Smarter Property Decisions: Lessons on Data Collaboration and AI Integration

Introduction to Shay Thalek and Cortland

Drew: All right, everybody. Welcome back to Peak Property Performance Podcast. It's me, Drew Hall, along with my esteemed co-host, Bill Douglas. Bill, welcome.

Bill: Good afternoon, everybody. Morning, whatever time it is for you.

Drew: Yeah. Well, we're teeing it up today with another guest. Shay Thalek is our guest. Hey, Shay. And then I'll introduce you.

Shay Thalek: Hey, there. How's it going, everyone?

Drew: Shay Thalek is a dynamic product leader at the intersection of AI, user psychology and multifamily real estate. He currently is a product manager at Cortland, a global leader in multifamily real estate investment management, where he focuses on leveraging AI and machine learning to solve complex operational challenges. Prior to Cortland, Shay pursued his master's degree in data science at Georgia Tech, go Jackets, while consulting at Rational Labs, advising senior leadership at companies like Coinbase and Walmart on behavioral science strategies. This unique blend of technical, behavioral and strategic experience enables Shay to bridge the gap between cutting edge technology and real world user needs, delivering products that drive measurable impact. Shay is passionate about empowering teams, fostering collaboration and shaping the future of AI driven solutions in real estate and beyond. So welcome, Shay.

Shay Thalek: Thank you. Good to be here.

Drew: Yeah. Thanks for joining us. All right. Shay, today we're going to talk about data transformation, like the real kind, not just on a dashboard in a presentation. So you understand how culture and data science come together to drive adoption and outcomes in commercial real estate. So let's put some questions in here. You said that data transformation requires a universal buy-in. What does that actually look like inside an organization?

Shay Thalek: Yeah, absolutely. I think in Cortland's case, I'm very lucky. We have a very, I would say, cutting edge CEO, Stephen DeFrancis. He's someone who, since he started this, has always been peeking around the corner, trying to figure out where he can basically have a dependent edge within the commercial real estate environment. And I think around five years ago or so, he really saw this emerging ability for data to be able to drive decision making within large portfolios, started really investing heavily within that. So I think it really started from the top. There's strong technology leadership, and then they're able to develop a team of in-house analysts, data scientists, and data engineers that can really act upon that vision. I think second to that, there's really a feedback loop of, it started from the top, but I think we've been able to deliver time and time again in a way that has delivered a lot of business value to the organization. And it's kind of proven our worth, so that, again, kind of encourages leadership to double down there. That's kind of like a nice feedback loop there.

Building Trust Through Data Governance

Drew: Yeah. Okay. So how then do you help non-technical teams understand and really trust the data?

Shay Thalek: That's a great question. I would say in terms of trust, some data governance practices can be really helpful there. So in a lot of our reporting, particularly some of the more critical ones, we really do a lot of emphasis on making sure that the refresh is kind of put in every day, which might sound pretty silly, but just being able to see the last time the data was refreshed can be a really big win for some folks because they can kind of like look between dashboards, maybe see that something is updated, something is not, et cetera. That's a big thing there. We've also found data dictionaries can be really helpful here. So really going through, defining the metrics that we're using within our dashboards or reporting, kind of like saying where we derive them from. Well, a lot of our less technical stakeholders probably won't go that far. I think just having that documentation, kind of knowing that there's been a lot of thought put into it would be really helpful. And then I think a third point, which is maybe a little tangential to what you asked, but at the end of the day, I've been working in technology, I guess my whole career at this point, and generally people care a lot more about the what than the how, so really focusing on people's desired outcomes, the problems they're trying to solve, and kind of explaining how data could actually serve that purpose and populating it in the right way at the right time. So I think that's helpful for building trust, let's say.

Drew: Now that's good. And honestly, I feel like you touched on all these different types of data. So this next question I was going to ask you is about, do you feel that most companies go wrong by getting too much data or the wrong data or just not being able to communicate why in the world we're going after this data? I mean, I feel like you touched on different aspects of those, but do you find one of those or even maybe another category being most prominent where people go wrong?

Shay Thalek: I don't know if there's ever too much data. I mean, there's definitely the cost component. A lot of these external databases are super expensive. So like anything, you want to do your cost benefit analysis when you're purchasing an external data set or putting in investments to create that. It is a large cost. But outside of that consideration, I think the data we're getting on the backend is kind of similar to what y'all are seeing with peak performance property. So you can just get a bunch of data in a data lake or in databases. From there, you can kind of go in, validate, aggregate in the ways that you want to. But to get to your point a little bit more directly, I think for the end user, it's incredibly important to be both showing minimal data and accurate data. So really, by the time that you're seeing a dashboard, you're seeing a finalized, productionized model. Folks should have a lot of confidence in that. And it should be as simple as possible. And someone should be able to pretty directly tie the metrics or the dashboard or the model that they're seeing to a business outcome or process that they're trying to drive.

Bill: I love the simplicity, Shay, because that is really hard to do. A lot of times people want to see everything and then the dashboard is, it's unfollowable if that's even a word, but it's a mess. So in turning data into action, what is good data governance actually look like in commercial real estate? That's a great question. I mean, we hear it a lot, but from a data scientist, could you explain what that actually is or that, or you think it looks like?

