Subscribe
Share
Share
Embed
A show to explore all data matters.
From small to big, every company on the market, irrespective of their industry, is a data merchant. How they choose to keep, interrogate and understand their data is now mission critical. 30 years of spaghetti-tech, data tech debt, or rapid growth challenges are the reality in most companies.
Join Aaron Phethean, veteran intrapreneur-come-entrepreneur with hundreds of lived examples of wins and losses in the data space, as he embarques on a journey of discovering what matters most in data nowadays by speaking to other technologists and business leaders who tackle their own data challenges every day.
Learn from their mistakes and be inspired by their stories of how they've made their data make sense and work for them.
This podcast brought to you by Meltano - "Unlock the Insights in your Data"
Aaron Phethean (00:27)
Hello Kevin and welcome to the show. Absolutely awesome to have you on and the beginning of season four. You're our first guest and I cannot wait to try out this new format. We've known each other a while and we've worked together but now probably more than ever I get to understand what it's like for you on the other side of the fence and really get to know what's going on in your working life.
Maybe you could kick us off with just a little bit of a description of you and the company you work for and just for everyone listening so they can get a sort of concept of what a day in your life is like.
Kevin Sampson (01:01)
Definitely. Yeah. So, um, Kevin Sampson, um, I worked for Vertex service partners. Uh, I've been with Vertex service partners, uh, coming up on two years now. Um, so I was their very first data hire. Uh, Vertex is a residential and commercial roofing company, um, that is private equity backed. Uh, and we're a P roll up of, um, a couple dozen brands across the United States.
So we've really created what we kind of called partner services, but we're kind of the backend office to a lot of these roofing companies providing their marketing support, HR support, finance support, data support, everything like that. So it's been a really interesting area to bring technology and data and analytics into traditionally not a very technology.
forward industry and also being their very first data hire, which has been an interesting problem set. Prior to Vertex, I got my career start at Amazon. I worked for Amazon for about four and a half years. Kind of started as a business analyst or a data analyst there, moved more into business intelligence engineering, which is kind of their word for essentially an analytics engineer.
⁓ kind of went more into the, the data engineering and data architecture side of things. feel like people, when they become a BIE at Amazon, typically either go like more data engineering or data science is kind of their tracks. went kind of more towards the data engineering side and then, ⁓ was looking for something new and, met our, our, VP of technology, Scott, ⁓ who's, who's my boss.
And he told me about this opportunity to kind of take their data and analytics platform from zero to a hundred. And it was really interesting. It's been really interesting so far. So it's a, it's a fun ⁓ area to be a part of.
Aaron Phethean (02:51)
Yeah.
Cool, you didn't.
Yeah. I mean, it's actually, it's crazy to me the way things work and it may be limited to the U S you know, this kind of roll up idea of, you know, lots of brands and you know, you kind of almost who would have thought that you could come from a tech job in Amazon looking at, you know, the, the scale of data that they're dealing with to roofing. And there's actually tons of data there that you're dealing with across, you know, lots of different, you know, regions and companies and you know, there's, there's an awful lot there to do for one person to build.
Kevin Sampson (03:21)
Yeah.
Aaron Phethean (03:30)
the ground up. So yeah, what a cool role.
Kevin Sampson (03:32)
Definitely.
Yeah, it's very related. At Amazon, I was part of their logistics organization. So kind of saw how data and analytics can be pushed into operations. That's all roofing is, is it's a sales organization and it's an operations organization. There's really no difference between that and other types of companies that are more like technology enabled. ⁓
Aaron Phethean (03:56)
You know, I
think it's super relevant to anyone. Actually, almost every company is boiled down to that. You know, you're making sales, you're making, you know, delivery of some sort that needs to be serviced. you know, that as a picture for everyone, I think they will be able to relate to. Yeah. The other thing that stands out to me about, you your role and what Vertex do is that, you know, you're looking at
You know, quite a broad set of data to make decisions. And you mentioned something there earlier about the making your entry into IT. If you start your career in IT, it does go one of two ways. Either you get to know the business more, or you get, you know, sort of more technical in your role. And I wonder, you know, just, just for everyone who's listening, we're trying to add a new format on the, on the podcast. And the idea is that we have a kind of fairly standard four set of questions.
Kevin Sampson (04:41)
Mm-hmm.
