Tim Gasper [00:00:01] Hello, welcome to Catalog& Cocktails. It's time for our season 6 finale and we're coming to you live from Austin HQ of data. world. I'm Tim Gasper, longtime data nerd, product guy over at data. world, joined by Juan Sequeda.
Juan Sequeda [00:00:16] Hey, Tim. We're here figuring some stuff out right now with, our LinkedIn is being funky. This is actually the first time in 160 episodes, I don't know that something's going wrong.
Tim Gasper [00:00:28] Yeah. Usually, LinkedIn is a reliable connection.
Juan Sequeda [00:00:30] Yeah. So, I'm telling people, " LinkedIn is being weird, join us on YouTube if you want to hear it."
Tim Gasper [00:00:36] Find us on Twitter and YouTube.
Juan Sequeda [00:00:40] Yeah. Anyways, but always the show must go on here.
Tim Gasper [00:00:50] How's that one? There you go.
Juan Sequeda [00:00:52] Anyways, this is weird. Anyways, okay. So, what are we doing?
Tim Gasper [00:00:56] It's been quite the season.
Juan Sequeda [00:00:57] It has been so much.
Tim Gasper [00:00:58] We've had some amazing guests across a lot of different topics and so we'll be talking through some of the big takeaways today, but also, we drank a lot of great cocktails.
Juan Sequeda [00:01:09] So, I'm really excited. I'm going to shout out for a friend who you met recently here in Austin, Omid Hi. We were just having drinks on a Friday. We'd go to my favorite brewery, which is next door to my house and he was talking about he makes this Earl Grey Tea infused gin and he gave me a bottle of it and this is really damn good and I'm kind of mixed up, sort of an old- fashioned type of thing, agave syrup with some bitters on this.
Tim Gasper [00:01:39] Cheers.
Juan Sequeda [00:01:40] Cheers. This is a very strong potent drink for a very long year that we've had. It has been crazy in the last six months.
Tim Gasper [00:01:48] It's been crazy and this is a great cocktail. You get a really strong tea flavor to it and a sweetness too.
Juan Sequeda [00:01:59] I was saying that this is a bit like a Long Island iced tea type of drink. I don't know if that's a good thing or a bad thing for the folks.
Tim Gasper [00:02:04] Yeah, I think it's a bad thing. Juan was like, " Tim, I'm going to make you a cocktail. It's kind of like a fancy Long Island iced tea."
Juan Sequeda [00:02:12] No, but...
Tim Gasper [00:02:13] Wah!
Juan Sequeda [00:02:15] I'm enjoying the, it's strong.
Tim Gasper [00:02:16] No, it's really good.
Juan Sequeda [00:02:17] It's strong. All right, so much to go do. So much to talk about. Kick us off.
Tim Gasper [00:02:22] Yeah. So, we're going to talk about four main kind of categories of topics today that have been themes across season six. So, there was a business value, kind of getting business value out of your data and connecting to the business. There was data fluency and culture, which we'll unpack a little bit more, technical aspects around data, and then AI, everything's about AI, right?
Juan Sequeda [00:02:47] Yep.
Tim Gasper [00:02:47] And so, we're going to keep you on the edges of your seats and we're going to talk about AI last. So, let's start off with our favorite, Business Value, because what's the point of everything that we're doing if it's not to try to create value for the business as data teams, as data organizations. And let's first talk about Ethan Aaron who is the CEO of Portable. He came onto our show by the way, Portable, super interesting company around data integration. And he also has really great LinkedIn posts. And I think Juan, you first kind of ran into him at some different events and things like that, but you also saw his post and you're like, " Oh, my God." You were talking with him and you're like, " We got to talk about that stuff, but talk about it on the show." And so, he came on. And one of the biggest things that we were talking about with him was especially around the lack of business value by data teams and how there's so many things that we're focused on that are not the right things there and we really got to change the way we focus and we prioritize our work. And I think one of the biggest things that we talked about on the show with him that was super valuable was sort of like this quadrant and he's got a really great LinkedIn post on it too. But that quadrant where you've got on one axis, you've got value, sort of low value on one side and high value on the other. And on the other, you've got effort, low effort, high effort. And where you want to focus is obviously high value, low effort work. That's the low hanging fruit.
Juan Sequeda [00:04:19] Kind of seems obvious but sometimes it isn't.
Tim Gasper [00:04:21] But sometimes it isn't. And some of our data teams, some of you are doing a good job of this, some of us are no. And some of us think we're doing a good job of it, but we're not. So, let's be honest with ourselves there and really ask our stakeholders what's valuable for them. And then, there's this other quadrant two, which is high effort, high value. It's hard, right? And so, you really got to be careful about the investments and the choices that you're making there. Let's say the data team wanted to take on finance. Finance is pretty important, but you might not want your data team to be focused on that right now. It's very hard. And so, it's better for the finance team to focus on it. So, a lot of people wonder what is the main point of the data team? What's the right way to think about how the data team should be focused on? And he had a great set of four things that he said. He said, first of all, analytics going to the executive and they might say, " Hey, what's the top sales or what's the top things for this particular person and how it impacts the business?" That's important. Automation of tasks, number two. Product, number three. And then, number four, risk mitigation. And so, I think that was really helpful. And then, finally, the smartest people in data can identify two main things. One is the levers that drive revenue and the second is the levers that reduce cost. So, think about the different data projects that you're working on, which ones are more around revenue and which ones are more around cost. And if you feel like it's not having a big impact on either, then it's probably stuff that is low value. The next person that I want to bring up that we talked to was Aaron Wilkerson. And Aaron Wilkerson is the Senior Manager of Data Strategy and Governance at Carhartt. And the episode that we had together was focused around going from the technical and technical strategy to business data strategy. And we talked a lot about how a lot of data teams are focused more on a technical strategy of we need to adopt these technologies and do the stack in this way. And it's very technical oriented versus really building out your roadmap and thinking in terms of the business. And he asked a really important question, as a data leader, do you have too technical a data strategy and therefore finding yourself having trouble connecting with the business? And he really said that a business data strategy is, so the salespeople may be talking about customers and how to get to them and doing customer journey mapping and these are all things that are part of the business strategy and a business data strategy is how is data going to help us get there. So, think about your own corporate strategy, think about the initiatives that are going on within your own organization. What are the things that are the most important? Those are the things that you should be aligning your data roadmap to. People need to talk about data the way that business talks about business in terms of the business.
Juan Sequeda [00:07:26] Man, a lot of these things are kind of seem obvious or common sense, but we are talking about it so it isn't, and hopefully, I mean this has been the year of business value, I think.
