Tim Gasper [00:00:01] Hello, everyone. Welcome to Catalog & Cocktails. It's your honest, no- BS, non salesy conversation with enterprise data management. I'm Tim Gasper, chief customer officer, and a data nerd at data. world. And co- host, Juan Sequeda. Hey, Juan.
Juan Sequeda [00:00:15] Hey, everybody. I'm Juan Sequeda, principal scientist at data. world. And as always, it is Wednesday, middle of the week. Today, we're doing it a little bit early because we're here live from Gartner right now.
Tim Gasper [00:00:23] Gartner Data & Analytics Conference.
Juan Sequeda [00:00:25] And super excited to have Cris Hadjez, who is the senior director of data governance at Norwegian Cruise Line. And we're going to be talking about our takeaways today. How are you doing, Cris?
Cris Hadjez [00:00:33] Good, thanks for the invite. It's always fun to listen to it, now I have to be a part of it. So let's see, I'm hoping to be as engaging as all the other guests you've had.
Juan Sequeda [00:00:41] Welcome to the alumni group of Catalog& Cocktails guests. So before we kick it off, it's still early, what are we drinking? What are we toasting for?
Tim Gasper [00:00:51] I've got some tasty coffee. I don't know, what are you drinking?
Cris Hadjez [00:00:54] I'm sticking with water for now.
Juan Sequeda [00:00:55] It's been an in intense couple of days for sure. But hey, let's go toast it. I love that we can be able to do this show live with folks at a conference.
Cris Hadjez [00:01:03] Cheers.
Tim Gasper [00:01:04] Cheers.
Juan Sequeda [00:01:04] Cheers for everybody. All right, there's so much to cover.
Tim Gasper [00:01:09] I know. It's going to be an exciting discussion.
Juan Sequeda [00:01:09] Let's hit the ground running. So what are your honest no- BS takeaways from these last couple of days here at Gartner?
Cris Hadjez [00:01:16] Perhaps the thing that for me was the most intriguing is that there were a number of sessions I had a chance to listen in on. AI was very big, data mesh was very big. But the underlying theme in almost every session I sat in on and perhaps it was what I picked, it was the importance of metadata management. I was very surprised by it. But I think that what really was unique is that I didn't pick the keynote. The keynote is the keynote. And in the keynote there was such an emphasis about the success of all of these initiatives that companies really want to do, and the importance of making sure that metadata management is critical for those success stories that they're trying to do. And it was as though they were saying, " If you haven't gotten that step right, these other ones, you're going to struggle." So it was almost like their no- BS talk. It was like, " These things are great, but you've got to get metadata management right first."
Tim Gasper [00:02:08] Yeah. What was your perception on why metadata management was such a common thread? Why does everything come back to that? Because I agree with you, and I'm curious, what was your interpretation of why that's the thread?
Cris Hadjez [00:02:23] I think that ultimately it is about people still understanding what their data is, increasing that data literacy that you hear about quite a bit. And people are drowning in data, but they don't even know what that data is. When it comes to some of the AI projects, we've had a chance to experiment a little bit, it teaches us a little bit more about our data that we actually should know more about, and we don't do that until you start playing and start doing some of these more advanced, let's call it, architectures or efforts.
Juan Sequeda [00:02:51] Because we've been talking about metadata for so long, but I always argue that it's treated as that second- class citizen. It's not the data architectures and all this stuff, that's a big thing. And is it finally that time has come that people are realizing, " This is what I need to be focusing on"?
Cris Hadjez [00:03:09] I would say yes. And I think that the reason that those topics were so prevalent is they're sexy, it's really cool, let's do this. You start doing it, you're like, " Why are we failing? Why are we failing?" And then you start seeing the themes come out and it's just like, " Oh, we probably should have done that foundational step first, and then we could have gotten further along." So what you're hearing, at least what I'm hearing a lot is, companies that are trying that, they're taking a step back and saying, " Okay, we can see what this power of this new path can be, but we have to get that basic step right." So I'm hearing a lot of people going back to that saying, " Okay, let's get that step right." So for those that haven't done it yet, it's almost like they're getting a chance that if they do that metadata management discipline right, and then they go do that project, they're going to be like, " This was so easy, I don't understand it." But that's because they did that foundation step.
Tim Gasper [00:04:00] I think it's very interesting that, especially now with AI projects, and I know we're doing some interesting stuff too, we should talk about that in a little bit, that it brings a very visible feedback loop around any kind of problems around data quality, around data understanding. I feel like we've had these problems for a long time. You mentioned data literacy and people are like, " Oh, I can't find data, I can't understand the data." But maybe it was a little fuzzier in terms of like, " Oh, what does that mean?" " I couldn't find the dashboard fast enough," whatever it is. But now with AI, literally the AI is hallucinating, and it doesn't understand because you don't have the data quality, you don't have the data understanding, the data documentation. Is this a good thing that now it's bringing more visibility, more of a feedback loop around the problems around our data?
Juan Sequeda [00:04:51] Are people like now they're crashing into the wall and like, " Well, you need to feel some pain"?
Cris Hadjez [00:04:57] Yeah, I do. I believe that that's the case. I think that the scenario where... Look, the reality is it's very challenging for anyone to be able to say that you start with a company and that's the company you retire. So the practical reality is you're going to have turnover, and as different people come and go within the organization, you use different terms, you use different words, you call the content slightly different. So AI is expecting you to have discipline, it's expecting you to have everything is called the same. And then when you're trying to use it, AI is coming back like, " I'm a little confused because there's five different things here that could be a customer. Which one am I listening to?" Which is ironically the same problem the business has.
Juan Sequeda [00:05:38] I like this. AI expects discipline, and I think I would unpack that and I would say for, let's call it consumer type of applications, you can be fuzzier, it's fine. But I think when it comes to the enterprise world, you need to have that accuracy. You need to be able to say, " Oh, this is the answer for these reasons," because regardless of the AI, this is what everybody's expecting, executives all the way up and down, they're saying, " If I'm going to ask a question, I want to have an answer that I can go trust," and so they can go explain that answer. And part of the explanation is that you know there's a person, and that person... and then there's accountability around this. So it's all about that discipline. I really like this, it's an important thing thinking about it when it comes to the enterprise.
