data warehouse, but then once the horse is out of the barn or the data is out of the warehouse, well, you know, it’s anything goes is is what I’m what I’m kind of hearing. I’m not sure that’s necessarily the case because, again, in theory, the governance policy should trickle down to every subsequent use of that field and every table or or I should say in in in in every group by, every sort by, every every analytical output, there should be rules that apply to all of that. But in a world of self-service analytics, I’m not entirely sure how that is doable.

I mean, if it right? If if I wanna if I wanna do even my own transformations or if I wanna do my own groupings, my own sorts, my own my own my own calculations, my own whatever, and I wanna do it on the fly, and I don’t wanna have to be subjected to to onerous rules at the time, then that doesn’t seem like it’s very practical. But I still hear every day vendors out there talking about governance, and they’re and they tend to be on more in the data warehouse side of the house and the BI and analytics side of the house, which is, okay. Well well, governance is access and permissions against a raw set of raw data.

It’s not really making sure that people are using it the right way. It’s not really about the operational use of the data even outside an analytical use case. So that’s I I I think I think you’re onto something. I I assume that’s what your company is is kind of trying to solve for. Correct? That’s what This is is doing?

That’s, yeah, that’s exactly Okay. What we’re trying to solve. So KAWA is the meeting of minds of, like, I’m definitely I view the world through the lens of the analyst. Like, how do I help a business user solve a real problem?

And I’m I’m the person that would, like, egregiously suck the whole data warehouse onto my laptop to then do whatever I needed with the data.

Oh, you’re that person. Oh, okay.

I’m that person. Yeah. Yeah. I’m I’m that person. So I, like, deeply empathize with, like, if you’re putting all these rules and restrictions in place where I can’t do a pivot because it’s gonna break the security model, like, I will break your security policy to get my job done.

So so but then on the other hand, the CEO of KAWA, Hussam Faas, was the global head of IT at BNP Paribas for fifteen years. And so he views the world through the lens of, like, how do you ensure, like, granular access rights management, for, like, a bank.

And so together, we’re figuring out how do we how do we enable, security and self-service to play together, which, like, there’s a million vendors that claim that they do it, but it’s actually an unsolved problem.

Well, right. There are a million vendors that say they do governance that have no integrations into downstream consuming system.

Right. You don’t you can’t do governance without solving the consumer problem.

Whether the consumer is an analytical have to solve them together.

Yeah. Whether the consumer is an analytical platform or actually a human in a CRM system or an ERP system or any other system for that matter. So this this this is a huge knock, I I would say, kind of writ large on on things that market themselves as governance solutions because when you peel the onion, what I see generally is access and permissions.

Right? And you can you can document a whole bunch of other policies beyond access and permissions, but in terms of enforcement, I guess we should draw a distinction between policy definition.

And I can write policy I can write policies on a a piece of paper. Right? It’s fine. I I could I could I could carve it into a a block of stone. But there’s policy definition, and then there’s policy enforcement.

Mhmm.

And what I’m hearing you say is that in your world, in what you see in the world of analytics at the very least, policy enforcement largely ends at the data warehouse, and it ends at access and permissions. And there’s maybe some other other policies that are being forced within the warehouse itself. But once the data’s out of the warehouse, forget about it. The the policies aren’t being enforced. So you can write whatever you want, or you can you can you can have whatever you want in Calibra or Alation or anything else defined from a policy perspective.

If there aren’t systems put into place to to to to manage the enforcement of it in a in more of an automated way, it doesn’t really matter. You’re right back to self-service.

Yeah.

You’re you’re you’re you’re not self you know, you know, self defined, you know, where where governance is is happening because people are just trying to get their jobs done.

Right. Exactly. Okay.

Well so I could understand how somebody in the bank would say, okay. Well, that you know, this this I this idea of self defined or self self governance of of data doesn’t make any sense if I were if you work at a bank.

Done.

I get this, but this problem is even resident at a bank.

It it is resident in a bank. It’s like banks practically invented this problem.

Okay. Well, so then alright.

So so I’m I’m do I’m doing more thinking here, but, like, what the more and more I thought about this, I I threw a couple things out on LinkedIn, and and I do the exact same thing, by the way. I’m throwing spag spaghetti at the fridge, right, to see to see if it’s fixed. I think sometimes people think I’m just doing it for clicks. I’m and I’m actually actually doing it to riff. I’m doing it to iterate.

Refine your opinion.

Yeah. Yeah. Yeah. And one of them was what I started kinda calling governance by design.

And and and I don’t know if if anybody else has said this, and maybe maybe many, many have. I don’t know. Maybe it’s even patented. I I I don’t know. And forgive me if I’m violating your your your copyrights.