Tim Gasper [00:07:38] Yes. I think that's been a big theme. I think even when we entered this year, it was like ROI and then business value. One more thing that Aaron talked about in his episode, he talked about making sure you get a seat at the table and get buy in. And so, he said, you've got to figure out as a data leader how to become a trusted partner. You're being asked to be in the meeting because a critical decision needs to be made. If you're in the room and you're not providing value, you're going to get kicked out of the room. So, don't just spend the whole time talking about servers or other things that people don't care about. Why streaming is the bomb. How do you know this is working? There is word of mouth that starts to happen. So, if you have this, " Wow, our data team is so impactful. I love working with that team." You know you're having an impact. And there's a difference between people coming to you with, " I have the solution and I just need you to do it versus I have an outcome I desire and I need your partnership to solve it."
Juan Sequeda [00:08:34] That's a very important one because it talks about the difference between being operational, being strategic, and I mean we need to be operational, but also if you're only operational then you're just being like, " Yeah, you just do things." Instead of like, " No, we need you to be part of this conversation. We cannot go forward and figure out how to make a decision to be the best in our company without you." That's the position you want to go be into. And I think one of the things that Aaron highlighted was another comment from another guest of ours. Joe Reese was like, " We've lived in this lost decade." We jumped into the era of big data and we focused on the technology for so much time and people made a lot of money and stuff around that. Salespeople made a bunch of money selling tech and selling services, but the value didn't come. We just lost and we got used to that stuff and that's it. Game over.
Tim Gasper [00:09:21] Millions and millions of dollars were spent on Hadoop infrastructure. Where has it gone? Where has it gotten us? Right? So, great session there with Aaron. One last thing with Aaron's session, I thought that was some good, honest no BSS that we hit at the end of our session is that we were talking about CDOs and how CDOs only on average will spend 18 to 24 months of their tenure. And came to the point where we're like, well, maybe we're hiring the wrong people in CDO.
Juan Sequeda [00:09:47] I think this is a trend that we're seeing is no more CDOs who will only have a technical background. Either CDOs are going to have business backgrounds or they're going to be technical people who can bridge that gap. But if you are a CDO that you only talk tech and come from the tech background, I guess you've got to start packing your back pack.
Tim Gasper [00:10:05] You might be in trouble. You need to realize learning the business.
Juan Sequeda [00:10:06] You need to realize that too. You need to realize this right now.
Tim Gasper [00:10:09] Yeah. Yeah. I think we're seeing that as a trend across a lot of the industry now as these CDOs that are coming in that may actually not have, they're not a deep expert in the technology, but they understand the business deeply.
Juan Sequeda [00:10:21] That's what's important. You can get your collaborators, your deputies who will know the tech for you.
Tim Gasper [00:10:28] Yep. Yep. Exactly. So, next episode, Alexa Westlake, Senior Data Analyst at Okta. Also, just in general, a great data evangelist. Connect with her. She is awesome. So much energy, so much excitement and ideas around data and in particular around how to drive outcomes over outputs as a data team. And so, outputs are leading indicator of outcomes. We really need to focus on outcomes. If we aren't understanding the outcomes, what's really the point? Data has to help with customer success, with marketing, with whatever it is that makes your business tick and drive those results. And you have to create a culture of joint ownership around success. So, data and business. The teams are working together. There's joint ownership over driving the results. She said that there are three categories of focus. One of them is around system improvement. This is kind of non- negotiable. You got to reduce load time, reduce costs. There're all these things that you have to drive which are core in order to make things more efficient. Process automation. If we're doing those outputs correctly, we should see this getting impacted, and then people experiences. So, you should see that you're driving better customer satisfaction, that you're decreasing churn. How is it impacting the end consumer or the people ultimately? So, I think that was very important.
Juan Sequeda [00:11:55] And those are the outcomes, right?
Tim Gasper [00:11:56] And those are the outcomes, yup, exactly. And we also talked a little bit about tools and processes and methods to be able to drive alignment around outcomes in the organization and create better partnership between data and the business. One of them was around OKRs and how it should be really important that the data team help different teams around developing and managing and tracking their OKRs rather than just be a downstream consumer of what happens there. And you really need to reverse engineer from the goal. It isn't rocket science to do that. The data team needs to be the best friend of automation metrics as well. Another tool we talked about was creating a council to measure scale. And that council can really help you to address the tech debt. It can help you to drive growth and innovation. If you're a small company, maybe you have to take on some of that tech debt and so there's an opportunity cost around that. But as you scale, you need to really address that. So, building that council creates a group of people that allows you to really look at the bigger picture and prioritize. Lastly, strong leaders. You need strong leaders in data and in the business, who understand data, look for empathy, look for self- awareness. Strong leaders are going to make you feel comfortable asking hard questions.
Juan Sequeda [00:13:25] That's another trend that we've seen, empathy, definitely empathy.
Tim Gasper [00:13:29] Yep. Empathy has been big. All right, last on business value. So, we had Chris Tabb who is a founder of LEIT Elite data over in the UK.
Juan Sequeda [00:13:42] He's on the mean data streets all over the world.
Tim Gasper [00:13:45] Hashtag mean data streets, bringing it to the mean data streets. And there we focused a lot around driving value from data, and in particular, sort of the honest no BS around business value from data.
Juan Sequeda [00:14:00] What is your definition of business value?
Tim Gasper [00:14:02] Right. And so, he started with the definition of business value and he had very specific words. I was surprised at how specific his words are. Evidencable, positive impact on your company performance. So, evidence driven, positive impact, and company performance.
Juan Sequeda [00:14:24] And by the way, it's not always profit, right? Because his example is, look at Uber. That wasn't their goal for that. That's not how they were measuring the value.
Tim Gasper [00:14:31] It's not just company profit. For Uber, it was all about growth. It was take over the market as fast as possible, getting every city.
Juan Sequeda [00:14:38] I mean the way you define impact and performance is different for every type of company.
Tim Gasper [00:14:42] Exactly. Yep. And another thing that he brought up that was interesting around business value is depending on what mode you're in, you're going to value different things. So, if you're in more of a fiscal mode, then business value is going to mean you're trying to save money on the budget. You're looking for more quick wins, low hanging fruit. If you're more in a growth mode as a company, then you don't necessarily have the luxury of waiting to see if things can organically grow and self- fund themselves. You got to get customers as fast as possible or you want to get market coverage as fast as possible. Maybe you want to develop a competitive advantage as fast as possible. That's what business value means to you. Profit might actually be a bad thing to optimize for. You intentionally don't want to optimize for profit, right? And then, finally, there's also a hybrid mode where maybe you're trying to extract some of the gains from the cost savings or some of the things that you're doing to then reinvest in innovation. And if you're in this mode, you need to create incremental value, you need to have a good story every quarter so it becomes very time bound and you really need to keep your stakeholders close on the journey because if they start to worry that it's not making progress, then it could start to switch back into fiscal mode. So, I think these are some good things. Think about your own organization, what mode are you in?
