Hello.

Good morning. Good afternoon. Good evening. Good whatever time it is wherever you are. Whatever you are, I am Malcolm Hawker. I’m the host of the CDO Matters podcast.

I’m also the head of data strategy at Profisee software.

Thought leader on LinkedIn, creator of a lot of content focused on you, the chief data officer, and people who wanna be you. VPs of data and analytics, managers of data and analytics, data stewards, data governors, data quality people, you name it. If it’s data, we’re gonna be talking about it here on the CDO matters podcast. I’m thrilled that you are joining me today.

As I talk, it is the middle of December. And when you see this, it will be towards the end of December 2023. So for our friends that celebrated, I hope you had a Merry Christmas for others. I hope you were having a wonderful holiday season. I hope you’re having a wonderful end of the year.

I’m looking forward to 2024, but while it is still 2023. We’re gonna kinda do a little bit of a end of the year round up here. I like these things.

Whether it’s a list or whether it’s, you know, your top whatever, I was looking at my top place from my Apple playlist this year. No big surprise. I mean, I could have guessed them anyway, but I don’t know. There’s something cathartic about retrospectives, right, and to be able to kind of look back. Maybe it’s a time to celebrate. Maybe it’s a time to mourn. Maybe it’s a time to learn.

I don’t know, but I fear anything like me, I I like the retrospective stuff. I think it’s I think it’s fun and I I I think it’s a good use of time. So that’s what we’re going to do today on the CDO matters podcast. We’re gonna talk about the things that I’ve saw the most. In 2023. Some of the key themes as it were for data and analytics leaders throughout the year.

A few kind of housekeeping items.

I suspect you will be listening to this in the last week of December.

By which point I would have already have had a live event with Erik Zwiefel, the CDA of Microsoft.

And we are going to post that episode two weeks after this one. So January 4th, I believe.

Maybe I can’t quite track time early January.

Look and early January, the first episode of twenty twenty four about this podcast will be a live conversation. Of course, you’ll be watching it or listening to it as a recorded conversation. If you don’t attend these live event on LinkedIn on December fifteenth, which is in the past for you now.

But if you didn’t attend that live event, you’ll have a chance to listen to my conversation with Eric that we are recording literally tomorrow, because today is Thursday and tomorrow is Friday, which we’re gonna, we’re gonna do it. Eric is just an awesome guy. I I’ve I’ve got a chance to know him a little bit and get to know him a little bit over the year.

Had a wonderful conversation with him over dinner in Boston, I think. I don’t know. All the cities are melting together. That sounds horrible, but it’s true.

Just just a just a super, super smart guy. A really nice guy. Somebody I enjoy talking to, I hope that comes through tomorrow in our recording. When we’re talking about the state of the data nation.

And, of course, he’s a CDO for Microsoft. We’re gonna talk about open AI. We’re gonna talk about AI. Well, we’re gonna talk about the Microsoft fabric. We’re gonna talk about other things front of mind to CDOs. We’re gonna talk about what’s going on at Microsoft.

I think Microsoft is head and shoulders ahead of the cloud let’s just call them the cloud hyperscalers, when it comes to, data management, data structure and creating, what would otherwise be called a modern data ecosystem? I think I think Microsoft is leading the pack, and I’m really looking forward to my chat with Eric tomorrow. And I hope that you can catch that.

When it comes out as a recorded version of CDO matters, because I think it would be well worth the the hour to listen to it.

So I think that’s the only housekeeping item for for today. Please look forward to that. Let’s talk about twenty twenty three. Shall we?

Wow. What are some of the key themes? One, when I look back in the year, I think the year really, really started with the the one word that still sticks in my brain because it was sticking to my brain in January of twenty twenty three and it’s still now was uncertainty.

And we started the year with a lot of uncertainty. You know, we were coming off of the kind of the reopening of economies after after COVID and things were going reasonably well.

But China was still kind of mired or China was still stuck you know, shut down. There were still at the time supply chain disruptions and shortages. There was still a lot of uncertainty of what that would mean for for kind of GDP nations around the world, consumer demand around the world. There were some kind of dark signs in the horizon with the beginning of discussions around interest rate hikes, and on and on.

And we started the year with a lot of uncertainty. There were a lot of people talking at the beginning of the year about, the recession. That was a a word that was being bad at around a lot at the beginning of the year. And there were a lot of people who had fairly gloomy outlooks of twenty twenty three.

I can say that at least from my perspective of my experiences here in the United States, in Canada, and, to a lesser degree, I I I I wanna say Europe was was arguably hit a little harder, just a little.

But for the most part, it wasn’t as much of an economic meltdown as a lot of people thought.

Yes. People, you know, are losing their jobs. You know, the news recent news is a lot of people being laid off from Spotify and and I personally know many people who were negatively affected this year by downsizes in in kind of in in the the corporate world, but at the same time, it wasn’t nearly as bad I mean, I was around folks for two thousand and eight.

I was around for for recessions in the mid nineties.

I have been laid off more than once.

It’s a growth opportunity and the time I look back and it was like, wow, I learned a ton from that. I learned a ton about how to manage my career. I learned a ton about how the right job and learn a ton about how to pick the right leader.