Aaron Phethean (04:51)
you know, four questions to ask. And, you know, the first question is really about, you know, the moment in your career where you thought data's for me or, you know, the kind of point at which you thought, yeah, this is, I absolutely love this. And I wonder if you can think back and, and, you know, bring it to life for us. So when, when did you think I must do this more? And he's got to go the buzz for it.
Kevin Sampson (05:15)
Yeah, I mean, it was pretty early on in my career at Amazon. had a, they had a traditional problem that a lot of companies have where they have a dashboard sprawl. They have, you know, metrics looking different in different places. Really kind of leaning more into like that data architecture, developing a strong data model. So I think that was kind of the point when I was
really interested in it, being able to, you know, they did a big transition from, from Tableau to QuickSight, AWS' dashboard offering. And a lot of the dashboards were just like built like for like, as opposed to kind of taking a step back and being like, what do we actually need? What are these delivering and how can we consolidate? So kind of being able to, you know, track usage across the dashboards, identify what we need, dig into each one of the dashboards.
Where is it pulling data from? What like insights are being delivered on this platform? What can be consolidated? How can we kind of like reorganize the data marts that are feeding into these things? And I really kind of found that like data architecture lens to be really interesting because you're essentially organizing. Like you're just reorganizing the closet of your data and
Also, kind of starting the communication again with stakeholders as to like what this metric is, what does it mean? Like literally what, you know, where conditions and case statements are in the SQL Code that are delivering, you know, this metric to you. And is that what you like currently view it as, especially in an organization that has a very mature data model, you know, like someone could have
put in an assumption years ago and we're still working off that assumption and we really challenge that assumption since then. So that was kind of a time that I thought was really interesting. And I also, you know, coming over to Vertex, that was kind of one of the big draws to me was I could build a data model from scratch. And I can take all the things that I've learned about what is a sustainable
data model that is hands off, has a lot of like insights, can give people the information to know why, you know, a particular job falls into this metric or doesn't fall into this metric. So it was kind of an interesting problem set. And I've really like kind of taken that like data architecture lens with things. And I think that's becoming more and more relevant now.
especially the data model side as we're moving into this kind of AI first world where you need to have like a really solid data model semantic layer to your data model that you can pull into BI, you can pull into AI, you can do ad hoc reporting on and you're going to be getting the same results from all these different sources.
Aaron Phethean (07:46)
I totally agree.
Yeah, I was actually going to make that comment when you were telling us about the other, the way your journey started at Amazon, that it was actually sounded quite meta actually. Like it was data about data, data about data being visualized by people like, who is using it? What does it mean? Like, actually I can totally see the relationship there that you're talking about, know, actually describing what a thing is and then unbundling it. Who cares about it still? what, who's even looking at that report?
Kevin Sampson (08:18)
Yeah.
Mm-hmm.
Aaron Phethean (08:33)
But yeah, it makes total sense that you'd relate that to everyone's challenge today, which is definitely about what data means and how to...
Kevin Sampson (08:42)
Yeah. And there's a big
trust component to it, right? Like if, if your semantic model is not organized in a fashion that can be pulled into multiple different tools and there's any sort of distrust that begins with it automatically as a blocker for people like adopting what you're building. so I think it's, it's more and more relevant as we go forward in kind of the data, the data world, as we're advancing these tools.
Aaron Phethean (09:07)
Yeah, I wonder whether you found that, let's say in the pre-AI world, when you showed a metric to someone that was, you know, they didn't trust or lost faith in, like, how did they respond at that time?
Kevin Sampson (09:21)
⁓ I think, I think it's an evolving thing. Like I think that in the past, it was just like, this is, this is wrong kind of thing. And I think it's, it's, I've learned to like ask the follow on questions of like, well, what is it wrong? How do you expect it to be? Give me a specific example of what you think would be right. Let me see how that deviates from how, how we have it. And, you know, in a pre AI world, you know, that would be.
30 minutes of my time to pull that together. Now I just get to ask quad Code like my question and it spits out an answer in two minutes and does 30 minutes of work for me. So it's kind of evolved in that way where it's it's understanding what their expectations are and then telling them why or why not the metric falls within their definition.
Aaron Phethean (10:01)
Interesting.
Yeah. Is the conversation pattern quite similar? sort of I'm wondering like, so you're utilizing chord in a very similar way to you would have thought about it yourself. It sounds like you're saying like, do you actually find the similarity in the pattern there?