Juan Sequeda [00:15:58] What mode are you in?
Tim Gasper [00:15:58] And maybe, even depending on the initiative, you might have different things going on at the company, different modes going for different initiatives. Last thing for data products, we talked a little bit about data products and he talked about the wrap, sell, improve model, which I think was kind of fun. Which is thinking about, when you're trying to create value out of something, the processes is that you need to think about the packaging of it. You need to really focus on how you're going to get people to want it and then constantly be iterating and improving on it, right? And so, in order to do that, thinking about data products and thinking in that way, that's going to allow you to think about it things more from the business perspective. Take a product, make it stickier, make it more user- friendly. So, data products I think is another big theme that came up a lot this year. Maybe last year it was more about data mesh. This year, the specific part around data products and then things like data contracts. That has been much more of a focus.
Juan Sequeda [00:17:00] Yeah. And one of the things that I really liked about Chris's episode is that we started talking about this analogy with the legal space, right? He's like, oh, they have to think about it like making the case, just like you make a case in the legal scenario, you have a small case, right? You got to win over the CRO. The CMO, right? That's how then you put the big case forward and that big case, the judge is the CEO, and then the jury right there is all the C- suites. That was a great analogy of the things.
Tim Gasper [00:17:24] A good analogy, yeah.
Juan Sequeda [00:17:25] I think later in the episode we go like, well, is everything supposed to be legal? Then I think we turned into the EV model and electric vehicle.
Tim Gasper [00:17:32] For those of you that are listening that have listened to episodes where we go real deep on an analogy.
Juan Sequeda [00:17:35] Real deep into an analogy. That was one inaudible.
Tim Gasper [00:17:38] This is a good one. Yeah. Check out the episode with Chris. Also, I mean business value came up all over the place, but specifically culture indeed of fluency came up. I know fluency is a topic that's very near and dear to your heart too.
Juan Sequeda [00:17:51] It's not inaudible because it's not literacy, because what I'm glad is that many people call BS on the word literacy because it's disrespectful. Several things. One, we had a conversation with Wendy Turner Williams. She was a former CDO at Tableau and many others. She was at Microsoft for a long time. Her stats, 92% of businesses fail to scale data analytics and 95% of that 92% basically almost everybody blamed culture. And if you can't scale data in this new world of AI that we're in, how the heck are we going to go scale AI? Now, the reality is that most people don't even understand what data culture is. So how about high- quality business decision data that aligns the strategy that the business is focused on and enabling people around that strategy. So, for example, if you hear like, " Oh, here's a lead." Well, in a business, how does that travel through the business? Let's understand that flow, that is critical to the data culture and that's literally the knowledge around the business. Data culture is really business knowledge and literacy is not just about, oh, I'm going to learn SQL. It's really about learning the business, the business strategies and knowing how to ask those right questions. So, part of also in thinking about the organization structure, there's no one size fits all, right? So, you think about where does the CDAO or the CDO report to or the CIO and so forth. The CDO should be the CEO's best friend because the CEO is the number one customer for data. But what's interesting too is the highlights is like, look, software companies, especially big tech companies, they don't have CDOs. I know a lot of conversations, people are like, the goal here is to kind of get to a stage in the company that you don't need CDOs anymore because the data is already so over the place. That's an interesting point about it. Be aware that data is political. I think we forgot to start talking about our T- shirts. I think that's another T- shirt phrase right there. Data is political. So, we're talking about how do we turn data to a first- class, data maturity frameworks, things like the DCAM. Be able to understand these models, these maturity models, educate yourself, understand the internal org structure, network. Who do you influence? Who gets it, who doesn't? You want to be connected to those people. What are the values of your organization and how does data help to enable those values? How can data help others? Find the fans, build a community internally, make data intuitive, shift they've left. And at the end, another T- shirt is we're shifting from data talkers to data walkers.
Tim Gasper [00:20:24] I love that.
Juan Sequeda [00:20:27] Another conversation we had was with Simone Steele. She gave a fantastic talk, I'll never forget at the CDOIQ conference and I invited her saying, " Please, can you give that talk here to everybody here on the podcast?" So, we talk about data sustainability. And sustainability isn't just about the natural resources, but it's also about those social aspects. So, are we training professionals appropriately? Do we have enough of them? Is it sustainable to manage the business with all the legacy debt that we have and all the flaws that we have? Is it sustainable to keep consuming more? So, she has this great diagram that she went over. She says that there's these two things that are happening. One, there's a development of tech and all the potential benefit it brings, right? So, cloud computing, generative AI and LLMs, right? These are all big inflection points and then there's a second curve which is also improving, but at much lower rate and this is the actual and the real benefit that's coming from that technology. But we also know that there's regulations and things that will start to flatten out these curves. The issue here is that things are not being designed correctly in corporations that we can keep up. The real value is not keeping up with the potential benefits and that's that sustainability gap and her position is that this is going to be the focus of the next generation of leaders. But this is going to take time because following Planck's principle, science progresses one funeral at a time. So right now, everyone is so very short- term focused. This financial quarter, you just have to wait and this is probably going to be wild. This goes back to finding the balance between efficiency and resilience.
Tim Gasper [00:22:07] That was a very interesting conversation with her because I kind of ended it feeling like we're in this for life. This is an ongoing journey and obviously we're making a lot of progress here, but it made me really appreciate the long view on what we're trying to do around data and sustainability.
Juan Sequeda [00:22:29] And I wonder how much if we're going to see it in our life, I mean all the stuff that we complain and rant about, when will we see in the next 10 years, 20 years, 30 years? I mean the honest no BS is going to take a couple of decades and I wonder if I'll be retired or not.
Tim Gasper [00:22:37] I feel like you keep on mentioning the Einstein quote, right?