But at the time, it’s absolutely no fun at all. It’s it’s it’s it’s a horrible experience. It’s, causes an awful lot of self doubt, not to mention a lot of stress, particularly if you have a family particularly if you’re mid to late career like I am, late getting laid off is just it there’s it just sucks and it’s easy to look back and say, oh, well, the group growth experience and, you know, beyond every closed door, there’s a new open door and yada yada yada.

But the story here is is that at a at a at a macro level, I think we can say that it wasn’t the news wasn’t nearly as bad as we thought it was gonna be. Of course, for those of you who may be in Israel or who may be in the Ukraine or maybe in other border areas of the world.

The the news is not Rosey.

And and it breaks my heart that these things are actually happening in our world today. I I really wish we would have evolved beyond that, but you know, I I I mentioned this in the last episode. I really wish that we had more peace in the world.

But I think at a high level, insofar as there is suffering around the world and so far there is war from an economic perspective, I think the year was not nearly as bad as we thought it was going to be.

Particularly when it comes to corporate America, corporate demand people needing the services of data people like you and I.

The services are needed. Right? And they remain in demand and whether you are providing software or whether you are a data professional, maybe you’re a consultant or an analyst, twenty twenty three, I suspect was generally a pretty decent year for you, not a horrible year. And for that, I think we can all be thankful.

Number two. Number two. The return of the live conference, the in person conference. This started, of course, in twenty twenty two.

We started going to to conferences again towards the summer of twenty twenty two, the first one I went to. In, in July of twenty twenty two, I got COVID, as did a lot of people that I knew as well. Through no fault, of course, of the, of the conference, you know, coordinators.

But but twenty twenty two is still a little tenuous.

Twenty twenty two, there were a lot of people who were still staying home for for very obvious reasons. The twenty twenty three, I’ll tell you, man, whoo, wee. And I’ve got the proof to I’ve got the proof here. I’m gonna this is my mister t moment.

There we go.

This is this is my proof that this is and by the way, this is just a sample. This isn’t all of them.

And if and if you’re if you’re listening to this and you’re not actually watching the video on YouTube of me making a fool of myself, I just put all of the lanyards conference passes of a sample of events that I went to this year around my neck and I kind of looked like a a data mister t instead of gold necklaces that well, I bet many people don’t even know who mister T is, right, Google it, the eighteenth, mister T. He used to wear a lot of, a lot of jewelry around his neck, but my jewelry is, passes. As you can see, some of them have some bling. Well, most of them probably do, speaker and and whatnot.

Anyway, yeah, the the live conference return, I’m taking off my out of my passes.

Next time I call somebody a data hoarder and you hear it, Yeah. You need to tell me, hey, dude. You’re hoarding, your conference passes. That was fun.

I need to put those back on for a photo. I think that’s kinda silly. It also kinda fun. And yes, those of you who are environmentally sensitive, and more aware of these things would be right to ask, hey, why don’t you just get one line here and take it from conference to conference?

That’s what I should do. This is what my friend wants to cater at Data dot World does. He goes to even more conferences than I do, and he has just one lanyard. I guess he’s more environmentally sensitive than than I am.

Anyway, conferences returned and they returned in a big way.

For this, I’m really grateful. You know, I I think that conferences are a wonderful opportunity to learn. Conferences are a wonderful opportunity to interact our peers, conferences, or a wonder of full opportunity to just get away from your day to day environment, like just breaking out of your maybe that’s who you work at home, maybe you work at an office, doesn’t matter. But getting out of your those surroundings and going to different surroundings is an opportunity for you to just think different. Play. It’s it’s an opportunity to to not be bound, I think, to buy a lot of the constraints that go into a nine to five job. That’s one of the things that I love the most about conferences is that it gives me the space to think.

And for somebody who does what I do, right? We create content, share best practices, share insight, talk to people.

Everything I know is a function of what I’ve learned from others.

So for the opportunity to go and talk to people, is is unbelievably invaluable. It’s invaluable to me, but it would be invaluable to you too if you were going to conferences. So I recommend them wholeheartedly.

That said.

I do think the conference is I don’t wanna say they’re ripe for disruption. I would just invite conference organizers to take a fresh look at your content. I think we really, really, really need to revisit some of our content because I think a lot of it is getting a little old and tired. Now that said, I can say, and this is an amazing thing.

And maybe this is a function of of of of of of a good job that the conference organizers are doing.

For many of the conferences that I’ve been to this year, a good fifty percent of them were first timers.

Which is, I think, in total, a really, really good thing, you know, that you you could kind of neg look at it negatively and say, oh, wow, we’re really turning over. And maybe that’s a bad thing, or maybe it’s a good thing. I think having more and more people in our space is generally a good thing. I know for a fact that, a lot of the new people who are coming were not taking jobs of old people. These are new people who are coming because companies are investing in data companies are doubling down or investing in data quality and MDM and data data integration because I heard that at these conference over and over and over again. I was blown away by how many times I heard we are beginning our journey on.