Kevin Sampson (10:26)
I do. Yeah. mean, I think that the someone that was kind of like a mentor to me at Amazon, like gave me the hint that like, you just need to get down into raw data and understand like what they're, what they're viewing is right. And also, it's also a good like pushback to the stakeholder of like, okay, if you're telling me this wrong, then tell me what's right kind of thing. So I think you're able to
give the same sort of prompting to Claude in order to like be like, hey, here's like the actual job idea they think is right. So like, tell me why it's wrong. And ⁓ it's able to process that pretty well.
Aaron Phethean (11:04)
Yeah, it's sort of, yeah, sort of wonder, you know, sort of wondering out loud here. If you didn't have that experience, it would be quite challenging to ask the stakeholder the question to push back, but it also would be quite challenging to ask Claude Cohn the question to get, you know, to get the discovery. So, yeah, there's no substitute for your experience in that domain. That seems quite insightful.
Kevin Sampson (11:20)
Yeah.
There's, yeah,
there's definitely a level of experience. And I think that's kind of one of the challenges that are probably a lot of like junior folks have nowadays, like coming into the workforce is they just don't have that pre AI experience to kind of rely on in order to tell AI to give you the results that you want, you So I think that'll probably be a little bit of a market differentiator between
know, folks that survive in the data world and the post AI world and people that don't, you know.
Aaron Phethean (11:54)
Yeah, yeah, yeah, exactly. And maybe that is a perfect time for a second question. So I wonder what you think the data industry as a whole is getting completely wrong at the moment and what should they be doing?
Kevin Sampson (12:09)
Yeah. I mean, I think there's a little bit of a confusion in the data world around like delivering insights and delivering answers. So like, you know, especially in like a non-technical organization where people are like a little bit less in the weeds, they just want to have an answer to the problem, right? But not necessarily know like the insights that come from that. Like they're not spending the 20 minutes on the dashboard.
drilling down, looking at raw data, going back to the CRM to see how it differs, this and that. So I think that there's going to be a time when operators kind of have no idea how the business actually works. They're just getting these like pre-chewed answers as to what's wrong. So I think that's kind of one of the things that's going a little bit wrong. I mean, every single company, right, especially in the BI world is like,
Aaron Phethean (12:46)
Mm.
Kevin Sampson (12:58)
leaning into AI and like building LLMs on top of their BI platform. And I think that's kind of a little bit of a dangerous line of like, you know, how much like an LLM is going to give you like the answer you want to hear, you know, and it gives this it gives this very like confident answer as to what's wrong. But if you don't understand the inner workings of
Aaron Phethean (13:04)
Mm-hmm.
Yeah.
Kevin Sampson (13:23)
your business and how the data is collected, then you'll just accept those answers as truths when they may not necessarily be like the actual answer to the question that you're trying to have answered.
Aaron Phethean (13:35)
So if I could play that back to you, I think what I'm hearing you saying is that generally the data industry's optimizing for answering a question, but actually there's a sort of fundamental problem with understanding whether the question's a good question or understanding the results. that sort of, know, that's generally, everyone's just optimizing to make question answering faster or easier, but there's no substitute for the research, if you like.
Kevin Sampson (14:02)
Yeah,
I think that you're losing a lot of that research. Exactly. That's a good way to put it. You're losing the critical thinking lens that you're looking at the data and the results that we're giving you by just solely relying on an LLM to give you answers. So I think that there's a lot of...
education that can go into stakeholders as to how to like research things that they're trying to figure out, you know, ⁓ by having a healthy balance of like LLM given answers and actually yourself going in and like trying to derive answers yourself with your own
Aaron Phethean (14:32)
Mm-hmm.
Kevin Sampson (14:43)
understanding of how the business works, your own understanding of the SOPs that you have, and like how data could be either falling into the metric or falling out the metric that you're trying to get an answer to.
Aaron Phethean (14:58)
Yeah, that's interesting. The kind of second part of the question was, you know, what should we do? We'd be doing instead. I kind of have a notion of like that, you know, that investigative process. I wonder if it's in your mind a technology solution or is it more like a training and you know, human solution that you feel would be the approach there?