Juan Sequeda [00:22:40] "Keep doing the same thing over and over again and expecting different results." Definition of insanity. We're driving ourselves insane here. All right, so another conversation we had was with Doris Lee who's the founder of Ponder and now at Snowflake. So, talking about data science and how that has been evolving, what are data scientists today? I think, for me, the definition, the summary of that podcast episode was the blurry lines. Everything is getting blurred from people from roles to tools. So, the lines are getting blurred from business analysts and data scientists to machine learning engineers. I mean, now, banks, spreadsheet users, quants or they're learning Python and they're doing things for time series, forecasting. Everything's kind of blurred right there. Her point is like everyone's going to be a data scientist no matter your role or title. If you think about the three categories, the data science capabilities go into low- code, no- code tools, auto ML tools, and then Python and notebooks and even there's lines being blurred across those different categories that serve different personas. Python versus SQL, the lines are getting blurred there too because you can use Python to query a database, you can use SQL to call script functions, right? At the end, this really shouldn't matter. The features are, there's APIs that are agnostic to what the data's backend is and kind of dot, dot, dot going forward. LLMs are the APIs for humans. That's another good quote from Mike Dillinger, I think a T- shirt quote that we'll talk about. And then finally, I like the open source landscape around this, right? One, I need a load in my data and transform it. That's what pandas are for. I need a compute statistics or machine learning model. We have scikit- learn, all that stuff. And then third, you visualize things. And at the end, you have IEDs and development environments like Jupiter, notebooks and so forth.
Tim Gasper [00:24:30] One big takeaway that I like that of Doris's chat is that the idea of a data scientist is becoming very flexible and it overall seems like a good thing because it means that it's a big tent that a lot of people can come into. It's no longer this idea that you're a big data machine learning type person. They can come in many different flavors.
Juan Sequeda [00:24:49] Agreed, agreed. So then, we had another conversation with Krystin Kim who is at Post Holdings, and this was all about the power of collaboration.
Tim Gasper [00:24:59] This was full of T- shirt quotes. I got that one.
Juan Sequeda [00:25:02] Right. So, one of the things is the power of collaboration, it's really the undervalued value that matters. So, her point is, if you're implementing a data catalog, you end up with a treasure trove of just so many different use cases that you can stumble ideas, you can amplify other people's ideas. There's just so much stuff that you can go do that serendipity, that's hard to put the value on, but that is actually the true value. So, when people talk about value, her quote is the what, the why, and the wow. And her big rant there is like you just go sideways when you start talking about the how. Oh, people get into the canonical models, the taxonomies, the lineage. She's like, no one really gives a fuck about that stuff. At the end of the day, it's a how. Give me the what, the why, and the how.
Tim Gasper [00:25:51] The wow.
Juan Sequeda [00:25:53] The wow, the wow, I'm sorry.
Tim Gasper [00:25:55] Not the how.
Juan Sequeda [00:25:56] Yeah. Later on, she says back to the what, the why and the how. And then you can talk about the how. She says, we need to become storytellers, make it relatable, have analogies. Her analogies, like data's an ingredient. The catalog is a grocery store. You get a piece of data ingredient, you mix it up, you create a forecast. The catalog is like your cookbook and your recipes live in there. And another one, she's like, " You have to be like Paris Hilton. I don't get out of bed for less than a million dollars." So, focus on the stuff that makes a difference. That is the wow.
Tim Gasper [00:26:24] That connects back to Ethan and some of the business value stuff.
Juan Sequeda [00:26:28] Exactly. I mean this is all getting, this is all connected here. Another good point to think about the Levi Strauss model. He didn't go to the mine for gold himself. He put the tools and the genes and hands of people that wanted to go mine the gold. We got to influence the influencers in that way, right?
Tim Gasper [00:26:46] Yeah.
Juan Sequeda [00:26:46] That's how we want to go find ways to scale. Another good quote, " Don't be the data scientist that gets lost walking in their random forest." " You either win or you learn." Another great quote from Krystin and what grownups call change management, right? It's a paying attention and it's all about the people part. " Fail fast but not too big." And yeah, " Project management is hard. It's an art and a skill. Don't overdo it."
John [00:27:14] Yeah. Krystin was a great guest at a great episode. She has very professional irreverence.
Juan Sequeda [00:27:22] Yeah, that episode we were in Orlando, so I was outside at the pool and you were in London, I think.
John [00:27:31] Yeah, I think I was London. You guys were in Vegas. Yeah. What a weird thing. That was awesome.
Juan Sequeda [00:27:35] And then, another thing that's coming up, it's data storytelling. We're super excited. We had the chance to chat with Kat Greenberg who just released her book on data storytelling. What is it? How do we effectively communicate our data insights? And this actually starts becoming an issue as organizations start to mature. Why? Because you have so many dashboards because they're so easy to create. So, you really need to understand your audience. So, if you're presenting the dashboard to folks who have no idea, they need a story. You're presenting to an expert, they probably don't need a story. And there are three reasons why you need to create data visuals to discover, to inform and to educate. But be careful about another T- shirt quote, " The chart vomit." What are antipatterns of data storytelling? Thinking that a dashboard is a data story. Really the data story is something you should be able to go write. You need to be able to write it down, understand it before you can communicate it. And then, telling the story is really focusing on the impact. You tell a story so you can have an impact on the business. So, what are the products that they're working on? What are the goals? How are you going to impact that goal? And another great quote that I love, " What is the goal above your goal." And something that we love, " How do you tell a good story?" The three acts structure from Greek theaters, the and, the but, and the therefore. So, like, and, and, and, but therefore you have to do something, right?
Tim Gasper [00:29:01] So, if you're doing and, and, and, and that's chart vomit.
Juan Sequeda [00:29:04] That's chart vomit, right? People get struck because yeah, it's an agreement. Yeah, she went and, and, and. But is that contrast that gets me into that next play, that next act I need to go do, therefore I need to go do this. So, I think these were all so much different aspects about, we talk about data storytelling, right? The collaboration, how the people, the roles are all being blurred. Think about sustainability and in just about how to turn data as your first class. And these are all the aspects of culture that organizations start thinking about. So again, it's people, process, technology, and this is all about the people in the process aspect.
Tim Gasper [00:29:40] Yeah, I especially enjoyed the end of that episode when we did a lot of the and, but, therefore experimentation. And it makes you realize that a story is embedded in almost everything. And what we're trying to do in navigating these data politics to connect the effect to that is we're trying to tell a story that people believe and want to and see creates value. So, even think about in your own work, we had the best sales quarter ever and our marketing team is awesome, but the pipeline is not looking very good, therefore we need to make sure we focus on marketing, right? Everything is a story and this data storytelling is what's going to make you really successful. All right.
Juan Sequeda [00:30:25] Now it's time into some more tech stuff.
Tim Gasper [00:30:27] Now, let's get nerdy.
Juan Sequeda [00:30:28] I mean, this has still been nerdy.