We’re beginning our journey with MBM. We’re beginning our journey with data governance. I heard that often. Right? So This is a really good thing. And again, going back to point one about growth in twenty twenty three, the fact that the conference circuit is alive and well, The fact that we’ve got a lot of new people coming into these conferences, the fact that we still have companies who in the year twenty twenty three are starting their kind of data and analytics journey, they’re at least their data management journey, data governance, NBM, data quality. The fact that we have companies that are just now investing in these things and standing up, data and analytics organizations, awesome stuff, really, really good stuff.

We are growing as an industry.

The CDO role is growing. Man, I would like particularly in the government space Holy cow.

What did the CDO IQ conference in, in July and Boston.

And one out of every three people I would say that came up and introduce themselves or that I had a con conversation with because we had a booth there, and I was working the booth shaking hands. As I do, One and every three was. Yeah. I’m the CTO for insert federal state county or city government. I heard it over and over again.

And that’s a wonderful thing.

One of the things that I found interesting from that though is that is that often you were talking to people who were clearly more mid career. Right? Typically, you would expect a CDO to be late career, later career.

And for a lot of these folks, sorry folks. I’m sniffling.

Usual, usual seasonal allergy stuff.

Two weeks ago, it was getting over a cold now. It’s allergy. So it’s a frying pan fire.

A lot of these CDOs are Mickerer.

And that’s a wonderful thing. But it also means that often, you know, I would say that there may be a relative shortage for lack of a better word of some of the real world experience that, you know, CDOs may otherwise need for success. This is an opportunity for consultants. It’s an opportunity for people who like me who share best practices and or devoted to helping CDO succeed in their role. So, it’s just a it’s just a good thing. Conferences are back, Our space is growing. There are positive signs in the horizon.

Can’t talk about twenty twenty three without talking about artificial intelligence.

Boy.

This is this is, of course, is the number one topic, by the way. I mean, it’s, but, like, miles ahead all the rest of them. It’s it’s the number one topic without a doubt.

I’ve seen such an interesting arc in a really short time, whether we’re talking about the arc of the technology itself, or whether we’re talking about the arc of just kind of general awareness of AI within the data and analytics space.

Let’s talk about the arc of the technology.

Well, yes, we still are mostly talking about generative AI. And large language models. I think that makes sense. There are a lot of companies that have been doing AI for a while. Depending on who you ask, it is about fifty percent of companies have some form of data science function.

That function can be formalized in the data analytics team or it could live ad hoc more ad hoc for lack of a better word within specific business functions. You could have a data scientist living in a marketing function, for example, working on propensity models, that kind of thing. But give or take about fifty percent fifty percent of companies have some form of data science role in them.

For those companies. I I think they kind of had a little bit of a head start.

But when talking about kind of this arc of the technology, data science has always been there. People have been doing it a long time. They’ve been building models a long time. This has been known, and it was really LLM kind of stole the show and continue to steal the show and are really groundbreaking for a lot of different reasons.

I’m not gonna go into why LLMs are transformational by nature. I think that kind of the proof is in the pudding. Right? This isn’t just another shiny object.

Right? This isn’t another hadoop.

How do we know? Well, because teenagers are using CHFTP to do their homework.

There are real use cases here. I mean, the list the list is very, very long. I won’t go into that.

But what I’ve seen, you know, in in a year, we saw tetchy BT, free was was released in November twenty twenty. Two and in a year now, I think we’re on to version four, version four turbo.

And a whole bunch of other elements have come along. Right? There’s been there’s barred, there’s llama from Facebook, which is open source.

X, Twitter just came out with one called Grock.

Google has upgraded. Well, they didn’t upgrade Bard. It’s still out there, but Google has their, the deep mind product, which, which is multimodal that is text and video. And, I mean, this looks unbelievable.

So there’s more and more to choose from out there and more companies are racing into the space.

The arc of kind of knowledge within the data and analytics space we started, I would argue incredibly flat footed.

Because I was having a lot of conversations at the beginning of the year, with CDOs who were saying, well, okay. Yep. There’s this thing. There’s these LLMs, and that’s great. But we’re not ready because I can’t use my data to train a model.

And what I would say to that is, well, you’re not going to train a model.

The the the the the the P in GPT, generator pretrained transformer, the piece debts were pretrained. So if you’re using a commercial off the shelf for profit, LLM, like a chatty PT then you’re not gonna be training it. It’s trained. It’s out of it comes trained out of the box. Right? And these training processes are insanely resource intensive and, cost intensive.

Right? So as a simple comparison, a GPT four has seven hundred billion parameters, LAMA has seventy billion parameters and to create LAMA required six thousand GPUs, graphical processing unit, six thousand GPUs to be running non stop for twelve weeks, using the entire internet.

As it’s fact based. And and by the way, using a whole bunch of text data, not using any rows and columns to do it. So the beginning of the year, I was hearing CTO say, we’re not ready. My data is a mess. My data is there’s so much garbage in my data, and I can’t use it to train any LLMs, and I was hearing this over and over and over again.

And I was saying, hey, don’t worry about it because you’re not gonna be training LLMs in the short term. Nobody will. And chances are you never ever will need to because by in a matter of months, LMMs are gonna be largely commoditized.

And and and and I think we’ve seen this. Right? LOMs are now competing on accuracy. They’re not competing on how much more of the internet can I go grab an index? They’re not indexing, but you get my point.

Nobody’s competing on, well, can I find more data than the internet yet?