But but where governance maybe instead of thinking governance as this after the fact, you know, thing that that to your point, you what you said earlier, you know, why is this all of a sudden a thing because IT people have put a system around it?

Mhmm. Right? Like, why is this now IT’s problem, just because we have some software that’s managing the data.

Maybe governance, not all of it, but maybe a portion of it or something related to it should be part of a a workflow and a process design, part of the the the design of how I onboard a customer or in your case, capture reservations in real time in the system that does reservations management for the hotel.

Mhmm.

It’s like yeah. To me, it’s you have to solve the two problems in the same system. You can’t buy a governance system and then buy a self-service analytics system and then try to integrate them.

It’s it’s like that creates an immediate barrier between the two whereby governance ends with the government’s application and self-service then is totally ungoverned and unrestricted.

So so, like, the way, like, we we’re thinking about how we would implement this is you know you know, like, what I I kinda think of it a bit like what DBT did with the ref function.

They enabled by create by adding this one ref function to So sorry. To for our listeners that don’t know what d b t is Oh, sorry.

Okay.

D b so DBT is a data transformation platform.

It’s given rise to a whole new field called analytics engineering, which is the the intersection between data engineering and analysis. K. Thank you. The key innovation of DBT is that they added one function to SQL. They added the this ref function, which means, dbt is automatically able to construct the dependency graph of all of your tables.

And then they can, like, auto sync the refreshes when one kicks off, and it finds all the dependent ones and then kicks off the refresh of those. At least this is my understanding of, like, what’s the key thing DBT did.

And I’m kinda thinking we could do something similar with security policies and self-service analytics because we have the entire query tree of, like, what did what did the user do to the data?

And we have the original policy, and so we can just trace that policy and transform it through the transformations that they apply. Like, if you do if you do a pivot, your row level security column becomes a column level security policy as an example. Mhmm. That’s the kind of thing that we we’d have to basically do all the algebra to figure out what transformation how do how do data transformations impact your security policies, because your your security policies need to essentially go through the transformation as well. That’s why it’s a hard problem. That’s why nobody solved it. Is it’s it’s a really it’s a really gnarly technical challenge.

I mean, yeah, it’s it’s a very unsexy problem.

I don’t know.

You never know. This this this I I think we may be onto something here.

And and we could be creating new buzzwords, like, on the fly. What I just heard you say, if I were to paraphrase using kind of current terms that are pretty hot.

Yes. Thank you. Please paraphrase me.

I From a hype from a hype perspective, not from a sexy perspective.

What I what I’m hearing you say is that so there’s this concept out there today about data observability.

Yeah. Where where in theory, if you’re capturing enough metadata and if you’re use and and if you are using some of the right technologies, including AI, knowledge graphs, Right?

Others, you you could derive, infer, or deduce situations where your pipeline is breaking Mhmm.

And automatically fix the break. Right?

Mhmm.

We’re we’re we’re and and it’s typically in a pipeline process.

It doesn’t have to be, but this is generally they’re talking about data data observability.

What I’m hearing you say, I’m paraphrasing, that conceptually, you could do the same thing through the lens of governance.

And if you have the baseline governance policies defined and established, you could run some processes that after the fact that say, okay, that observe the use of the data, that observe these pivots that you were talking about, that observed how how how data is being transformed or or used or digested in in analytical systems to say where it was aligned or misaligned with those data policies.

Observing observing government.

That’s exactly right. It’s it’s basically automating the transformation of your security policy Just follow the transformation of your data.

Okay. Yes. So it’s basically kind of automated governance alerting.

Automated governance. It’s not I don’t know if it’s alerting because there’s no you don’t need to notify anyone.

You just need to If if you’re out of if you’re out of compliance with your policies.

Right?

If if in that pivot, something you did you did that was out of out of policy, I think he’d want somebody would want it to be seen where saying.

I’m not yeah. That’s interesting. It depends what you wanna govern. Like, I’m not suggesting that we govern SQL operations that someone can do. I’m suggesting, like, we give we give full free rein to a business user to write whatever SQL they want.

We just need to make sure Yeah.

That, like, as they write new SQL, they’re not they’re not granted access to data they shouldn’t be allowed to see.

Well yeah. Okay. So arguably, I I don’t know how you could break a, well, you could break a governance policy on a read if you didn’t have access to read the data.

Well, we’re falling down a rabbit hole. There may be something here, though. And and and what what what this to me, what this kind of sounds like, frankly, is is data fabric e and and data mesh e as well.