Tim Gasper [00:30:31] Well, that was business nerdy. Now we're going to get, I don't know, what is this kind of nerdy? Tech nerdy?
Juan Sequeda [00:30:37] I guess.
Tim Gasper [00:30:39] All right, so technical aspects around data. So first, Joe Reis. Is it Reis, Rise?
Juan Sequeda [00:30:46] Reis.
Tim Gasper [00:30:47] Reis. Joe Reis. I think so. Right, Joe? Joe, correct us if we're wrong. We've been saying it wrong too many times then. He is the CEO of Ternary Data, instructor @ deeplearning. ai and also wrote the book, The Fundamentals of Data Engineering. Great book, make sure you check it out. It's more than just the nerding out. It's also best practices and really thinking about approaching data engineering in a smart way. And we focused our talk with him on Catalog& Cocktails around data modeling, sort of lost art around data modeling and how important it really is. And he highlighted that the art and the philosophy behind data modeling has been lost and he stressed the need to really reconnect with these fundamentals in order to not just have a better data foundation in general and have cleaner data, quality data, but just to realize the AI potential.
Juan Sequeda [00:31:39] This is the AI ready data world, man. I mean, if you're not focusing on the semantics, on the modeling your data, you're going to live in crap and that stuff is not sustainable. You're going to fail.
Tim Gasper [00:31:48] When I first met Juan, so Juan's company, Capsenta was acquired by data. world. You had a couple of catchphrases and one of them was the beautiful data and the opposite of beautiful data was the...
Juan Sequeda [00:32:02] Shitty data.
Tim Gasper [00:32:02] The shitty data. The data that was the friendlier version or not friendly version. The inscrutable data was inscrutable, right? Inscrutable. Inscrutable. And I always think about that word inscrutable and we want AI to come in and understand our enterprises and understand our lives and just become this magic pixie dust. But honestly, the metadata and the data that we have available for it to use is inscrutable. It is not beautiful. It is not modeled. And if we really want to capitalize on AI's potential, it has to become beautiful data. It has to be AI ready. So, we're so used to seeing modeling as an exercise of capturing ad hoc queries and responding to ad hoc requests and not focusing on traditional conceptual modeling and then logical modeling and then physical modeling. And it's not to say we have to swing the pendulum all the way back to let's spend two years arguing in front of a chalkboard or a whiteboard about what's the architecture going to be and what's going to be the snowflake edges versus the hubs. And we have to make sure that we do a balance and that we stay agile. But we've become so focused on the tactical. What is the risk of not thinking enough of modeling? Well, you're going to do dumb things more quickly and you won't know that you're failing?
Juan Sequeda [00:33:20] And you'll just continue doing more dumb things and dumb things keep just driving yourself insane.
Tim Gasper [00:33:25] Yep, exactly. So, there's not one single way to do modeling I think was another great takeaway from Joe, right? There can be these religious wars about should you do Inmon, should you do Kimball? Should you do this? So, don't get caught up in the religion around that. They're good for different use cases. Each one was optimized with something in mind. So, think about how are you using the data, pick the right mode for you and perhaps it's even a little bit multimodal. The analogy is think mixed martial arts, right? He said one martial art is not always the winner, it's the ones who know how to take the right skills from different parts of martial arts and combine them together. Those are the ones who tend to win the martial art competitions.
Juan Sequeda [00:34:08] A hundred percent. This is something like I go through all the time and they're like, " Oh, we should do it this way." And then, I catch myself as being pedantic like, " Hold on, wait, wait. We need to understand the use cases around this stuff and let's be very explicit about it." So, this is a really important takeaway.
Tim Gasper [00:34:23] Think about the trade- offs related to modeling, like time versus efficiency versus money versus quality versus rigor. Get an answer out the door as fast as possible versus being more rigorous. Do the right thing but don't over index on just doing the fast thing. In data, you can mask over debt with another query.
Juan Sequeda [00:34:41] And that's I think one of the issues. I mean, we're seeing this more and more where you have a bunch of models that you're creating modeling and then it's just a query over a query over a query and then it's just making a big pile of inaudible.
Tim Gasper [00:34:59] Yup. All right. Next, data contracts.
Juan Sequeda [00:35:03] That was a recent one with Andrew Jones.
Tim Gasper [00:35:06] Yeah, so Andrew Jones, the guy who wrote the book on data contracts, go check out his book, go check his posts out. Very interesting conversation. We started off with what is a data contract? It's an agreement between those who produce data and those who consume it. Simple, clear. There are then details of course to how that's actually communicated and enforced. It could be service level objectives, SLOs, it could be service level availability, SLAs, it could be owners. It lets you manage responsibilities around data and have expectations and dependability around it. It helps you to understand it. So, data contracts are probably things that already exist in your organization. It's just a lot of it is implicit. It is not explicit. It is kind of unknown. It is a source of distrust and complication. We should get a little bit more concrete about these things. A data contract needs to be something that you can depend on. Anything that describes the data and sets expectations around the data is not necessarily a contract. I think is one of the things that we really came away with this from. So, for example, a schema, a data schema. It has some contractual aspects to it, lowercase C, but it is not a data contract in the way that we think about data contracts. And that's because it alone does not guarantee anything because you can change the schema. But if somebody says, " We will never change the schema except with 30- day notice and this is the way we're going to communicate about it." Okay, now you've created a data contract around it. What does it mean? He talked about shifting left around responsibility. What does it mean to shift left around responsibility? What actually reminds me of an episode we had a while back with the gentleman from Indeed, one of the leaders that was there, right? Or was it Intuit?
Juan Sequeda [00:36:58] Intuit.
Tim Gasper [00:36:58] Yeah. And we talked about shifting left and there's a little bit of a software engineering mentality here, which I think was an interesting conclusion that we kind of came to at the end when we were thinking, " Okay, interesting." Data contracts have slightly different meaning in the context of tech organizations, software organizations where it seems to have become more predominant. And now, other industries are trying to figure out how they take the best ideas around data contracts, which might change the shape of it a little bit as you talk more just about a contract on a warehouse. When Andrew was talking about contracts, he was talking about when you have software that is serving customers, it is creating data. And the people who decide what data is captured is the engineer who wrote the code for that software.
Juan Sequeda [00:37:48] This is something that when you're having the conversation about things like data contracts is like you really need to understand the context of where those people are coming from, right?
Tim Gasper [00:37:59] Yeah.
Juan Sequeda [00:37:59] So, if you are actually in control and writing the software, you're a tech company, the contracts for you is different than, " No, I just consume data that comes from Salesforce so I can't go tell, there's things that I can't control." So that's something that we need inaudible
Tim Gasper [00:38:13] You're an insurance company and you've got all these files that are coming in from different places and they're messy and you have to massage them, but you can't tell them to change the file format.