So I would argue the LMS have become largely commodity and now they’re just competing mostly on accuracy in terms of things like how they score from a computational perspective, from the mathematical perspective, how accurate they are in their responses and on and on. So you’re not gonna be training in LLM unless you live in the business of of of selling AI solutions. You’re not gonna be creating an LLM. What you’re probably going to be doing with LLMs, which is probably true for ninety five percent of companies is finding ways to extract value from them.

Right? What you’re probably gonna be doing is finding ways to get an LMM to give you more accurate predictable consistent results.

There are ways to do this. Yes. Complex prompt engineering. Today is the way to do it. And I’ve been making a lot of posts through the year about this thing called the rag padding retrieval retrieval augmentation where In essence, what you do is you pass a known factset into an NLM. If you say if these things are true, then summarize this for me, then make conclusions based on this, do some math based on this. That’s one way that you can most certainly reduce what we call hallucinations.

So the arc of the technology here, we’ve gone we’ve gone from and the arc of the kind of the the the knowledge of of CDOs in the space has gone from Okay. I really don’t know how this stuff works.

I I think I’m gonna be trading LLMs even though I’m probably not.

To now we’re like, oh, okay. I get it. We’re I I think we’re probably gonna have to find ways to utilize these these systems. And now we’re rer towards the end of the year is that I think that most people are starting to write see that how they will be interacting with AI is through chat bots and through agents.

Where these agents are the ones, and I will just loosely use the word agent You could call it an application, actually.

Probably a better way to look at it. If you look at LLMs as an operating system, and agents or software on top of the operating system, I think that’s a more maybe a more useful metaphor here.

I think what we’ll be seeing is that people will be interacting with LMS through these agents. The agents will kind of bundle up prompts And GPT four, by the way, can accept three hundred pages of text in a prod. Now there’s a lot of tokens there, and that’ll be expensive. So so warning, don’t don’t go encourage everybody in your company to go submit three hundred page prompts multiple times a day because the meter’s running. That’ll be expensive.

But you can’t. Right? So so I think what you’re gonna be seeing is is more and more apps for lack of a better word. Where smart agent technology is used to give you a response that is highly accurate, highly consistent and highly predictable.

And where those agents are really probably kind of tuned towards specific use cases.

So that’s where things are going from an AI perspective. We’ve gone from a little bit of the wild west. Don’t know how this stuff works to okay. Maybe LLMs are how how most AI is gonna be operationalized scale, which I believe to be true, by the way, the short term, for most companies, AI will be operationalized at scale through these LLMs and through these smart agents.

Into that world of smart agents where you were basically passing data to an NLM.

Right? Maybe you’re looking at a power BI report and you’re asking an LMM to summarize it. Maybe you’re looking at some other data that came from a sequel database and you want to draw some conclusions from it. May you maybe you need to help doing some analysis. This is another thing that’s happened a lot with AI here. And now all of a sudden, we have data citizen data analysts, like, where in the past that you had to hire somebody to do the analysis, now you can have the l o m’s to do it for you. But in this world of using smart agents to do this, What you pass through the smart agent, what you assume to be true or what you tell them to assume to be true is highly highly relevant to the quality of what you get back.

So we started the world with we started the year with this mindset among CDOs that, oh, garbage in garbage out, but it was through the lens of training things.

And I was the I was saying, hey, hey, hey, garbage and garbage out here is not nearly as relevant when you’re talking about training LLMs because you’re training off the entire internet. The internet is not known as a bastion of high quality data.

So when you’re talking about training LLCMs garbage out, I would argue it less relevant than it’s ever been. However, in a world where in a world where smart agents are taking data that you give them, taking facts that you give them and drawing conclusions on those facts, garbage in garbage out matters. And I hate the phrase garbage and garbage out, by the way. It’ll be another topic we talk about here shortly. But the quality of what you’re passing that agent What you were telling the agent is true is factual is highly relevant to the quality of the output.

So, yes, it does matter that we need to get our data houses in order.

However, however, I would urge CDOs to avoid the temptation of saying, Hey, we can’t get any value.

From AI until we get our our data house in order. What I would urge is CDOs need to find use cases that you can leverage now.

Using existing technology.

Right? There are most certainly going to be some use cases in your organization that could benefit through the automation of some sort of natural language experience, whether that is maybe some sort of customer service chatbot Right? Maybe there are other potential use cases here where you could leverage this technology where sensitivity to data accuracy is less material than it would for other use cases.

So this is point number three of the year twenty twenty three in review.

What I saw in this year as a part of this kind of this bumpy road of AI that we’re on was a lot of CEOs react with fear and and rightfully. I mean, I get it.

But a lot of them react with fear and basically say, stop.

Stop. We gotta get everything order we we need we need to slow down.

Let’s not jump off the cliff on this AI thing just yet. Let’s just stop and let’s slow down. And as recently as last week when I was at a conference holding a round table of about twenty five data and analytics professionals, I asked the question of how many have kind of like outright banned LLCMs within their organization, and there were a number of heads that went up.

Still. Right? And I get it. I get it. I’m not saying that’s bad and I’m not judging.

I am saying, however, that boards of directors are expecting movement here.