Cris Hadjez [00:06:23] Because the way I view it is that the human has the ability to do the interpretation. The AI is going to try to do the interpretation as best as it can, but it is, it's going to run into those challenging, let's call it, walls that it's going to hit. I dare say that if you have the same organization running an experiment, have them go do that AI experiment with a curated data set that you know is good or just let it go in the wild, and do you get the same results or do you have a different opinion of AI now? It's like inaudible.
Juan Sequeda [00:06:54] Is that something that you're seeing people talking about like, " Let's actually do experiments, all right. Let's go sit down and do the foundations, and let's go have fun, do the cool thing"?
Cris Hadjez [00:07:05] And see what happens yeah, just see what comes out.
Juan Sequeda [00:07:06] Are you seeing people doing this?
Cris Hadjez [00:07:08] I'm not seeing them do it, I'm hearing them talk about it. In the evenings when we're not drinking water and some people start saying, " Hey, what is it that you really do? You do data governance?" It's like, " Wow, if we could get some structured data, I think that we could really do some neat things." And you hear that quite a bit. And not structured data in the concept how we think of structured and unstructured data, but just I think a better choice of word is curated data. It's like you know that this is a safe walled garden, you know that data is good. You run your things against that. I almost make it akin to a joke that I always say is when vendors are out there selling their products and you see their products, their products look amazing. And my response is, " You have a curated dataset and a demo script, and that thing is going to look like the best thing you've ever had. Now grab that same product and use your dataset that's got no rules, everything's wrong, and what does that product look like?" And a lot of people tell you it's like, " Yeah, that product's horrible." It's like, " No, the product's fine. The problem is your data, you've got to get that right."
Tim Gasper [00:08:04] I think that's a big insight is that a lot of folks want to do some really innovative stuff around whether it's self- service analytics or around AI. And ultimately if you don't have that curated data, if you don't have that foundation, you're going to run into a lot of problems. So it's probably a good time right now to be in governance or data enablement, whatever term you want to use because now the focus is higher than ever on it. So what's your feeling on that?
Cris Hadjez [00:08:34] Again, I think it really goes to varying degrees of success. If you've had that discipline, that's great. I think it also, honestly, it opens up the door for the constant struggle that you hear about, which is partner with the business, IT, you have to partner with the business, reach across and work together because now it's truly a scenario, it's like, okay, the investments can be there, the efficiencies can be there. But if you can't figure out a way to get all of those groups to work together hand- in- hand and land on, " What do we call this? How do we use it? Where do we get it from?" and have that consensus, and honestly, a lot of these investments are going to fall short, and that hype circle is going to struggle like it usually does. So I think it really truly opens up that door to further reinforce the importance of that partnership.
Juan Sequeda [00:09:22] One of the things that we were doing before we got live here was going through the oldest sessions at Gartner and seeing what they were tagged with the topics. So not a surprise, the number one is AI with 79 things. So there's like 79 sessions that have tagged or whatever with AI.
Tim Gasper [00:09:39] That's a lot of sessions.
Juan Sequeda [00:09:41] Yeah, number two, I was actually surprised that it's the data analytics governance. So it's interesting, so people are like-
Tim Gasper [00:09:48] 67.
Juan Sequeda [00:09:48] 67.
Tim Gasper [00:09:50] Almost as many as AI.
Juan Sequeda [00:09:51] And then after that you have the strategy, which was 55, then sessions about seat for CDAOs, 52. Analytics and BI had 43 sessions. Leadership, 40. Data architecture, 31. So this is, we're seeing this slip of the core of data management architecture that is... we're talking about it, but there's more important things we be talking about. AI, obviously, so that's a hot topic, but governance is the one that's supporting. Interesting enough, data quality, 21, data literacy, 11, MDM, 13. And the one that had the least was real- time and streaming with three
Cris Hadjez [00:10:31] Sessions. Okay.
Juan Sequeda [00:10:33] So listeners, you can interpret that however you want, but...
Cris Hadjez [00:10:38] Yeah, I think it goes back to you've had years of those topics being the high volume topics, at least I believe that that's how the number would've played out. And now it's getting to a point where people are sitting there throwing their hands up and saying, " We can't get this to work. It sounds cool, but how do you get it to work?" And so I think that the focus has shifted to saying, " Okay, let's help people walk through what that would look like."
Tim Gasper [00:11:02] Yeah. It's no longer just a conversation about data lake architecture or something like that, that's just a means and kind of a given now. It's really, "Hey, I want to get to AI," 79 sessions, " and the only way I could do it is with effective governance," 67 sessions, right?
Juan Sequeda [00:11:22] So let's get into it a little bit deeper. What did you see in sessions about what should you be doing, how should you be doing with metadata?
Cris Hadjez [00:11:28] So on the metadata side, it goes back to the question or the comment I made earlier, which is it really is boiling down to their emphasis is that it's on the partnership. You have to be able to get these cross- functional, if you're multi- subsidiary or cross- branded... gets everybody together and let's start talking through. And again, looking at the basics, don't go out to the ocean, don't go after everything, but pick things that make sense. Usually the easiest path forward is... The easy first step is pick a report, pick a dashboard, pick something that everybody knows that's super important, and just define everything on there. And then over time just let that grow out a little bit more. I saw that as a bit of a theme where they're really trying to get more engagement from everybody. And it's also quite a bit about the prototyping in terms of, " Hey, try this experiment with this, play with this. What do you think about it?" And I think that the other thing that was also rather interesting that they were mentioning is that there's been, I guess, conversations where it's like, " Hey, do I do A, or do I do B, or should I do C?" Particularly when it came to the data architecture, it's like people were sitting there saying, it's like, " Look, do I do data fabric? Do I do data mesh? Do I do lake house?" And what they were trying to say is, no, it all works and in a way that you should be able to understand the complementary capabilities of each, and you know how far you carry it, and then you take it to the next step using the other architecture. And I think that that's an interesting path in terms of how people are looking at it because most people would debate and say, " No, I'm doing this," or, " I'm doing that." They don't look and say, " We're going to apply all of it." Because now when you're applying all of it, it brings in the question it's like, "Do I have the architects? Do I have the skills and the engineers to manage it? Do I even have the engagement from the business in order to make that model work?" The bigger picture, yeah.