Mhmm. Every time that I’ve and it hasn’t been many times. It’s maybe been two or three times when I’ve actually had the opportunity to have to have an interaction like this with with Zamaq who who wrote the book on the data mesh. Are you are you a mesh? Are you a mesh acolyte?

I’m so A mesh there.

Mesh in so far as when I I’ve given, like, demos of our products to a number of large enterprises, they’ve been like, this is the missing piece in my data mesh architecture. And I’m like, right. I need to go learn about this data mesh architecture because it sounds like we’re building something that’s useful for it.

Well, that makes sense. Right? Because if you ask me in in the limited conversations that I’ve had with Zimok, where I’ve always focused, where I keep coming back to is governance.

Yeah. And if if you live in a siloed world of a domain, what whatever that domain is. Right? Like, in the mesh world, it’s supposed to be business domain. Right? But it could be a data domain, but let’s say business domain. And and your business domain is marketing.

Mhmm. Right? And and that’s where you live, and and and you produce analytics. And by the way, this message is is an analytical architecture. It is not an operational architecture.

Mhmm. Another one of my gripes with the data mesh. But even if you assume analytics and you the the the only consumer of the data is an analytical platform, which is just not practical and kinda silly because data is being consumed outside of analytics. But even if you assume that and even if you assume that you live in this domain well, there’s a lot of data that is inherently cross domain.

And our business processes and our workflows exist cross domain.

Domains. Right? Yeah.

Marketing is about generating a a qualified lead that will become an opportunity, that will become a sale, that will become recognized revenue in some sort of downstream ERP system, right, where I could write a report today for a marketing use case. And in two weeks from now, when it becomes for an opportunity, then this there’s a sales domain involved. And then two weeks after that, after a contract gets signed, then there’s a finance domain.

So that so the information let’s take a customer, for example. The that customer or lead customer, whatever, is traversing multiple domains and and multiple use cases. And how are you going to allow how are you going to enforce governance policies at a cross domain level if you’re basically just saying anybody can do anything they wanted at domain level?

Because when the CEO Can you repeat the last question?

Well, because if you’re allowing anybody to do anything they want, right, it can complete freedom. From a governance perspective, from an analytical perspective. You own peep people you loosely and I think dangerously, yes. I said that.

Dangerously use the word owner of data within a domain level. Like, I own this is my customer data. This is my customer data. Customer data.

I think you can do that if you live and breathe in a business domain.

Right?

And if you never have to share data person to me, you own the data that you create in Salesforce.

That’s how I think of an owner.

Well, so so this is this is the this is why the word owner is so is so problematic.

Because everybody thinks it means something different.

Well, right. Because because it means multiple things. Right? And owner and owner for through my lens, an an owner could be somebody who defines a governance policy.

Do I own that data? If I define it if I define the access rights for data, do I own it? Well Mhmm. Maybe.

An owner could be the person who created it, just like what what you said. An owner could be a person who enforces the governance policy, which could be IT in the exam in the example that you gave. So you’ve got like, that’s three different parties. Or I’m the CEO.

Right? I’m I’m I’m the chief revenue officer, and I’m accountable for or or I’m the chief customer officer, and I’m I’m I’m accountable for customer success. Do do I have a say in in that in that customer data? Of course, I do.

Do I own it? Mhmm. Like, the the the idea of ownership to me is is completely preposterous.

And we throw we throw around this word.

And and then there’s there’s there’s another definition of owner, which is I have the physical data on a file on my laptop. That makes me an owner.

I’ve got I’ve got data sitting in my iPhone. I own that data, I think. Yeah. But but, anyway, it’s a little bit of a rant on the on on the ad idea of ownership. But getting back to your point about mesh.

I’m not sure how we got here, but I’m glad we’re here.

Mesh, my my gripe with the mesh is cross domain cross functional. Right? And and and how do and how do you have and enforce and manage governance policies that are crossing business domains?

And it seems like that’s what you guys are working on, so that’s pretty cool.

But I think I answered Yeah.

It is what we’re working on. So we would have like, in our system, you would have the marketing user creates a campaign’s data source, and they apply a role level security policy that says which campaigns can which users see, for example.

And then you have, like, a sales ops person that has your opportunities data source, and they apply a security policy of which salespeople can see which opportunities.

And then, like, an analyst can go and combine both, and they’ll they’ll only see whatever they’re entitled to see across both, poll data sources.

So we enable you to combine data that has different access rights created by different groups of people.

I’m starting to see now why people say your product is very meshy. Yes.

Yes.

I I mean, we we It’s like we kind of just approached it from, like, this is a problem we wanna solve.

And we didn’t really think too much about, like, how do we align to, like, the big tech trends.