Company senior executive c levels in the CEO, the CEO, others particularly boards are expecting your company to have a story around AI. And if you’re the CTO or in a CTO type role, VP of data and analytics, however you title it. If you’re in that role and you’re the one saying, no, we can’t because we’re not ready or we don’t have our data house in order, I would say that that represents a risk to you as a CDL, particularly given that that was the same thing that people were saying two to three years ago when a lot of those very same boards of directors were asking for digital transformation and digital transformation in a hurry.

Right? Twenty twenty when everything went sideways, a lot of boards of directors were saying, Hey, we need to find a way to respond to these disruptions quickly. We need to respond to global. Supply chain disruptions quickly.

We need to respond to changing consumer behavior quickly. Hey, data people. Can you help? Nope.

Sorry. I can’t. Our data house is not in order. Give me a couple of years to figure it out.

Which is what a lot of CDOs were saying in twenty twenty, which is a lot of the CDOs are saying now in response.

To a desire to go quickly on AI, I would urge CDOs to avoid that temptation find ways to deliver value now.

Even if your data house is not in the order that you wish it to be. Find ways to deliver value now because if you are that CDL, I would say that that represents risk to you.

So that’s artificial intelligence. That’s kind of some of the things I’ve been hearing around. CDOs and AI readiness. Yes. AI readiness will be a is a topic this year, and it will continue to be a big topic next year. I know this to be absolutely true.

But it doesn’t mean you need perfection.

You need to be AI ready, and there are things that you can do to enable that. The best thing I think that you could do to enable become more AI ready is learn by doing.

Go find some use cases where these LOMs can be helpful.

Find ways to not ban LOMs within your organization, find ways to leverage them within desktop applications, like maybe like Outlook or or within Power BI or others where some of your vendors are happy to help enable these capabilities because they can charge you more money. Find ways to leverage these these technologies now to deliver value now. So you have a story to tell.

While you figure out all of the bigger picture things, like, while you figure out governance and AI governance and ethics and all of those big picture things that will take some time.

Find ways to value to find ways to deliver value using AI now.

Number four or five, I’ve lost my track.

Because I can tend to ramble as any of you who have been tuning in to me for a while now, no. It’s part of my charm. I hope.

Help.

Number four or five. Let’s go with five. I like five.

The complete evaporation of the data mesh. Wow. My goodness. Twenty twenty two, you you you couldn’t go to any conference.

You couldn’t log in to LinkedIn. You couldn’t you couldn’t have a conversation with a lot of people without the data match coming up as an item. As an issue. I was still, at Gartner in early twenty twenty two and it was mesh mesh mesh mesh mesh mesh.

That sounds a little silly. Smash, smash.

But true.

It was it was right at the top.

Of Gardner’s hype cycle. It is kind of slid down towards the trough of disillusionment on the gardner hype cycle. I would say it is in the pit of disillusionment.

I think there’s a princess bride metaphor there somewhere with the pit of disillusionment.

I would say that that’s exactly where it is now. And and and and next year, it probably won’t even be on the hype cycle. I’d be willing to bet. That’s my twenty twenty four prediction for the data mesh. It won’t even be on, wouldn’t be on the hype cycle. And by the way, by the way, two years ago when it was on the peak of the hype cycle, Gartner had it labeled as obsolete before plateau.

So there was an x through it. Oh my gosh. The pearls were being clutched. What do you mean it’s obsolete? What do you mean the data mesh is obsolete?

Well, what what the x meant is that it it it will die before it reaches mainstream.

And that’s exactly what’s happening.

So vindication to Gartner and the creation of the date of the of the hype cycle.

There’s a lot of good reasons why the data mesh is completely and totally fallen off the radar.

The biggest reason why is is because it’s really, really hard. It’s it’s complex. To do a full mesh, full mesh, a real mesh.

At least insofar as a mesh would be defined by Zhanak daghady, the author of datamesh, You need to do a few things that are really, really hard and that most people don’t know how to do.

The biggest one being something called federated computational governance, which is in essence automated data governance.

How many of you have automated your data governance? How many of you even have people who are showing up to governance committee meetings on a recurring basis?

Exactly.

Right? So the idea that we could completely totally automate governance.

Okay. Another is you would need some sort of unified platform for managing and distributing data products across your organization that would it be helpful in in helping enable some of that automated governance as well. How many people have that? Well, some people have some data catalogs, but, do they have a full data marketplace?

Right? Have they enabled that? Well, no. Again, that’s hard.

Data contracts. How about that? You got those?

Those widely used within your organization? No. No. No. Probably not. No. Maybe here or there, and a few kind of machine to machine communications where you’ve implemented some form of a contract, via an API where data is being shared perhaps externally or maybe even externally, maybe onesie twosie, but under wide set braces, basis?

No. No. Don’t have that.

Do you have, you know, product ownership on a widespread basis?

Well, no, but we’ve hired a data product owner. Okay.

That’s it. Done. We have our mesh. And that’s really what ended up happening. Oh, and by the way, along with all of those things, you would have to dispense of what would call what Zemoc would call antipatterns, which includes any centralized pattern of managing data, including MDM and your data lake and data warehouse. Those are anti patterns to the data mesh.

And if you disagree with me, read the book, because I’ve read it three times, I had to read it three times to understand what was being said. Maybe I’m just not that smart, entirely possible.