Juan Sequeda [00:13:23] So one of the things that I've been talking with one of the analysts here, it was on the social side. So I love to chat with Mark Byer, all the time. And we've been talking about how that social aspect is missing. You're talking about partnerships, right?
Cris Hadjez [00:13:42] inaudible.
Juan Sequeda [00:13:43] How do we go find those partnerships, and who's doing all those types of work? And so we started the discussion saying, " Look, what we need is more of a social data engineering." And that was on Monday we were talking about that term. And then people were like, you actually said it too, well, that kind of seems like social engineering, like hacking.
Tim Gasper [00:14:00] Hey, listeners, what do you think of the phrase social data engineering? You like it, you don't like it? Let us know.
Juan Sequeda [00:14:05] Well, what came out was, well, Mark was also sharing with other people, and somebody mentioned the same thing, he said, " But what you're really trying to do is collaboration. So it's really like collaborative data engineering you would do." And that's the one that's stuck with me. I like this. Now, but at the end of the day, you have all this metadata, all this knowledge that's embedded not only in the systems and stuff, but it's also in people's heads. So we need to create catalogers or whatever extractors of metadata from the workers' heads. And I think going back to the point that you said is people are now... they come and go in organizations, they have all their own know- hows, the terminology, their processes, how they do things. That is something that I feel that is a and figure out how to go partner. And I think we need to have those folks who can do that role of that collaborative data engineering to be able to build more bridges across different communities and saying, " Oh, well, by the way, I'm not going to go talk to everybody else, I need to go figure out a focus. And here's this report that we're trying to go understand. Let's figure out that entire thread that goes from top to bottom and there's a bunch of people I need to go talk to to figure it out and understand." So that was the conversations I was having with folks here. I don't know, how does that connect with what you've been seeing?
Cris Hadjez [00:15:22] So two quick points that you really made me think about. So one is, if we go back 10, 15 years, a super popular topic used to be the concept of knowledge management in an organization, not metadata management, it was just basic knowledge management because what you said, it's like how do we do this process? What's the most efficient way to do the process? No technology involved, it's just how do you do this particular process and get it through? And it was just the basic concept of knowledge management, and how do you transfer it, and how do you make sure that when you're getting new folks in because either they get promoted or they go to another organization, the next person has to know how to do it as efficiently, hopefully not dropping that productivity capability. So that topic's actually been around for a while. I don't think that's what you were saying earlier. I'm not sure if everyone's taking it seriously and now people are realizing it's like, " Well, maybe we have to take it seriously." Juan, you made a very interesting comment about, " Let's figure out about this report and this dashboard." Last night, having a conversation over dinner, and someone shared the fact that they had read the post on a thread out there that said this new person went into the organization and they were asked to go ahead and review some of the contents as part of their ramp up. And what they did is they looked at a report that had been produced for the better part of two and a half years, and they realized that the person that had developed the report had done the join incorrectly, and was providing information that was incorrect to the organization for the better part of two and a half years, but everybody was using it as gospel. And the person's post was, " Do I tell them that it's wrong, or do I fix it and not say it?" So now this person is inheriting the issue and it's like how do they navigate that scenario?
Juan Sequeda [00:17:11] This is fascinating because this is what I call the politics of knowledge. What do you do?
Cris Hadjez [00:17:16] What do you do?
Juan Sequeda [00:17:16] Right. You're going to have to go do the pros and... Okay, let's-
Cris Hadjez [00:17:16] You play it out inaudible.
Juan Sequeda [00:17:17] Let's play this out. What do you do?
Cris Hadjez [00:17:25] Me, personally, I would just document and demonstrate why I think that number is wrong, show it because maybe I got it wrong, maybe that join, they already figured out that correct join would've produced the incorrect numbers. But what this person's position was, being again new, but being technically... they already told him that he was technically stronger than the prior person there. And so that was the conversation at dinner last night is what do you do? Do you sit there and say it? And then the debate became, it's like, " Well, what makes you think that that person's right?" So now you're questioning that person's technical capability even though he was told, " You're stronger than the last person that was here.
Juan Sequeda [00:18:00] So when it comes to politics, and I think we had this conversation last week too about something about politics, and it's like wait-
Tim Gasper [00:18:09] Around ownership, right?
Juan Sequeda [00:18:10] When it comes to... This is not something like, " Oh, I'm going to come in and tell you that you guys are all wrong." You're going to make enemies, especially if you're new. So I think part of it is understanding the landscape of your organization, understanding your peers. You need to make friends, you need to start socializing this stuff. So I think in a situation like this, one approach is, let's start socializing with the people right next to you. It's like, " Hey, so I'm looking at this, and what do you think about that?" and see if other people catch it too or not because then they may give you more context that you missed and you're like, " Okay, no, this is why it's correct there." So I think that socializing little by little. But here's the thing, if I zoom out and the honest no- BS thing is, that was wrong, but so what? The business was still functioning, right? Well, they're not on the news because something bad happened, whatever. So what? And I think we need to figure out the so what. And this is the important case for metadata and for knowledge management and so forth to say, " Look, the so what is we left so much money on the table," or this is where we need to go to with the question because otherwise people are like, " It's directionally correct, it's fine."
Cris Hadjez [00:19:25] That's exactly the term, right? And that's where the comfort level, I think, comes in, is that most organizations will sit there and say, " Look, if it's directionally correct, I'm okay with it." Unless I'm a publicly reporting organization, I have to get it down to the penny right, then per most people for their day- to- day decisions it's, " Look, directionally, as long as that's an accurate enough number, we can move forward."