Kevin Sampson (15:17)
I mean, I feel like right now in current landscape, it's more of a human thing. think that in the future, it could be a technology change that occurs. Like right now, I think like the AI is more of like a, like a exploration accelerator. Like I can ask it a question and get an answer very quickly. But it's not necessarily like a conclusion thing or conclusion machine that you should be.
necessarily relying on. It's kind of more of like it guides you down the path of what you should be looking into as opposed to like giving you the like explicit answer of what you should be doing. ⁓ So I think that's kind of the
Aaron Phethean (15:56)
Yeah, almost like you asked a silly question,
but should you have asked a better question, it would be perhaps a, know...
Kevin Sampson (16:01)
Yeah. Yeah. And it's so much of
it's prompting, right? You know, like, ⁓ there's so much of a, like even just comparing the questions that I would be asking an LLM, knowing our entire data and versus someone that doesn't know all of that. Like we get vastly different answers because I'm just inherently prompting it with like exact columns that I basically want it to be querying, you know, as opposed to someone that doesn't know that. And, and it's just kind of giving it a more ambiguous answer. And it's just.
Aaron Phethean (16:23)
Yeah.
Kevin Sampson (16:29)
It's just picking whatever it thinks is, is correct. A lot of that comes down to semantic layer stuff, right? Like, ⁓ but I think that like, semantic views that like dbt puts out stuff like that are very baseline, you know, it just, it just has a metric. It just has synonyms and it just has a definition. So, mean, that's what it's relying on. And you can obviously add additional context to it, but like,
Aaron Phethean (16:32)
Yeah.
Kevin Sampson (16:53)
gaining all of that context from years and years of knowledge that a lot of these operators have is really, difficult. So ⁓ I think.
Aaron Phethean (17:01)
Yeah, it's not a two line
description, it's a paragraph, it's a life story about a single column.
Kevin Sampson (17:06)
Yeah, it's
an entire 30 years of experience being distilled into a README file. It's a lot. So think that's probably where technology will catch up eventually. in current state, I think it's more of just a quick answer. Give me my little insight, but it's not necessarily a decision maker. We need to change the course, our direction, of course.
with what it's telling me.
Aaron Phethean (17:31)
Yeah, okay. Well, very excitingly, it's time for our third question, our kind of hot
⁓ just for everyone who's listening, what the attention, the hot take is that, know, Kevin's never heard this question and it's going to challenge the way he's thinking. And it's quite surprising that we were already talking about semantic layers. And I was at this conference, actually it's a meetup, know, couple hundred people in a room, a speaker on the main stage, and they made the comment that is going to be your hot take. And they said that they didn't believe that AI needed a
semantic layer. And I wonder what your reaction to that is, is how do you feel about would it work well without a semantic layer? Could you imagine like, does it serve a purpose at all?
Kevin Sampson (18:16)
think not having a semantic layer works in a business that is very simple. think that it doesn't, if it's a straight, we track sales the same way, we track revenue the same way, we track leads the same way as the industry standard for doing that, then I think not having a semantic layer could potentially do that.
And if you have enough like reference materials for the LLM to be like derived from If your business is not like that which is most businesses Then I think it does need a semantic layer because there's there's specific There's specific definitions and assumptions and ways that
people are interacting with your data collection software, whether that's a CRM or an ERP or whatever it might be, that having a semantic layer to define what the thing is that's being derived, the metric that's being derived or the process that's being derived, then having a semantic model is crucial to that.
⁓ also just like on the synonym side of things, right? Like, that's one of the big challenges that we have in our business is that there's like traditional, like roofing terms for what we would call things. And then there's like more business oriented names to things. And then there's all of the gray area in between those two things, especially when you're, when you're working in a business where.
your a roll up of where you have roofing companies that might be using more of those roofing terms and then you have all the business folks that haven't worked in roofing their entire lives and are coming in with other terms for the same thing essentially. ⁓ So
Aaron Phethean (20:05)
And I made it across companies.
So I've often seen that, that, you know, even the same industry, but different company has a, their own acronym, their own description of a thing. So yeah, if you've got a roll up of 14, 20 different companies, there's going to be bound to different.
Kevin Sampson (20:15)
Yeah.
Definitely, definitely. Yeah, there's
a lot of just lingo things that ⁓ like a semantic model provides that you just can't get if you don't have that.
Aaron Phethean (20:23)
Mm.