But dispensing of anti patterns. Right? So if you wanna talk about a full mesh, you gotta go peer to peer completely.

Like, we’re talking full spider web here people. Right? But spaghetti bowl, right, where it is like sharing, peer to peer sharing, or domain to to domain more arguably. Correct. Domain to domain sharing where the infrastructure, the governance, the operating framework to enable cross domain sharing just doesn’t exist. Well, it exists, but it exists in this really conceptual theoretical model that nobody’s ever implemented before.

So you throw all the things together.

All of the first four that I mentioned hard, hard, hard, hard, what you’re left with is data products, which leaves me to topic number six. The bell of the ball of the industry conferences this year that wasn’t AI was data products. And I think the data products are really kind of what’s left over from the implosion of the mesh. The data mesh was kind of like the situation where a whole bunch of things were loaded onto the back of a flatbed truck, and then somebody hit the gas really hard. And then ten miles later, the only thing that was left in the back of the truck was data products.

Uh-huh. Here’s our mesh.

And really it wasn’t even data products, but it was a data product ownership, which I would argue is is is from the mesh perspective, falls woefully short of what real product meant, the value that real product management could deliver in the data and analytics space.

But what a lot of companies were left over with was, you know, data products. I thought the only thing I can kinda get my hands around. I like the idea of assigning ownership. From a governance perspective ownership of governance policies and procedures and the enforcement of data governance policies.

I like the idea of enforcing that at a domain level. My customers in sales and marketing have been complaining about a lack of control of their own data. So I’ll assign a data product owner. You can go figure out the governance within sales and marketing.

Knock yourself out. Uh-huh. Here’s our data mesh.

That’s really what happened and I think that the craze this year around data products is a leftover from that. And as much as it pains me to say this, because I really, really, really do believe that product management when applied to data management can have transformative impacts. I think there’s so much goodness here. The list is long. I’ve done a podcast on this already earlier this year. Check it out.

Really think that there’s gonna be some goodness there, but frankly there’s just not enough juice in the squeeze to keep data products on people’s minds because I think in twenty twenty four, it’ll just it’ll it’ll be the last thing that falls off the data mesh truck, but it’ll fall off. Because most people aren’t doing it, what I would say the right way. All they’ve done is made somebody in who is in a early to mid career position, they’ve made them the owner quote, quote, of data, particularly shared data.

That could never ever truly be owned by a single person.

Untenable position.

So will be my that’s a prediction for twenty twenty four. We’ll talk about that in a couple episodes, but data products will will will go away. And as much as it pains me to say that because I wish they didn’t.

That’s data products. That was number six. I will be talking about that in, at a number of different conferences this year in twenty twenty four. The the the title of my presentation is Hooray data products.

What’s a data product?

That’s the reason why it’ll probably go away next year, which doesn’t bode well for people coming to my presentation.

But we’ll see how it goes. Maybe I’m wrong. Won’t be the first time. Ask my wife.

But yeah, there’s a lot of people out there talking about data products that really just have no idea what a data product is. And and frankly, I just posted something on LinkedIn about this. I really don’t care. It was just so much just just hand wringing around. How do we define a product?

A product is something a data product is something created by a data team that solves a specific business problem that business stakeholders would otherwise be willing to pay for that customers would be otherwise willing pay for. That’s all it is. There doesn’t have to be this elegant, hierarchical, definition of products that works going up or going down. You don’t have to define the lowest atomic level of a data product for it to make sense.

You can just say it’s something that solves the problem. That people would otherwise be willing to pay for it. That’s enough. We don’t need it to be this great wonder ontology that that, that that can be traversed forwards and backwards.

It doesn’t have to work that way.

Products can be things that are more loosely defined and more loosely classified yes, there are product families. Yes, there are hierarchies and products.

But the fact that a raw material can be both a raw material and a product shouldn’t be confounding to data people, but it is.

What else twenty twenty three? Well, When I was at all those events, see my bling before.

I was talking about a lot of new people in the field.

Which is a wonderful, wonderful thing.

But that means that there is still a real first and a real desire for what I would otherwise call best practices around data and analytics and data management foundations.

And as much as I’m prone as the next person, because I I love talking about some of the higher level stuff. As much as I’m prone to that, there is a desire to continue to focus on best practice is now that said that said, I think some of the old ways of doing things are really, really broken.

The idea that we can say some that data can be owned by one person is a classic example of how broken things can be. That is that that idea of data ownership being a necessary dependency for data governance or data the idea that that is a widely held belief is indicative of some of the problems we’ve got in this space. It’s not working. Data ownership is not working and it never was able to work for any data that is shared outside of one system. If your data lives in one system and it just never goes outside of your CRM, it’s never used by anybody outside of sales and marketing. Sure. Give it an owner in sales and marketing.

Fine.

Right? If it’s one of the two hundred fields in your customer record in Salesforce that nobody cares about other than the CFO, go ahead knock yourself out. You can give that one person on that data. But the minute that data starts to go somewhere else, the minute it needs to go in into an ERP or into some into a GL wherever, the idea that one person can own data is a recipe for success because the only person just keep going up, the org chart. The only person who would really ultimately own that data is the seed.