Juan Sequeda [00:19:48] This is what we need to go figure out.
Tim Gasper [00:19:49] And knowing that downstream usage is pretty important here to help inform some of the politics. If you know that that number was being reported in the annual stock report, then you're going to handle it in a different way than if it's a casual dashboard that occasionally the executives like to look at once a quarter, but it's non- critical, they're just looking for directional trends, maybe you just update it and move on. So it's interesting to navigate the politics around all of this. And a lot of that has to do with how are we using it and who's using it.
Juan Sequeda [00:20:24] Continuing on this, what you realize is that if people they find issues, they think it's an issue like, " Data quality is a big issue." Like, " Yeah, that's an issue, but how is that really impacting across the lines of different businesses?" People are like, " It's not a big deal for me. I understand it's a problem, I have bigger problems I need to go deal with." So I think that's, again, that partnership is super important to understand, " Okay, great. I'm done with this, I'm going to put it aside, I'll create a log about it, and then I'll go to the next problem." So I think that's the type of politics we need to figure out inaudible. All right, what else? I know you were talking about the data mesh, data fabric, and that was a very popular presentation.
Cris Hadjez [00:21:10] Yeah, that presentation, I didn't think it was going to be as big of a deal, it was an enormous deal. I think the room held 700 people, room was full and standing room only along the walls. Everybody wanted to listen in on the session. So it was perhaps, other than the keynote, I would have to say, the busiest session I saw, the most amount of people in a session at all, it was that one. So it's interesting, I think, that people are trying to gravitate to get a better understanding. I'm not sure how many of them are doing it, or what I think may have very well happened is that when they put the architecture up there, I think people were trying to kind of say, "Hey, I resonate with that because I actually have two parts of that, or six parts of that," because they would sit there and say it's like, " Look, there's seven components in order for this to work." And I think that what was an interesting thing that they tried to make the audience at large feel a little bit more comfortable is of the seven components that you need in the architecture, three are very mature in the partner space, the other four, they're either brand new or they're trying to get a level of maturity. So you have to be somewhat selective in terms of what you think your success is going to look like, and how that plays out. So it was interesting to hear that almost to kind of remove a little bit of the pressure to say, " It's okay, go experiment. And so set your expectations correctly when you're trying to promote or advocate for this in your organization."
Juan Sequeda [00:22:35] So another theme here is a lot of experiment like, " Go off and experiment with it-"
Cris Hadjez [00:22:39] Just try it.
Juan Sequeda [00:22:40] Try it out. I think one discussion, the session that was too is the main thing about data mesh is the data products. So I think that is now the thought of a data product, I think, I would consider is mainstream. And I think, actually, people have gone over like, " Oh, what is it?" But the definitions, they've stopped being pedantic about it, which I think is a good thing, " Let's just get shit done, and provide business value." But I think the notion of the data product and having that separation of... what is the infrastructure you're going to create that? And then you have that delivery mechanism with these data products. And I think that's something that I'm starting to go see more and more now that people are getting that mindset right there. I don't know,
Cris Hadjez [00:23:23] They are. The biggest pushback that I would consistently hear after the session, just chatting with different peers, is they're saying, " Okay, so now I have to divide up my existing data engineering team out to different places or we have to bring in more resources." And then the second- biggest concern that I heard was, " You're never going to ultimately be in a full data mesh strategy. So if you're still keeping your central data warehousing concept, and you're applying the data mesh strategy, you're having parallel teams, so you actually increased the headcount that you needed to support the data." And I think that that's where... I'm not sure exactly how you walk that line, and provide the efficiency, and provide the data products, and allow the businesses to get what they need without understanding that there's an incurred cost to that style of agility.
Juan Sequeda [00:24:14] This is interesting because at the end of the day, this is definitely not a technology problem, and it's truly just the people and process, and specifically here it's like, "How do I get all these people? I've got to pay for them." So it's that worked-
Tim Gasper [00:24:30] Or to enlist them somehow if you're borrowing or collaborating with people across their organization.
Cris Hadjez [00:24:36] Well, and so it was interesting because one of the topics that they reference in that is that the struggle that they've already identified is that when you start working across the organization, the technical acumen of the other parts of the organization typically are not as strong as what you hold in IT. Okay, that makes sense. So are we saying that you want to create a federated IT department? Are you saying that, to your point, you're going to enlist and recruit, and now you work for me, but it's in your cost center, but we teach you the practices of how to do? Again, how does that social construct start to work in an organization? And if they're trying to shoot for efficiency, does that efficiency... What is the incurred cost to get it? Do you have the palette for it? Are you willing to pay for that? I think that that is the consistent chat I kept hearing. I think in the next couple of years we'll see how many companies are more willing to go ahead and commit to that.
Juan Sequeda [00:25:35] So people are more talking about it inaudible.
Cris Hadjez [00:25:36] They're talking about it, for sure.
Juan Sequeda [00:25:38] But any people actually having an answer like, " The cost of efficiency was X, for us."
Cris Hadjez [00:25:44] Exactly. " And here's what I got out of it."
Juan Sequeda [00:25:47] And I got Y, and Y was this little bit or big, way bigger, I don't know.
Cris Hadjez [00:25:52] Marginally or exponentially, what was it? So I think that that's what still stands to be determined. I think that they're not entirely certain. In fact, if you even have the opportunity to speak with some of the Gartner analysts, they will sit there and share with you, it's like, " Look, there's really nobody that's 100% full data mesh and it's working perfectly. They're just not there either." So they'll tell you that. So in all fairness, and from a full disclosure perspective, it's like, "Well, people are absolutely," call it experimenting, call it looking at it, call it researching it, put whatever adjective you want, but it's definitely being discussed a lot more than I would've expected, so yeah.
Juan Sequeda [00:26:30] Okay. All right. What other main themes have we seen?