Yeah, it's a, I mean, feel like it's a new thought to me, at least like, you know, when you're in a country like Switzerland, or someone that's quite isolated, you know, the local dialects develop because they don't really talk to one another. And that's sort of isolated. Like, that's what's happening with companies who are obviously experiencing the same thing when they're just holding a roof, but call it different things. You know, that's, that's super interesting.
Kevin Sampson (20:46)
Yeah.
Exactly,
exactly. For sure. I mean, there's, there's, yeah, and then then there's like metrics that are unique to your business, right? Like, we have a metric that that we use a lot called ⁓ two legged or one legged. And all that means is that there's a there's a belief that a salesperson has a has a better chance of closing a lead if like
the two people of the household are there as opposed to just one person of the household. Cause then the excuse can't be, well let me like go back and talk with my wife about this and we'll get back to you. Instead you have both people there that can make the decision on the spot and you can walk away with a signed contract. like stuff like that is unique to the industry, right? And ⁓ not having a semantic layer or a semantic model that defines that would be lost.
Aaron Phethean (21:44)
Yeah, that's really interesting. That's really, really interesting. So now the, you know, the kind of third part of the third question in the podcast, and we've sort of flirted around, you know, talked about AI a little bit, now we're kind of taking it head on. What's
AI being, what was the impact of AI on your day to day? Like how has it changed how you do things? And it's been so recent it feels, but yeah, probably is evolving. yeah, how has it changed things for you?
Kevin Sampson (22:10)
Mm-hmm, yeah.
Yeah,
I really feel like in the second half of 2025 is really when these LLMs got like good enough and having like ⁓ skills that you can add on and MCP servers and all these other like additional things that you can add on to your like Claude Code and other CodeX and whatever you use have really changed things. For me, it's made me a ton more efficient. Like I no longer...
actually do any of the mundane work where I need to add a column to a model, have it flow down into the reporting layer table. I don't do any of that anymore. There's no reason to, especially if you know exactly what you want and you're able to prompt it for exactly what you want and it does it and you review it and then you commit it. There's no reason for that.
And then on another side too, mean, it's, it's allowed me to push more into kind of the software development or SDE world where now I'm able to build react apps off of things. I'm able to, um, yeah, this there's an Azure skill. use, we use Azure for our Claude infrastructure. There's an Azure skill package, which is like amazing. I no longer have to like mess with Azure at all. Um, I no longer have to.
great containers, just creates containers for me and gives me the keys that I need in order to have data pushed to it. So it's allowed me to be able to push more into other areas of development and engineering that I'm not experienced or skilled in where like my ⁓ skill set has mainly been analytics and data engineering side.
but I've been able to kind of push more into that front end, which has been a big game changer. Yeah, I know what the outcome is that I want. I'm able to test it. I'm able to go back and forth. I'm probably a little less efficient than someone that does have that actual front end experience, but it's an enabler for me. It allows me to do that kind of stuff that I'd have to...
Aaron Phethean (24:02)
Yeah. And you know the outcome. Yeah.
Kevin Sampson (24:23)
take months for me to figure out. Instead, it's taking me days to do it. You know, it's incredible.
Aaron Phethean (24:26)
Yeah. Yeah.
And there's someone who's coded in React applications. It's, you can spend an enormous amount of time looking at some tiny state issue that has just completely gone away. You that's, yeah, never had an environment that would be such a big step.
Kevin Sampson (24:38)
Exactly, exactly.
Yeah.
Aaron Phethean (24:43)
Interesting
that we asked everyone who comes on the show to think up a question for the next person. But of course, being the first person of the first season where we're doing this, I had to ask LinkedIn what the question for you should be. I gave them a little bit of a description of your situation. Quite a lot of pipelines, quite a lot of data.
Kevin Sampson (24:59)
Yeah.
Aaron Phethean (25:05)
quite a lot of things to look after. And it seemed that pretty unanimous that everyone wanted to know how on earth you did your job with only you. And so the question is, how do you push back on new requests when you're already at capacity as a team of one?
Kevin Sampson (25:23)
Yeah, mean, prioritization is the answer to that pretty simply, there's a lot of, and this is kind of where LLMs have helped most recently, ⁓ a lot of ad hoc answers or like things like that. I have been pushing people to LLMs to answer for themselves. That was, I think that the ad hoc side of it is what can really bog you down as opposed to like the prioritize.
quarterly projects that you need to work on. you know, Cloud Code has been a part of that most recently. I think that my life has gotten a lot easier on the development side where I'm able to push a lot of the, develop a framework of what I want, spend a couple of days kind of basically developing like what's going to be my initial like Cloud Code request.