And how’s that governance committee gonna meeting gonna look? There’s only one person sitting there and it’s the CEO. Well, it’s not gonna happen because they got better things to do.

So that’s, you know, the idea of data ownership. And the list here is just so long. There’s a thirst for for best practices. We need to figure out some things that have been holding us back for years and years and years. This is number eight for me and Number twenty number eight in twenty twenty three, I’m just gonna loosely call modern data leadership. We need new ways of thinking in the data and analytics space.

We’ve been trying the same things for twenty years.

This round peg and ramming it into a square hole. I I I like to pick on data ownership because it’s a layup.

It’s a layup, but I could keep going.

I could keep going about how we approach things. How about the words that we use, about the mindsets that we have in the data analytics team?

The idea that data quality issues are a burden and not an opportunity. That’s an example of a mindset that just does not service.

The idea that data can be structured one way in a CRM and data can be structured a different way in ERP, that that is a data quality problem. To me is a total disservice. It is not actually accurate. You can say that that represents a data quality problem, those records could be completely and totally accurate. They just happen to be structured in a different way because those applications were built a different way. The applications were built a different way because they had different users and different use cases and different processes, meaning they were built by design.

So the idea that we as data leaders would say, oh, well, that that’s garbage in.

When it might not actually be garbage in way may actually be data that was structured that way by design to serve a specific use case to help our customers accelerate their business processes.

That’s what that’s an example of a mindset. Be calling that data garbage is an example of a mindset. I think we need to change. It’s not helping us.

It’s not helping us because frankly people are tired about hearing us complain about data quality, particularly companies that are successful see my first point. A lot of companies are this year and a lot of companies had great years in twenty twenty three.

So You know, we need a different way of thinking about data. We need a different way of thinking about data quality. We need a different way of thinking about to management. We need to be more flexible, more adaptable.

We need to think about truth differently Right? The fact is truth is contextually bound. Just is. Always has been. We’ve always known this yet. We continue to cling to the idea of a single version of truth.

At a CEO level, that’s a context.

A CEO CFO, that’s a business context, and it’s a valid business context. What’s true to that context may not necessarily be true to an operating level. And we need to recognize both and to say that one is true and one is not true, Again, that’s an example of a mindset really not serving us. So twenty twenty three, I think, will be remembered by many, including me, as the year when we started to look things a little differently. And we started to challenge legacy approaches to how we think about data, how we think about leading a data function.

How we think about culture.

We like to talk a lot. I had to check out my episode from last week. If you wanna hear more about data culture, I talk for an hour about how we need to think differently about culture. So I won’t go into detail here.

But we use, sadly, culture issues as an excuse to explain why.

Our data teams are not delivering value. And that we we just need to stop. We just need to stop because chances are it’s probably not the culture. Probably isn’t.

Check out last week’s episode more on culture. We need to be thinking more about new ways of managing data and analytics functions, we need to think about new ways of serving our customers. This includes product management.

It includes a lot of different things. It includes the structure of the team. It includes all sorts of things that I’m so excited to be talking about in twenty twenty four because there’s tons of opportunities here folks.

Few other things. Kind of like, I’ll get into a little bit of a lightning round.

Data modeling, dead or not. Boy, there’s a lot of people who care about data modeling. It seems like this this this data modeling machines can do it. Are you crazy? Machines could never do this. This is pure art.

I I fall on the machine side. I’m gonna I’m gonna give this one of the machines. Sorry. Like, the the idea that human beings could be faster, more scalable, more accurate in the long run than machines, I mean, please.

Doesn’t mean there can’t be a human oversight, and it doesn’t mean that this is happening today. I think that humans can can deliver some pretty good results today. But in the long run, I I I cannot I cannot envision a world where data modeling has not been mostly automated Put your comments below. There are some there are some people out there that are just VM at data modelers that this is like scoff at the idea that this could be something the machines do. I think that’s largely a reaction, to potential job loss. Just just is what it is. If you do data modeling for a living, of course, you’re gonna say the machines can never do it.

So data modeling dead or not, it’s not dead. It’s critically important. It is literally the frame of the wing of the flying plane, so don’t get me wrong. All about data modeling. I just don’t think that humans can do it as well as machines can.

Sonantic layers. A lot of people talking about semantic glare This kind of falls in the data product bucket. Yay, semantic glare. What’s a semantic layer? Sounds great.

How do I actually do it?

I really envision or I would like to see a future where semantic layers and MDMs come together because MDM software, it’s a semantic layer. Okay, it it it allows common definitions to be created and maintained. Right? In fairly complex ways, Right? Same as true with semantic layers, although they do it at a scheme, at a schema level and not an object and field level.

I love to see those things kind of come together in the future. So maybe we need to get semantic layers and MBM on a first date to see if any magic happens.

Data fabrics I’m a believer. See any previous podcasts related to data fabric demystified. It’s one of my top performing podcasts. For good reason, I share a lot of great insights on data fabrics.

See my conversation happening tomorrow, that will be released the first week of January twenty twenty four with the CEO of Microsoft talking a little bit more about data fabric. So I’m a believer. Now, because my my company is a partner of Microsoft, which we are, so I guess that’s the the disclaimer part here.