Cris Hadjez [00:26:37] So the only other main theme, and granted only because of my role, it was a bit biased for me, I was trying to a little bit more where people are going with master data management, trying to understand if it's the typical, do you buy it off the shelf or do you build it? What are the challenges? How many of the domains do you go into? Do you go into your traditional domains for master data management? Do you create a new concept of the domain of how it applies to your business world? But the interesting thing, even in the master data management side when I was listening to those chats is, guess what was another theme that was in that chat?
Juan Sequeda [00:27:09] Metadata
Cris Hadjez [00:27:10] Metadata management. So they go back to the same thing, it's as though if we think about that concept of the language, we're all speaking, right? We're speaking in English, people are understanding us in English, but before we could speak, we all had to learn words, we had to learn letters, we had to learn how to put them together. It's not how it usually goes in the world, " Let's go slap data together and just see what happens." And then could you imagine us trying to communicate, we didn't have that foundation. How do you do that? So I think now it's coming back. Everybody's taking a step back saying, " We've invested countless dollars, countless people, countless hours and we're getting marginal resource, including on master data management."
Juan Sequeda [00:27:53] Right.
Cris Hadjez [00:27:54] So I think on the master data management side, what they were finding is that if you can find how to get everybody to agree is this is call it your system of record or this is what we're going to call it, they were seeing better levels of success and that dovetails into some of these other things on the A side, it was something that they did try to explain how it all plays together.
Juan Sequeda [00:28:18] Well, I think part of the so what here to give an answer is we just need to sit down and say, " Okay, here's how much money we've spent on all this infrastructure and all this technology, and look how much we have not been spending on, it's called the knowledge management, on the collaborative data engineering thing is like, " Hey look, we know how much we've put in and we know how much we've gone out of that. We have those numbers. Now let's go shift some things and talk about prototype and experiment. Give us the opportunity to go shopping because we believe, make the hypothesis that if we invest in more on this knowledge management, people side and stuff, we're going to go get an increase.
Cris Hadjez [00:28:57] Get higher return.
Juan Sequeda [00:28:59] And then we can go do that because we have a base that we can compare to. I think this is the, "We should try out these..." We should do this more.
Tim Gasper [00:29:07] Try to do some ROI analysis.
Cris Hadjez [00:29:09] It is. I think what you're seeing though is that you're seeing organizations that they've made the commitment to go on and down these paths. And they're not just ready to throw in the towel. They're saying, " We don't think it's so much that the technology didn't work, we just don't think we were ready for it. We think we missed some steps that we should have probably thought through initially." Most people-
Tim Gasper [00:29:32] Is that the right way to think about it?
Cris Hadjez [00:29:34] I haven't gone through it. I can tell you what we try to do-
Juan Sequeda [00:29:38] But I was wondering, it was sunk cost versus are we falling into a trap of fallacy, or sometimes we do need to be patient and just invest and maybe adjust our approach a little bit, right?
Cris Hadjez [00:29:49] Agreed. I will tell you that it was very long time ago, probably been 15 years ago or so, I was at another conference and a federal agency went up and explained how it was something to the effect of it was their$ 67 million learning lesson that they figured out after... 67 million that was thrown at the problem, is that they had it wrong and they have to start over. And how do you have that conversation? You spent$ 67 million on something. It took you that long to figure out, " We got it wrong, we've got to do it again." That's a very difficult conversation to have. So I think that you have some organizations where they haven't perhaps spent$ 67 million, but they're just like, " Look, we believe that we actually could get this to work if we can go back a little bit and do some things again." And in fact, if you listen to a lot of the case studies when people go to conferences and they want to listen to a customer talk about their scenario in the case studies, what's the one slide everybody's waiting for? The learned lessons. " What did you learn?" " If I could do it again, here's what I would do." That's the whole value in the whole presentation is they'll show you all the really cool stuff, but at the end, the one slide everybody's looking for is like, " What did you learn? What did you learn? So that I don't make that mistake."
Juan Sequeda [00:31:05] Well, I think what's also important, especially when we had this conversation last week too on the podcast, was that we need to also be sharing all the bad stories, which we don't. We go off and we're like, " Oh, everything is just unicorns and rainbows. Here's all this thing, it just works perfectly, blah, blah, blah." I'm actually more interested... We should give all the failure talks like, " Here's what I did, and it didn't work, and it cost us much, and don't do this. Don't do this, please." I think that's more... We need to have an anti- failure conference.
Cris Hadjez [00:31:36] I had a term for that in the past life I used to call this, look, we need to be comfortable having people tune into the horror channel. These are the horrors of all the deployments of what has gone wrong. And let's try to make your story not be another anecdotal story that we add and talk about. Let's just figure out, these are all these people that have already tried this. And in the same breath there's going to be some folks that are like, " This is new, and we're going to try it." And as long as you go in eyes wide open, and it's like, " Look, there's a chance you may not get it right on the first path, that's okay, but you know what you're getting into, right?" As opposed to being sold or provided a vision or dream of how it can work, and then you're slightly disenchanted because it didn't live up to your expectations. So now what do you do?
Juan Sequeda [00:32:22] Going back to the master data management, I was in one of the talks it's, Oh, it's not Dead. What I found interesting about it is I think when people say, " It's dead," it's because they're thinking about just the old school tools. And what was really interesting is that people talk about MDM, and specifically in this talk, there is no speak about, " Oh, a vendor product to go do this." It's like here's a concept of it... and I think everybody will agree like, " No, no, no, we should make sure that our most important data about our master data or customers, we need to manage that." I think nobody is going to disagree with that. And you like the notion of keeping it lean. So it's like, "Oh, if we're going to talk about customer data," which you're going to master that, " well, you don't need to add all these other attributes to that stuff. You don't even to have the sentiment of the customer. No, let's keep the very middle then keep it lean. So that was good. That was one that I like. And then connecting it back to that's just the data product. It's just something else that is, go back to your terms. It's a very highly curated thing that this should be beautiful data inaudible can go use. So at that point I'm like, " Yeah, master data is not dead." You can do it in multiple ways and it's not always you have to go get a vendor, you can figure out your own. So I think that's the balance inaudible.