Well, it’s it’s it’s it’s certainly high in the hype cycle. And and it and when you use data mesh in your marketing materials these days, this is like, yeah, gold because there are there are people who are, like, becoming fervent supporters of of the data mesh, right rightfully or or or or wrongly.

Anyway, the the ownership concept to me is a great segue, because we’ve only got a few minutes left, and I do wanna talk to you about product management. But the ownership concept that that that that that conversation is a great segue because I think when I when I see data people try to do product management, they use words like owner.

Right? And and and they throw around well, maybe they have a racing matrix, and I’m not opposed to racing matrix. They’re fine.

But they will use this idea of data owner. What data people will also do is that they’ll say, oh, okay. Well, as a data owner, I get to define data products.

So then let me go figure out, like, what all my raw materials are. Let me go figure out what all the fields I have, what all the tables I have, what all the rows I have, all the APIs, all of the the outputs, all the dashboards, all the everything. And and they will inevitably embark on these, like, these kind of never ending journeys to understand what are all the raw materials that I have in my in my sandbox, right, in my tool chest. And then then I will find ways to combine them together to make my data products.

And I see these things, and it’s like, you guys don’t know anything about product management.

Right? Because there isn’t there isn’t a product management how would you as a product manager?

Let’s say, Bethany, you you you are now the product owner or maybe even the chief product officer of of a company, and you’re gonna be very data centric, and you could build maybe some software, you could do anything you want, how would you go about defining what is your product?

I think the the first question data people need to ask is do they need to build a product or not?

Because there there’s this, like, there’s a lot of hype cycle around, like, data as a product, and, like, part of it’s very expensive to build a product. It’s, like, much more expensive than it is to just, like, build a bunch of tables.

And so there needs to be enough value and enough leverage. Like, to me, a product is I’m trying to solve a a problem at scale for a very large set of users. That’s when I need a product.

Whereas if all I need to do is, like, automate something for one person, that’s not a product.

So so the first question is, like, is the thing you’re building, like, high leverage from a user usage value perspective? And if it’s not, then it’s not a product. It’s just a table being refreshed or, you know, whatever data asset it is.

And that’s why, like, the the biggest the biggest, yeah, question is always about value. Like, that’s what the job of product management is. It’s like trying to ascertain value, and you’re not gonna find value in schemas and fields and assets. You’re gonna find value by speaking to customers.

And this actually comes to one of the questions. I’m gonna preempt one of your questions, which is why did CDOs have such a short tenure? Can I can I pre can I answer that question now?

Should we wait till the end?

No. No. Well, go for it.

It it’s related to this. So I think, chi like, chief data officers to me are, like, kind of an analogous to, like, a CTO. They’re like it’s a technical role.

CTOs work alongside a CPO.

Right? You have, like, the CPO to figure out why and the CTO to figure out how and the combination of them to figure out what.

And CDOs are just left on their own to do everything, like why, what, and how. And I think that job is impossible for a single person.

You actually need a CDO to have a CVO, a chief value officer, to work alongside them to figure out the why. And that would be a person who’s, like, an expert in the industry that the CDO is operating in.

So, yeah, that’s that’s why I think that role is failing is because there should actually be two.

I love it. I it’s so so what you said, right, I I was kinda started this on a mini rant by by some things that I see as frustrating in in the field of people saying their their product managers working for c you know, chief data officers, who are saying, well, you know, I’m gonna spend months my data and figuring out what I can do in doing this bottoms up process. What I heard you say over and over again is that it’s about value and ROI and two, solving a customer problem.

Right? And if you’re not solving a problem and if you’re not delivering value, then it doesn’t really matter. But that’s that is the genesis. That is the north star.

It it is it is the the essence of product management is is delivering value to your customers.

And if you don’t take that approach and if you don’t take the approach of the why and quantifying the why through the lens of value, well, then you’re gonna have a hard time.

And I couldn’t agree more. It’s interesting. So, you know, saying saying you’re the you’re the you’re the what, you’re the why, and you’re the you’re the why, you’re the what, and the and the how, is an incredibly tall order.

And you would be talking about even outside the the CDL world, you’d be talking about a unicorn that that that rarely existed. Right?

That’s right.

Right? And then and then put it into a brand new role where, you know, for a lot of these organizations, this is this is the first time hiring a CEO.

Right? For for a a a vast number of them, it’s the first time. So you’ve got a new role, a a misunderstood role that is that is incredibly wide where the expectations are incredibly, high, and it seems like it’s a kind of a bit of a recipe for for for disaster. So what you would your recommendation is to separate the value component Mhmm. And measuring a value.