Juan Sequeda [00:38:22] This is probably one of the reasons why you see that big tech companies don't have really CDOs because it's like the data function, the work of data is really embedded so much in just the culture of how engineering is being done.
Tim Gasper [00:38:33] Yeah, it's the product and engineering organization that owns the creation of the data.
Juan Sequeda [00:38:37] Exactly.
Tim Gasper [00:38:38] Whereas, most companies have to deal with the pain that comes with the data, how it comes. And so, there's a different problem there. So, super interesting. Well, so what are not data contracts? Well, not data contracts is where you don't shift left far enough. And also, it's where it doesn't improve the data quality. Where do data contracts live? Well, ideally the truth is code. Data contract as code, which is again a little bit of a software centric view of the world. But even for the rest of us, the code, can we encode it in something, right? Put it in GitHub and how do you make it discoverable? Put it in your catalog. Make those data contracts discoverable via your catalog and connect it and make it related. How do you incentivize people to adopt data contracts? Every organization has some kind of a goal that they're trying to accomplish. Connect data contracts and providing higher quality data to that goal. We can't achieve that goal unless we have high quality data. And the only way we can have high quality data is if we have data contracts. That's your argument. Data contract doesn't have to be complicated. His book is a 15- page chapter focused on the implementation of data contracts and he really wants people to know this, like, " Don't overengineer it. Don't get overcomplicated. It can be simple."
Juan Sequeda [00:39:55] It's a T- shirt right there. " Data contracts are not that complicated. It's a short chapter in my book."
Tim Gasper [00:40:00] Yeah. It's only 15 pages. That T-shirt's going to have a lot of words on it, but it's still a great T- shirt. You should buy it. All right, data quality. Tom Redmond, the data doc came and talked with us and he is a guru around data quality and his first and biggest observation was that the vast majority of data management is by people without data in their titles.
Juan Sequeda [00:40:27] I love that.
Tim Gasper [00:40:29] So, it's marketing, it's ops, it's finance. This is actually a fun contrast to the Andrew conversation where there was a little bit more of a software angle to it So, what Tom is basically saying is, well Andrew, he's dealing with the engineers that are creating the data. For most of us who's creating the data? It's the marketing department that's creating the data. It's the ops department, it's the finance department, it's the sales department, customer success and customer service. They're the ones creating the data.
Juan Sequeda [00:40:57] They're creating and they're the ones who are doing interpretation, doing inaudible.
Tim Gasper [00:40:59] They understand the semantics of the data.
Juan Sequeda [00:41:01] And then, they're taking the data from some other department, which they don't know how it does. They're the ones who actually all those spreadsheets that we can plan, they're being shifted around because of all the people who don't even have title data, don't have the word data in their title, they're the ones who's doing that data work.
Tim Gasper [00:41:15] And one of the things that's driving us crazy, and I think Tom really unearthed a golden truth here, is that we're driving ourselves crazier around data quality. And it's because there's a control problem here. These people, marketing people, ops people, finance, they're doing the work of confirming the data, but they're not trained in data quality or data management. They're just trying to do their job day after day. And so, the second observation is that a data scientist cannot properly understand a business problem that they're working on unless they understand what's going on with the regular people, with these people that are just trying to do their job and make decisions and run the business.
Juan Sequeda [00:41:55] The true data people who are doing the data work who don't have the word data in their title. That's Tom's work.
Tim Gasper [00:42:01] Exactly. So, if we're going to address data quality, it needs to be bigger than just something the data people take care of. It's a company- wide initiative and probably a cultural change.
Juan Sequeda [00:42:11] Yes.
Tim Gasper [00:42:12] Third observation from him. Technical debt is out of control and all of this cost money, money, money. So what needs to change? Well, everyone in the org has roles today, but they aren't really conscious of those roles. They're not trained in supported in those roles. So, 95% of the people that are doing data management are untrained without support. So, it's no surprise that one third of our time is being spent on these issues. So, this is the piece that we need to address. Bad data. Is that a thing?
Juan Sequeda [00:42:42] Yes.
Tim Gasper [00:42:43] But it can be nuanced in terms of is that an address I can send stuff to? If no, it's bad data. So, there's some interesting semantics and also some subjectivity to what quality is, but you can define it.
Juan Sequeda [00:42:59] And what I liked about that is that there's a spectrum of what's easy and hard. If you ask is this good or bad? And there's a clear answer that says, " Yeah, this is good or bad." Then yeah, that's an easy thing. And probably 80% is going to be like that, but the other 20% is like, " Oh, it's going to be so complicated." Then don't spend time on that philosophical part right now.
Tim Gasper [00:43:22] Well, and one thing that I thought was fun about this conversation is that, Juan and I have some go- to sayings. I know Juan, you especially say this a lot is that, " We don't even know what customer means. And so how are we supposed to do things if we don't know what customer means?" And I think you brought that up in the episode of like, well, what does customer mean? And he was like, " Wait, hold on. You're focusing on a hard problem. That's a hard data quality problem." If there's a hundred issues, 20 of them might be deep ontological issues, but 80 of them are actually low hanging fruit. So, solve those 80 low hanging fruit data quality problems first.
Juan Sequeda [00:43:57] That's so true.
Tim Gasper [00:43:59] So, I thought that was good. Then finally, he said, " It is absurd that today people have to talk about business value. It's the 2020s for God's sake."
Juan Sequeda [00:44:06] Oh, God. So true.
Tim Gasper [00:44:10] The future is here, but sometimes it doesn't feel that way. All right, last one on technical. Ari Kaplan, who is one of the data evangelists over at Databricks, came and talked to us about the data lake and also about data privacy and some different interesting examples around data. And his background actually before he was at Databricks was in sports analytics. Literally, he is the moneyball guy. He said it was remarkable, first of all on sort of the data lakehouse side, it's remarkable that not everyone has heard about lakehouse. Despite how us in the data industry, we feel like we're so immersive. Like data lakehouse, that's so 2017 or something like that. A lot of people haven't. And so, we have to remind ourselves to get out of our bubble.
Juan Sequeda [00:44:58] Oh yeah, that's an important one. The entire world knows what's Snowflake. I've had these conversations like, " Oh, Snowflake. What is that?" You what Snowflake is, and then I remind myself. " Yeah, Snowflake, Databricks, those things that we hear all the time." Yeah, everybody knows about that stuff.