⁓ doing a very, spending a lot of time in like plan mode and developing what the plan is and then letting it execute, on that plan and then kind of clean up from there. but I think in in the other part that's allowed that is, you know, I've, you know, I've been at Vertex for almost two years. The first year of that was just building was, was getting data ingested doing the EL part.
And then doing a ton of transformations and really, really building out the, the data model in order to be, very like sustainable and, don't know the term is, but where it can like kind of self-suffice self-suffice issues. Like I have, there's no hard coded anything and anything that I've built so that it all kind of flows upstream and downstream correctly.
Aaron Phethean (26:30)
Mm-hmm.
Kevin Sampson (26:53)
⁓ So that if a change needs to be made, I'm able to go to the top of the funnel, make that change, and then it flows into everything that I need it to flow into. ⁓ So I think that's kind of how you handle a lot of it as like a team of one is having a really solid data model that you can rely on ⁓ and then leveraging tools to kind of work on some of that more tedious ⁓ development work.
Aaron Phethean (27:02)
Yeah, I feel like that's an interesting.
Yeah.
Yeah, I feel that's a really compelling insight there that, you know, if you're going to serve the business at pace, then you do need a solid foundation. And maybe early on.
I hazard a guess that the business expectations were lenient or lower because they knew you were building some infrastructure, some foundational stuff out. And now that you can give rapid answers, they're delighted and they're seeing the benefits of that. Maybe the fact that you've also got AI is like you're helping think, you're now amazing to the organization. You're doing so much stuff. It's such a game.
Kevin Sampson (27:38)
Yeah.
Yeah.
Yeah.
Yeah,
I think the only other thing I'd add onto that is like your data stack really matters. And like an example is when I first started, we were using like Power BI as our BI and I had no experience with Power BI before. And I just found it really cumbersome to use and it was hard to deliver on things. So I quickly kind of realized that
One, it was difficult for me to use. Two, it wasn't really giving people the visuals and insights that they wanted. So we went hunting for another BI platform. We landed on Sigma, which has been really great for us and is a lot easier and quicker to iterate on. So I think just kind of finding things in your stack that take a lot of your time and shopping around for other vendors that are gonna take less time for you to use is also pretty crucial.
Aaron Phethean (28:41)
I can definitely imagine the benefit of that. You know, that the interface with your users, if they can be more productive, the requests are not coming to you, you know, you're the best tool for your job. You you use dbt, you build the dbt. That's, you know, that's probably the best tool for you, but the user needs to be the best tool for their job, you know, for them to get the answers.
Kevin Sampson (28:41)
and being able to develop as a team of one.
Yeah.
Yeah.
Definitely. Yeah. And yeah, just creating, you know, a real solid like gold layer group of tables that then people can create their own visuals and draw their own insights. And you know that it's coming from like a solid gold layer or semantic layer of tables that have common definitions. ⁓ So it all kind of comes back to data model. So that's why I think kind of that data model looks ⁓
Aaron Phethean (29:21)
Yeah.
Kevin Sampson (29:28)
experience was kind of crucial in my development as a data engineer and analytics engineer.
Aaron Phethean (29:33)
Yeah.
There's tons of things I could go into there and our experience of working with clients and trying to explain that they need to invest in that layer because they'll reap the benefits. Perhaps one for another show and another day. I want to get your question for the next guest, if you could. So this is your chance to ask a challenging question or a question that you'd really like to hear from someone else, very similar roles across the series.
Kevin Sampson (29:47)
Yeah.
Aaron Phethean (30:02)
If you could share this, what would you ask the next guest?
Kevin Sampson (30:02)
Yeah.
Yeah, definitely. mean, kind of going along the lines of how quickly things have been developing. But what's a belief that you held strongly about your data work three years ago that you now think is wrong? And what came to change your mind about that belief that you had?
Aaron Phethean (30:22)
Cool. It's a question. I can't wait to hear the answer. It's very impressive. Very much everyone's battling with it now. So thanks, Kevin. I ⁓ really appreciate it. And thank you for coming on the show. It's been a huge pleasure.
Kevin Sampson (30:36)
Definitely. Thanks, Aaron.