And I actually am a shareholder at Microsoft. Okay. So I’m a shareholder at Microsoft. But data fabrics are a concept that it that pre existed anything at Microsoft doing.

Microsoft just happened to release something in in in the product this year called the data fabric, and I would I would I would argue to v one a data fabric. When I talk about data fabrics, I talk about, like, the the future state, the big thing. And I’m a total believer. And I was at Gartner.

I was a more of a contrarian at Gartner, but I saw the problem.

I think we’re gonna hear more and more and more about Data Fabrics in twenty twenty four.

Other things. Blockchain. Just evaporated. Just went away.

Some people went to jail. Some people fled to other jurisdictions.

The crypto market’s completely totally crashed aka crypto winter.

Yes, block chain and crypto are are when is it enabling technology and another is a is a use case on on blockchain. Crypto is a use case on blockchain FYI.

But, yeah, it it largely went away, but sneak peak twenty twenty four get ready for it again. I think we’re gonna see ETFs, spot ETFs change traded funds released, and approved by the SEC, early in twenty twenty four, and that will cause an explosion in crypto prices, which will cause a reinvigoration around the whole idea of blockchain, which I actually think holds a lot of promise.

Frankly, I think the I think blockchain is an unnecessary. If you really see the benefit of data mesh, like if you actually see the benefit of peer to here, data governance and data sharing, data exchanges, widespread data ecosystem data sharing internally, externally. If you see those things as valuable, which I do.

I think blockchain necessarily has to play a role there. Maybe I’ll talk about that in a podcast coming up, but I’m a believer in blockchain absolutely positively, because I do think that it is a wonderful wonderful solution for the problem of how do we enable peer to peer sharing at scale.

We’ve always tried. We’ve always wanted to Right? I think data marketplaces in to a certain degree are a reaction to sharing, but it’s old school. It’s web two dot o stuff. Meaning centralized centralized infrastructures to do widespread sharing. That in and of itself is a little bit of a boxy moron. Is it not?

Centralization to share.

But, I mean, that’s kinda how things have historically been. Right? We centralize things, put things into one bucket, and then everybody can go drop in the bucket. Well, if you believe that there’s a better way, I believe there’s a better way.

I think peer to peer, Napster proved it. Good grief. Napster. Yeah. That’s how old I am.

Nastro proved that peer to peer can be done and can be valuable and can be done at scale.

And I think that need, whether it is sending currencies worldwide, whether it is speeding up the international financial systems, the the the monetary systems, the what, I mean, the list of use cases here is long. Good grief.

I think I think blockchain, well, we just can’t ignore it and we won’t ignore it in twenty twenty four. It’ll come raging back.

Other kind of runner ups?

It really kind of that’s it. I mean, I still I still see people struggling with how to value data, and I’m still getting pushback. I’ll I’ll post things on LinkedIn about, hey, data leaders, we really need to figure out how to value data, and it’s still getting the pushback of, well, it’s an intangible, and it’s really hard. And we can’t do that. And it’s it’s not an asset according to the CFO, and We can’t can’t can’t can’t. That’s part of the old mindset.

That’s part of the old mindset. The I can’t value data and I don’t need to value data and I don’t want to value data. Old mindset stuff.

If our sole purpose is to deliver value and we don’t measure it, What does that say?

Well, it says it’s it says it says a lot of different things. One, it says, I I I would would argue it doesn’t bode well for a long term tenure, right, if the very thing that you exist to do is not being measured, that’s a problem. I mean, I’ve got measurements as a part of what I deliver day in and day out for my job, most everybody else does.

So if you don’t get measure it, right, if you can’t value it, your sole purpose is to deliver value, I see that as a problem. And I think that there are a better way, and I think we can value data. That is most certainly something I’m going to be talking about twenty twenty four.

So let’s wrap up. I think twenty twenty three for a lot of people was a difficult year.

I think twenty twenty three for more people, thankfully, was a pretty good year. I know for me it was just for me.

To the degree that anyone cares, I had a pretty good year in twenty twenty three. We got a couple of new, new fur babies in the house, a couple of little dogs that we, we picked up at our local Revard County dog pound.

They’re just awesome. They they enrich our lives. We we we love them to death. They’re fantastic. We’re dog people. Love those two.

My job is amazing. My company is amazing. Doing this is just so wonderful and it’s so amazing, interacting with people, meeting people, talking with people, It’s just it’s twenty twenty three. I I will remember it very, very fondly and frankly without sounding overly glib.

A big reason why I will remember it fondly is because of you.

Maybe you’re listening to the podcast. Maybe you’re giving me a comment on LinkedIn. Maybe you’re coming to hear me speak at a conference. Maybe you’re coming up to my booth at a conference and having a conversation. Maybe you’re sending me an email. Maybe you actually read my monthly newsletter. The CDO matters round up.

And maybe you provide comments to it. It keeps me going.

It it it keeps me going. I’m my sole purpose here is to help chief data officers extend their tenure to make data and analytics leaders more successful in what they do, and I only able to do that because of what you provide me and the trust that you give me. So, I don’t take that lightly.

Thank you. Happy holidays.

Happy new year. I will look forward to seeing everybody very soon in 2024. Until then, Bye for now.