I think that makes sense. Right? I I could see that. Now now whether that’s actually a c level person or not, but but a a hiring a value engineer and these people actually exist, by the way. That’s that’s a real title, a value engineer.

Yeah.

Yeah.

Did you say something?

I was gonna say, I think, like, that is a c level person because there is a higher level objective than achieving value. I would I would almost say, like, if there were a hierarchy, the this chief value officer would be, like, the boss of the chief data officer if we had to order them.

So that is Because why is always a higher order question and how?

So, so not but that’s not a chief revenue officer. Revenue would be the manifest how of the why of value from a sales perspective.

It’s not a chief financial officer who But there’s other ways of of delivering value, like, removing risk is is delivering value, for example.

So value doesn’t have to just mean money. It could mean it could it could be measured in other ways. It it’s like the chief value officer could also be somebody who, like, massively eliminates risk for the organization just as a as a yeah. So they’re they’re just about, yeah, either delivering revenue, removing costs, removing risk.

So I I love the idea.

Through, like, innovation and data.

I I I I love the idea. I think we I think we need to press on this a little bit more, maybe maybe in a completely separate episode. But the idea value and measuring value and managing to value, it is it was when I was a Gartner analyst, it was the number one theme of my conversations with my clients, talking about why their data and analytics initiatives were failing. It’s because they were failing to articulate value and to deliver value and to quantify and measure value.

Completely and totally agree. Let’s segue a little bit. Let’s say I’m a chief data let I’m a chief data officer, and this is just in our last few minutes because because we’re running out of time. I’m a chief data officer, and I’m excited about kind of integrating product management into how I do a data and analytics function.

And I even recognize that instead of having product owners, maybe I need to have product managers and that I’ve I even recognize that this is a completely separate role with different different competencies.

From your perspective, what are the kind of the two the top two or three behaviors and or attributes of a product manager would you be looking for that you would that you would recommend that CDR go find? What do those people look like?

Somebody who can say no to ninety seven percent of the stuff that lands on your plate.

Because because I I I think that’s one of the challenges in data today and why it doesn’t deliver value is because we’re trying to process every request. And it’s like, actually, you only need to process three percent of, like, work that’s gonna deliver the largest amount of value.

And so you need somebody who has the strong Pareto understanding of the world, who can just say, like, we’re just not doing that. We’re gonna focus on three things instead of trying to respond to one thousand ad hoc requests.

So that’s because that’s, I think, the biggest obstacle facing data teams today is they operate like service teams just churning through, I like, requests in a Tickets.

An IT system as opposed to product teams thinking about, like, we’re we’re gonna build something very specific that does, like, a couple of things. And then I would I would also hire an industry expert, and I say that because I was I was in a failing role at my last company where I was the I was a product director for analytics inside of a hotel tech company.

This is why I talk about reservations.

And and I realized, like, I just totally lacked the industry expertise to move the that role forward. Like, they they needed somebody who really understood the business to be like, these are the top three things that we should be doing. I didn’t have that that expertise, and so all I saw everywhere was tech debt and problems.

And so you need you need someone who’s almost, like, ignorant to, like, almost ignorant to tech debt and problems who just sees the opportunity, and, like, they can then lead the charge forward.

I think sometimes understanding too much what goes on under the hood can, like, lead you to be lead you to a level of despair that, like, that blocks progress.

Yeah.

And that was kinda what happened with me in my last role was I all I saw was problems.

So, yeah, that that would be another thing is is hire an industry expert. Like, hire a customer is is honestly what I would do. And this is like this is what this is my approach to product management. It’s not to go find professional product managers who’ve worked in completely orthogonal, like, the adjacent industries, but they’re experts in product management process. Like, those aren’t the PMs I wanna hire. I wanna hire people who, like, live and breathe analytics, and then, like, train them in product, essentially. So that’s but that that’s just my own personal perspective.

So I’ve I’ve been recommending the exact opposite. Like, find Oh, really? People in training on data analytics, but, again, separate conversation.

That may I’ve I’ve I’ve loved it. Let’s go break some stuff. Let’s go keep breaking stuff, and I wanna see more I would love to see more spaghetti on the fridge on LinkedIn or anywhere else because I think I think we need to go break some more things. I think we need more new ideas, more fresh thoughts, more ideation, including out on LinkedIn.

And, hopefully, this this will motivate some of our listeners to, to to do the same because I think we desperately need it in the in the analytics space. Thank you so much for your time. I hope we can do this again sometime soon. And for our listeners, our watchers, our viewers, our subscribers, thank you so much for tuning in to another episode of CDO Matters.

I’m Malcolm Hawker, your host. Thank you again, and we will see you on another episode sometime soon. Thanks all.

Thanks.