Cris Hadjez [00:33:35] For sure. I would tell you that, at least, the leaner, you start with your master data effort, you're going to have a higher probability of success. So to your point is, I don't need to bring in 12 attributes, I'm going to focus on five, I'm going to focus on four, whatever that minimal number is. And then even when you're trying to do those match rules in terms of how that works out, it'll actually show you, " Are my matchings right?" so before you start going down the path because invariably at some point you will potentially have either a data ingestion problem or a data interpretation problem that might have that attribute or value that makes no sense. And now it's like, " Well, my MDM project is paused because we just can't get these things to match." And so I do, I think it requires some thought when you're going through it, but if you start small, you can get that thing actually moving forward. Now to your point, don't get so many attributes.
Tim Gasper [00:34:27] Is that a very common messed up, you would say, in general that when people tackle these master data management type initiatives, they're thinking too big, let's boil too much inaudible.
Cris Hadjez [00:34:38] I think that people think of the concept of the golden record, and they really like the concept of the golden record, one place where I can get the information that's right. And I think it goes exactly back to what Juan saying earlier is like, " How many attributes do you need for that record to be golden? Is it sufficient for you just for me to be able to say, 'Let's use four or five attributes, I've got the person'?" But perhaps even more importantly, I can then create, let's call it like a neural network, if you will, of some kind that just gives you the pointers. If you want more detailed information about that person, just go to that system to pull it, but I'm not going to bring it in because then it throws off all these rules that you're trying to match that you can't.
Juan Sequeda [00:35:16] And this goes back to why metadata is so important because at the end you want to have all that across the social collaborative because you want to know for all these different business units, these are the things that they care about the most. And so that's the social part, that's what's in people's minds, we both track that, and then be able to connect it and realize, " Okay, these are that we need for these reasons." And you should be able to say, " Because each of these attributes drive some sort of revenue or whatever, and if I don't have this..." So that's how we should be able to make that case.
Tim Gasper [00:35:48] And if you want more detail, you can go to that other system and you can go there-
Juan Sequeda [00:35:54] Yeah, exactly.
Tim Gasper [00:35:54] As long as you don't need to take all 300 attributes and try to master them.
Cris Hadjez [00:35:56] Yeah. Just get the ones that help you match that person because throughout the life cycle, in most organizations, if you think about it, an organization is going to have some kind of marketing or sales platform, almost everybody has a CRM tech platform. Then you'll have your sales platform, whatever that is. So that in and of itself, by definition, depending on where that life cycle is for your customer, could live in all three platforms potentially, or a version of them, or they can only be in one or two platforms until they kind of continue that journey with that particular company, and now they start to expand, and now you're getting more actions, you're learning more about them. I think that it's going to be a very interesting next two to three years to see how transformative some of the things that people are trying to do really go back to the basics again. It's like it's almost circular. A lot of these things, it's like, " Hey, that was really new, now it's not new, but guess what? It's new again."
Juan Sequeda [00:36:52] What comes around goes around, we need to know our history. And I think what we're discussing here today, and what you've observed in the takeaways of the conference, which has been AI and the metadata and the theme of everything, and everybody's mind is right now with AI being able to think about how to chat with their data and stuff like that, all the conversations I had, it was very clear to them that we need metadata, we need that context. If we want to be able to accomplish that vision of to chat with our data, especially when it's with the structured data, it was clear to them that metadata and context needed. I did not have to have any conversation with anybody trying to convince them that metadata was important for that. So that is a level of, I don't know, maturity or people are now connecting the dots and like, " Okay, I know what that future looks like, I can see that, and I need that foundation."
Tim Gasper [00:37:49] Yeah, I think that's true.
Juan Sequeda [00:37:51] So that makes you feel comfortable.
Tim Gasper [00:37:52] I did have some conversations though over this conference where the impact of AI was a little bit not fully understood by folks. I did have quite a few conversations where people were like, " Oh, I'm really hoping that AI is just going to solve quality, or it's just going to solve stewardship inaudible."
Juan Sequeda [00:38:10] That'd be great. Let me know when it happens.
Tim Gasper [00:38:13] Obviously, I think there's going to be a really positive impact around that. But I was a little worried because... So there are some people who I think fully understand and appreciate that AI is more of a copilot and it's going to help, it's going to be an accelerator. And it's also earlier in our conversation where we talked about how it's also going to be a great lens back onto how is the quality of our data, how curated is the data force us back to that right foundation. But I think there is a little bit of confusion still by folks who are hoping it will be that magic bullet.
Cris Hadjez [00:38:42] But the interesting thing about that, if you think about that is even in ChatGPT, it says, " Please double- check the answers we're giving you." And ChatGPT is arguably perhaps the strongest, most popular AI platform out there right now. So even they say that. And I think, Juan, you used a very specific term a little earlier, and I think that that's really important is the context of what that data actually means in the use because from company to company, it's going to change. Within your company, even among different areas, you could have that context problem. And I think that the ability to have the understanding of what your synonyms need to be like for a term to allow everybody to understand, " Hey, this is what that means," I think that's going to be pivotal, even more so, again, if people are going to start leaning on AI, AI is going to need to know how those four things could mean this one thing.
Juan Sequeda [00:39:37] All right, time flies. Time flies. We can keep talking about this stuff, but we need to go to our next segment, which is our lightning round questions.
Cris Hadjez [00:39:47] Okay.
Juan Sequeda [00:39:48] All right, I'm going to go first. We talked about data literacy being important and I think... What about metadata literacy? Does everyone in the enterprise need to be metadata literate, or that just goes over people's head?
Cris Hadjez [00:40:02] I think it does go over people's head. I think a good drinking game is if they could say it.
Tim Gasper [00:40:09] Metadata literacy.