Tim Gasper [00:45:13] There's a whole world out there, folks. It's not all just about the modern data stack. Which is funny because then when we talked about the lakehouse, he's like, " Well, the lakehouse is the modern data stack." And I was like, "Oh, we came full circle." But he's really talking about the lakehouse, this is the modern way that you want to do data management because it allows you to bring two different paradigms together, which we had sort of this online paradigm and a very transactional paradigm. The pendulum swung to Hadoop and more of the unstructured paradigm. The lakehouse was the marrying of these two things together where you can have the structured data warehouse and the data lake to bring together the structured and the unstructured. The best way to make predictions is when you use multimodal data, structured and unstructured. This is why lakehouse is important.
Juan Sequeda [00:45:58] That's what we're seeing right now in the AI world.
Tim Gasper [00:45:59] And that's what you're seeing with the AI world. People want to combine unstructured and structured data together in order to create better predictions and better automation. The lakehouse, you can have it in one place. It's also a single approach to governance, which is important. Also, knowledge is critical. The model may recommend things, but it doesn't know the context of your business.
Juan Sequeda [00:46:18] This is the big thing which is coming soon in our next section for the AI part.
Tim Gasper [00:46:23] Exactly. And then, we talked a little bit about sports analytics. If a player is injury prone, so tell them to throw the ball softer, but this doesn't make sense.
Juan Sequeda [00:46:37] That's why the context is so important.
Tim Gasper [00:46:39] The context is so important. The data itself is not going to tell you even if the context is critical. You need people to collaborate, paint the picture. And he said, " Be vulnerable. Go talk to people."
Juan Sequeda [00:46:48] Yeah. This is an episode which is, it's actually hard to summarize because it's just go listen to it because you go into the whole history of how he got into it and I don't want to spoil it to people.
Tim Gasper [00:47:00] We won't spoil it. Go listen to the episode. If you enjoy sports, you enjoy history or you just enjoy data, which should be all of you. Check out that episode if you haven't heard it. He tells some great stories about his moneyball past and experiences.
Juan Sequeda [00:47:12] All right, let's get into the AI part. Let's do it.
Tim Gasper [00:47:13] All right. So, I think AI was all over the place, but I think specifically we had some really great episodes. One with John Cook. I think this was a combination of business value, generative AI and data products altogether. And I loved how we asked him, " So, do you have a definition for data products or do you have a pedantic definition?" And I love how he says, " Yes I do and it doesn't matter." But the verb, managing data products, that is what matters. And that's the data product management. And then, so think about data product and generative AI and LLM, they come together. This is what I'm really excited right now. What we're seeing is this whole notion of AI ready data is like how do we build all these AI ready data products?
Juan Sequeda [00:47:56] One, we're going to see these co- pilots. I'm going to use LLMs to help me go create these data products themselves. Second, we're going to be using them for some sort of classification. And third, the LLM itself is going to become a data product. Your chat with your data itself is going to become that product right there. And what we really need to start thinking is how to use LLMs in every step of that data product management life cycle. And I think this is something that we're starting to go see right now. I know I'm working on this in my day to day. The stuff that we're doing in our lab is like, okay, here are the steps that we go to solve this problem. Where do I put LLMs? How do I put this and how do I measure actually the benefit from that? So that's what I'm really excited about. So, the question here is, as data professionals, how do we start taking advantage of all these generative AI innovations combining with data product management? So, his advice is start to push on things, test and see how much you can push on this and see how folks are reacting. So, you really need to be, start talking to your folks around you. Understand that very few people, very few companies actually nail software product management. So, set the expectations correctly because you're probably not going to, you will definitely not nail this in the first time. It's going to take a while until you figure it out.
Tim Gasper [00:49:09] So, give yourself a little grace.
Juan Sequeda [00:49:10] Yes, for sure. Anyway, setting the expectations is really, really important and you have to develop a language to speak to that business value. Again, in the end, we're all super excited about AI and all the cool things that we can do. We need to connect that all with the business value. And then finally, John, you have to go.
Tim Gasper [00:49:30] He was wearing a T- shirt.
Juan Sequeda [00:49:31] He's wearing a T-shirt, the great data race T- shirt, which is like you start first with the business problem. But if you start with the data first, it's a dead end and you're done. Game over. So, then from the business problem, you go to the wilderness of the data sources and you hit the roundabout of the data modeling and then you go through this whole area of contracts as you're testing and then you hit the production and then it continues to keep going to that duration. So, I really love that.
Tim Gasper [00:49:55] I love board games and I'm really hoping that somebody will try to create, it's like life except it is data life. It's this. It's the great data race.
Juan Sequeda [00:50:04] It's actually not a bad idea.
Tim Gasper [00:50:05] We'll play it one day, right?
Juan Sequeda [00:50:06] I'm sure that we'll have a market for that. I don't know how big, but I think there is a market.
Tim Gasper [00:50:09] I think, at least 10 people will buy it.
Juan Sequeda [00:50:12] Well, we have, for the Spotify stats, I think there's 112 people who have us as their number one.
Tim Gasper [00:50:20] That's true, 112 people, we are their number one podcast. So first of all, thank you all of you.
Juan Sequeda [00:50:27] This is amazing.
Tim Gasper [00:50:27] Some of you are probably listening right now, so we're talking to you. I see you, not really.
Juan Sequeda [00:50:31] Hopefully, buy the board game that we're going to sell inaudible.
Tim Gasper [00:50:33] Buy this board game. Yeah, we're going to stick it in our T- shirt store.