Cris Hadjez [00:40:10] I think that would be a great game, then you qualify to go ahead and participate. But I think, again, you have to be selective as to who you engage into that process. Similar to come boil the ocean, you working at a 10, 000- employee company, you don't need all 10,000 in, let's get five or 10, and then from there, let them choose who are the best ambassadors and like the old commercials, and I may be dating myself, it's like, " I know two friends, and you know two friends, and they know two friends," and something to that effect, and then just figure out how you build that. I agree. I don't think you can get everybody involved. And for them there probably won't be as much value as there would be for someone else.
Tim Gasper [00:40:47] Yeah, don't try to boil the ocean in terms of-
Cris Hadjez [00:40:49] With people.
Tim Gasper [00:40:50] ...getting everybody to fully understand what we may know on the data team side. All right, second question here. So one thing that's been coming up a lot recently is the topic of AI governance. Is AI governance a unique discipline, or do you just see that it's just part of data governance?
Cris Hadjez [00:41:07] Actually, I personally see it different only because there are so many additional things that an organization can choose to do with AI in terms of what they're allowing that platform to do. So in a traditional governance scenario, I typically only have to worry about things that I have within the walls, and how we go ahead and transmit things out, what needs to be encrypted. With AI, you introduce a capability of potentially the true LLM. Now how much of my data is going out? How does that work? And more importantly, is it introducing a bias that we did not anticipate? Because that wouldn't fall under the traditional governance the way people have done it for the past couple of years because they haven't had to think like that.
Tim Gasper [00:41:53] Yeah, okay.
Cris Hadjez [00:41:54] So now you've got a new spin on it. And I think it's a very strong possibility it could emerge into its own discipline. I've been asked to get engaged from the organization perspective, but I also think that as that matures, I could see how it could evolve to its own.
Tim Gasper [00:42:08] Yeah, okay. That makes a lot of sense.
Juan Sequeda [00:42:10] All right, next question. AI, AI, AI, right? So many sessions about it. Did you come away from the conference feeling that they're teaching practical next steps around it?
Cris Hadjez [00:42:21] Practical? So in order to be able to go back and do it?
Juan Sequeda [00:42:24] Yeah.
Cris Hadjez [00:42:25] The sessions I sat in on, not so much, right.
Juan Sequeda [00:42:28] So still high level blah, blah, blah.
Cris Hadjez [00:42:31] I would even say a little blah, blah, blah, but they were trying to bring it down to, " Here are what some people are doing, and here are ways to go about it." I would also say that there's potentially other conferences that are hyper focused on AI where Gartner has to touch everything because this is the data and analytics summit, so they have to focus on a little bit of everything. I think that if that's the space that you're really interested in, you're probably looking for a conference that's super hyper focused-
Tim Gasper [00:42:59] You need to find a conference or content that's going to let you go to that deeper level.
Cris Hadjez [00:43:02] Exactly.
Tim Gasper [00:43:03] Yeah, that makes sense. All right, last question here. So we talked a little bit about master data management today. Do too many companies buy for master data management when they should actually build?
Cris Hadjez [00:43:15] So-
Tim Gasper [00:43:18] I know you've had an interesting journey around this.
Cris Hadjez [00:43:20] So from my background, having done it, I can tell you that, look, ultimately if the MDM solution that's on the shelf addresses your problems and it gets what you need, perfect. But if you have unique scenarios or you need a particular, let's call it, frequency of innovation because you have unique things that take place in your landscape, you probably should build it. Now part of that is you will have to use other tools in order to make that happen. But it can be built, it can work. And some companies will sit there and say it's like, " Yeah, we built it homegrown, and it's going to be quite some time before we consider anything that comes off the shelf because it works." It doesn't need to be pretty, it just needs to work, right? Because the reality is that for most folks, they don't need the pretty gooey. Some places want it, others don't. And if you don't need the pretty gooey, then I just need something that works, yeah.
Juan Sequeda [00:44:12] All right, Tim, takeaways of the takeaways?
Tim Gasper [00:44:16] Takeaways and takeaways. We started off with talking about the Gartner conference, and what are the big themes that you've been seeing? Cris, you gave a talk as well, which was awesome. And the biggest themes that you mentioned were, okay, there's been a lot around AI, there's been a lot around data mesh, there's been a lot around all sorts of different analytics and governance topics. But the one common thread that tied everything together is metadata and metadata management. And you will struggle essentially without good metadata management, good governance to then be able to do all the things that you want to do around data mesh, around AI. The common thread is really understanding and managing that metadata. Understanding what the data is. You mentioned data literacy. And when it comes to AI, you can really use the metadata to actually learn about and understand your data better. And you mentioned like, why are we failing at this cool thing? It's because you haven't done the foundation. And that was a key thing. And I think a theme throughout our entire chat was around, hey, this whole focus around AI to move more of these advanced use cases, it's all going to drive back to the foundations again. And maybe that's a really good thing. It's going to really let us remind ourselves that, hey, the stuff that we've been talking about for 30- plus years here of what is good data management and good data governance, it's the same and we have to put more focus on it. You also talked quite a bit around partnership and how you have to partner with the business. Juan mentioned about social data engineering, you called it partnering with the business. Whatever you call it, connecting to the folks in the business where your stakeholders. You mentioned start small and build up from there. Don't boil the ocean, pick things that make sense. Start with an important dashboard or an important report and use that as your basis to start building the circle from there. Prototype and experiment. So I think it was great to hear the focus around metadata, focus around partnership. Juan, what about you? What were your unique takeaways?
Juan Sequeda [00:46:04] I think metadata is really going back to the foundations. So then you brought it up like, " Hey, 10, 15 years ago we talked about knowledge management, and then people didn't take it seriously and maybe now it's time to go do that." I think it connects to this whole theme of social data engineering, collaborative data engineering. I like what you said, we tell the story. Well, we found out that this report is wrong because there's this joint is wrong. So what do we do? And this conversation we had about the politics of knowledge. These are the types of things that we need to go figure out. How do we deal with the situations and play it out? And if it's wrong, so what? It's directionally correct. So what? These are the things that we need to go figure out, and then see if it's actually a big problem that we need to go deal with or not. Interesting to learn that the data mesh, data fabric still is a very strong topic. Standing room only in a 700- person room. Again, start focused on experimenting, iterating, you don't have to boil the ocean, you can be incremental about it. The concerns that people are having here, are you going to divide the data teams, or you just need more resources in your central team? Are you going to create some federated IT kind of infrastructure? And how does this whole social construct work? So what is the incurred cost for efficiency? So this is really important. You know what your costs are, X, and then you should go figure out what the ROI of that is and see if it was worth it or not. This is really important to go do and otherwise they'll be spending what, $ 67 million on something?
Tim Gasper [00:47:23] For learning.
Juan Sequeda [00:47:26] Yeah, learning. And then talk about MDM, master data management, the build versus buy. Again, metadata continues to be a big theme around that. There's been a lot of investment, but has there been a strong return on that investment? I guess we really need to go do more of this, understanding the cost and be able to figure out what has that return been. Start lean, higher probability of success. The concept of a golden record is great, but how many attributes does it need to be for it to be golden? That is a great question we need to go figure out. And how do we figure that out? Talking to people. And these attributes and stuff, this is all metadata. What you learn from when talking to people, all that should be connected by itself. We should be comfortable watching the horror channel. I love that. That's a T- shirt right there. We need to be able to go learn about all the failures, and we need to have more of those conversations. At the end of the day, I think the theme, we know, is metadata. But also metadata is not just a technical metadata, there's like that, the people, social metadata put it all together, and I think it's very clear that people are realizing we need to go back to the foundations. That's the context and that's the only way how AI is going to work. AI requires discipline. Discipline inaudible. All right, how did we do?
Cris Hadjez [00:48:40] It was great. It was fun. I can't believe how long it's been. What? An hour? That's crazy.
Juan Sequeda [00:48:46] But we took our notes. All right, wrap us up. Three questions. What's your advice? Who should we invite next? And what resources do you follow?
Cris Hadjez [00:48:53] Okay, so for the first one, what's next for me is being able to go back at least internally and just identify the saying, Hey, look, the practices that we're doing for metadata management, they're good, but we probably need to put more focus on it because as the organization is working on other initiatives, we've already put to the forefront just because of what we're doing in terms of some of the initiatives internally. So I think this really just galvanized that thought, it's just like, " Look, even after they finish this next step that's on the horizon, this needs to be ready for them at that point." So for me, it really helped me inaudible that. Who's next? I would tell you... And I want to make sure I pronounce his name correctly, but Ehtisham Zaidi.
Juan Sequeda [00:49:38] Ehtisham.
Cris Hadjez [00:49:38] Ehtisham. So, Ehtisham was, again, we could argue it's like, hey, I'm a bit biased, and maybe that's why I heard so much about metadata. But the reality is it was an opening keynote. I don't get to pick opening keynote, opening keynote is the opening keynote, and he hits it right out of the gate. And then he also did the data fabric with data mesh. He really went into a lot of these disciplines and AI. He went into all of those areas that we covered here, which is interesting. It's like, okay, so let's hear it from his perspective because he probably covered this from his view. So I think that would be an interesting chat. And then for resources in terms of what I do, I shared with you that for me, typically, I don't tend to frequent the same conference year over year, that's not something I tend to do. I pick what's the conference that gives me a different spin or a different view. If, for example, like I was saying earlier, if there was a much stronger appetite for us to have to understand AI, I would probably go focus and identify a conference, it doesn't have to be large, but what you're really trying to do is you're trying to get with other peers, it's like, " What are you guys doing?" So even a conference of 500 people, you could come away with value if you're willing to engage in a conversation. If you're going to be passive about it, then I think your return on those events are much lower than you probably want.
Tim Gasper [00:50:53] It's very wise inaudible. A lot of people tend to go to the same conferences, maybe it's Gartner, but if you really want to learn, you need to expose yourself to different content, deeper content, and different people, and be willing to engage in conversation.
Cris Hadjez [00:51:05] I think that's the key, honestly, to go to a conference and just not have a conversation, just sit in every session, but never ask a question, never go to a booth, never talk to a peer, never take the time while you're at lunch or breakfast or dinner to talk to somebody, you could have paid for the streaming and just stay home.
Juan Sequeda [00:51:21] That's true. That's great advice to be able to say, go pick your theme that you want to go learn because what you were telling me before too is if you come again to this conference next year, it's going to be the same type of... There's not a big change that's happening, so go do that. Hey, we met at the conference about Kafka streaming.
Cris Hadjez [00:51:41] Yeah, and that was completely... it was a purely focused Kafka conference, and that's where I bumped into Juan. And he was like, " What are you doing here?" I was like, "I'm just trying to figure out a little bit more. We're going into this space, and I want to get more educated." But in that entire time, I had shared with Juan, I didn't sit in on many sessions. It was a three- day conference, I think I sat in a total of four sessions, I spent my time on the floor just talking to all the vendors, " Tell me what you're seeing. What are you doing? Why do you think your product's good?"
Juan Sequeda [00:52:09] Interesting responses when you ask that.
Cris Hadjez [00:52:10] Yeah, Just put them on that, I guess, uncomfortable playing field and just say, " Okay, learn about it. Learn about it. Have the willingness to learn."
Juan Sequeda [00:52:22] This is great advice. Well, next week we have Scott Taylor, the data whisperer, that's going to be on, he's such an awesome guy, and looking forward to that podcast episode. With that, Cris, thank you so much.
Cris Hadjez [00:52:34] Thank you, guys.
Juan Sequeda [00:52:34] We appreciate it.
Cris Hadjez [00:52:35] I loved being a guest. This was a lot of fun.
Juan Sequeda [00:52:36] And as always, thanks to data. world that lets us do this every week. Good fun. All right. Bye, everybody.
Cris Hadjez [00:52:41] Bye. Thank you, everyone.
.png)