Juan Sequeda [00:50:38] In our imaginary T- shirt store. All right. One of the episodes that I really like was having Mike Dillinger, who was a former lead at eBay at LinkedIn working on knowledge graphs. So, we really discussed the relationship between LLMs about RAG retrieval- augmented generation, knowledge graphs and vector databases. One of the quotes for the T- shirt for our imaginary T- shirt store, " The LLM is the API for humans." And so, thinking about how do knowledge graphs and LLMs come together? It says knowledge graphs are great for reasoning and coherence. And LLMs, the large language models are great for language. Knowledge graphs, these are the semantic layer. And I think this is a conversation in another episode that we had with Dean Allemang going, what we're talking in a second is, yeah, knowledge graph semantic layers is basically the same thing. We have so much marketing going around. But so, vector databases, how does this come in? So first of all, vector datas are about focusing on the strings. You push strings in and you're trying to associate strings with strings. But this breaks down because you can say, " Hey, wine and cheese, those two things go together, but that doesn't mean that they're in the same food group." Doctors and nurses, those are related, but they're not synonyms to each other. So, this is when things go break down. So, there's some inherent limitations when you focus just on the strings. And this is where knowledge graphs come in because it provides that context that distinguishes all these different things. And the knowledge graphs are for bringing you the accuracy, which is critical for any type of enterprise applications you want to go do. I think if you're creating applications that are not for the enterprise, yeah, this is fun. It's okay if we have mistakes or stuff. But for the enterprise, accuracy is critical explainability even more because it's what everybody trusts. And at the end you want governance around all of this. So, what does the RAG architecture look like? And I like how he puts it up. Look, you have a query, you have this background that goes through some software and it has the knowledge graph, the database. You're trying to get all this context out that is relevant to that question. You add it back to the prompt that sends it back to your large language model. You have all this context. But the honest no BS here is that, that seems like a Band- Aid. And we were brainstorming during this episode of what does an architecture actually look like? And I think we're figuring this out. I think 2024 will be the year of figuring out what these architectures look like. And I think we'll see two architecture. I mean, there's the architecture of how we can do things fast and kind of just quick and dirty, the Band- Aid approach. But I think what we're going to start seeing is the enterprise is realizing, " Wait, I cannot really put this out in production and use it within my company using part of my product if I don't understand how this is really going to go work." And what we are starting to go see is the knowledge graphs are going to be that intermediate layer of providing all that accuracy and explainability.
Tim Gasper [00:53:35] Every company is going to try to build their own LLM or a set of LLMs that represent their knowledge and their capability and they will all run into the same problem.
Juan Sequeda [00:53:46] And connecting this back to the episode with Joe was about the data modeling is that we really need to focus on the meaning of this data. So, we know that we can effectively use AI. We are underutilizing knowledge graphs, and it's not just for RAGs. It's like we should be using knowledge graphs to help for the training in the first place. Other great quotes is, " What are we getting from knowledge graph is adult supervision." And the definition of hallucination, " Returning the best crap."
Tim Gasper [00:54:17] I don't think that was my favorite part of that episode.
Juan Sequeda [00:54:18] So, to wrap up on this, we had an episode with my colleague Dean Allemang, together with my other colleague, Brian Jacob. We had an episode just sharing about our recent work on our benchmark that we did. And that was comparing the usage of knowledge graphs when it comes to answering enterprise questions over enterprise databases. And one of the things that I think always has been troubling myself and the many people is like, " Hey, we're making all these claims that, oh, we can go ask questions and chat with our data, but all this chatting with our data is over unstructured, over the strings that are in vector databases. But what happens if you have your data in your SQL databases, in your lakehouse and so forth? What's going to happen?" And I think people acknowledge, " Oh wait, it works pretty cool and queued for small tables, easy questions. But how is this going to look like in the enterprise?" So, having that question in mind, we were like, " Let's set up an experiment and we come up with a framework where we're looking at enterprise schemas." In this case, it was about insurance using this open standard from OMG on insurance model, have a spectrum of enterprise questions, easy questions or harder questions. So, day- to- day analytics to more of the metrics and strategy questions. And then, also the spectrum of complexity on the schemas. Is it a couple of tables? Or I need nine, 10 tables to go answer these questions, and then, have explicit all that context, all that metadata, that semantics, all that knowledge. So, we put all these things together and we ask all these questions without a knowledge graph. So basically, using a large language model is a question. Writing SQL directly and using the knowledge graph and what we see is a 3x difference. So, you have three times more accurate responses if you have a knowledge graph versus if you don't have a knowledge graph. So that's really, really, really cool to see how that work has made so much impact so quickly. It went really viral and it's really cool to see.
Tim Gasper [00:56:10] How many views was it?
Juan Sequeda [00:56:12] Well, I mean, I made my post and there's like 100,000 views and between that and Twitter and then just so many people contacting us and folks like at DBT have already replicated, other startups are replicating this stuff. And so, it's really, really cool to see how people are acknowledging that we need to invest in knowledge graphs to have improved the accuracy of LLMs. Therefore, the takeaway, if we need to invest in knowledge graphs, we need them to treat semantics, metadata modeling as first- class citizens. And that was a lot. That was everything we've gone through.
Tim Gasper [00:56:47] Maybe to bring it all full circle, it seems like common sense because it is, but now we need to do it. That's kind of a theme across everything, right?
Juan Sequeda [00:56:58] Well to wrap up here quickly, what we've done here is that...
Tim Gasper [00:57:02] Yeah, you did something cool here, right?
Juan Sequeda [00:57:04] We didn't take the transcripts, as you all know, well maybe you don't know. But Tim and I, whenever we're talking, having the conversations on the podcast, we're taking notes live. So, when we provide the takeaways live, we've actually been writing them. So, I took all those notes that we have of our takeaways and then I just fed them into GPT.
Tim Gasper [00:57:25] So, this is literally chat with the takeaways of the takeaways.
Juan Sequeda [00:57:27] Exactly. So, I have some questions here. So, we're going to probably share this GPT with folks. So, whoever has the ChatGPT plus can be able to go access this. And you can ask questions like here, what are the takeaways for data practitioners?" And they say, well, the highly relevant to the work is fundamentals of data modeling for AI success, focus on data quality and people, align data work with business value, data as a strategic asset. You can keep going around these things. And what are takeaways for data leaders? Embrace core data modeling, semantics, prioritize data quality and human factors, align data products with business values and so forth. So, this is pretty cool that we're going to start doing this and kind of feeding our transcripts and stuff to GPT. So, you can, I guess please continue to listen to us, but if you don't listen to us, you can chat with us.
Tim Gasper [00:58:13] Share with us your takeaways because we want to see what your takeaways are of our takeaways of the takeaways.
Juan Sequeda [00:58:20] I'm trying to think about how to put another takeaway in that.
Tim Gasper [00:58:24] Sorry, I had to.
Juan Sequeda [00:58:25] Tim, it was a pleasure in this year's six seasons.
Tim Gasper [00:58:29] Cheers
Juan Sequeda [00:58:29] This is year. I forgot, four? We're on the year four.
Tim Gasper [00:58:33] Yeah. Year four, season six. How many episodes?
Juan Sequeda [00:58:38] 160, 170?
Tim Gasper [00:58:40] Something like that. I guess it depends on how you count. This has been so exciting. We can't wait to figure out what we're going to do next season. We're always trying to innovate, always trying to make this better. Tell us how you want this to evolve, how you want it to be better. And we're just so thankful to data. world for allowing this to happen and to you all for being an awesome audience and community.
Juan Sequeda [00:59:04] Cheers everybody.
Tim Gasper [00:59:06] Cheers. Happy holidays.
Juan Sequeda [00:59:06] Happy holidays. See you all next year.