Shay Thalek: Yeah, absolutely. I think a first step is really understanding the data sources that you're using and where they're coming from. So as I mentioned earlier, where are we getting external data sources? Like for instance, like market data, where are those vendors coming from? How trustworthy are they, et cetera, same for any of our internal systems. So when we're integrating with any like property management or financial systems, really understanding what those metrics are and kind of like how they're derived within the logic, I think it's really important there. And then in general, we focus a lot on just having a very clean database. So we do a lot of transformations in our tables and things like that. So we'll take raw data sources from a variety of sources and kind of have some master tables that are often used. So I think that's a very important part of data governance as well. So for instance, we have one table that really has a lot of information on our assets, for instance, all the communities, a big, like a lot of primary keys, et cetera, there. We have a lot of stuff built in our data from that table. And so having those kind of core tables that are really vetted and verified by data engineering there.

Effective Data Collaboration Strategies

Drew: Well, how do you move from insights implementation? Because you're going to be dealing with so many different groups. Like you've got asset managers, finance engineers, property managers. I mean, there's people that are technical, non-technical, some hate data, some love data. So how do you actually move to implementation and get everybody involved?

Shay Thalek: I think this is where kind of the product management hack comes on, I would say, just in that you never want to deliver data or a product in general that just for the sake of delivering data or the product in general, you really want to make sure that there's a strong business need, a strong business problem. So a lot of this really is just pretty informal conversations, reaching out to stakeholders. So for instance, we had one team recently where we kind of went over the fundamentals of machine learning, just to give like a sense of what that was, and that really ended up starting an interesting conversation about how we can start utilizing machine learning to basically optimize the retrofits that we do around like sustainability initiatives. So really the conversation just started around high level machine learning. This is what it is. She had some knowledge, so we were able to go into some business problems around what she was going through. Then we were able to kind of prioritize those business problems to say, hey, there are like 10 things she really wanted us to focus on, but what I like the top two based off her aspects and what are the business reasons behind that. From there, we can actually go back, kind of prototype, show like, hey, this is maybe what this would look like with fake data if we created this type of model that would help you out here. We can actually go and kind of work iteratively from there. So I think really starting with the business problem and then bringing stakeholders in every step of the way, essentially. So really, even before you do data pulls, kind of showing the prototype of what that might actually look like with fake data, making it as salient as possible, what the deliverable will be is very helpful here. Something we've kind of stumbled on, especially earlier in my time here, it's kind of delivering on a data product with unclear requirements, realizing that it wasn't actually what stakeholders wanted or needed. It's really taking that time to understand the problem really well. If it's an operational process, actually map out that process, understand where any friction is, and then go in to find those data solutions is helpful. I think like the data pull and the actual analytics, building the dashboard, building the model should really be kind of the last step after you've really validated that it's a worthwhile endeavor and the direction that the stakeholder actually wants to go.

Bill: So what I heard you say is don't solve the technical or the data problem until you know the business outcome that's desired, because it might not be contiguous.

Shay Thalek: Absolutely.

Bill: Okay. I think going back to the conversation on trust, like that's just going to build a lot more trust. If someone feels like they're included in the process every step of the way, you're really including them in understanding their problems, understanding the business that they're trying to solve. If we just come in with our data expertise, but don't understand what they're doing as a part of the business, it's not very helpful for them.

Case Studies: Data-Driven Decisions in Action

Drew: Well said. So can you share an example where Data Insight materially changed an operating decision in the organization?

Shay Thalek: Okay. So one example actually from PadSplit, which is a company I used to work at. So that was actually my entry into real estate. It was a co-living marketplace, largest in the country now, that really focuses on basically helping local property owners rent out their house by the room, kind of everything that would go with that, et cetera. We ran a fair amount of experimentation at PadSplit. That was, I think, a very helpful way to gain data, so you're able to get causal evidence, see if something's wrong, something's not. One big thing we focused on a lot was cleanliness, so kind of figuring out how do we actually...

Shay Thalek: folks in a co-living environment where no one really owes anyone anything to actually go ahead and do that cleaning. We ran quite a few different experiments, actually set up an A-B infrastructure to be able to catch that. The one that ended up really doing well was essentially we took a group of test homes, did nothing with them, took another group of test homes, we gave them an additional application where they could essentially check off the chores that they did every week and then look at other folks' chores as well. This is actually super inspired by Kristen Berman at Irrational Labs, did a similar thing at her co-living environment that she lives in. Folks would brag about chores, the thought being that essentially making domestic labor more visible and kind of creating this sense of reciprocity would really help. We actually saw around a 90% reduction in cleanliness complaints in that group over a period of three months using our A-B infrastructure. Yeah, which is pretty massive. So this is a pretty just like obviously huge signal that something we wanted to roll out. So since then this has been rolled out on the team. I've left obviously since then, but I believe they've done some iterations and stuff since then. This has kind of become its own product line. So causal evidence can be really helpful there.

Drew: That's awesome. Wow, thanks for that. Something that sounds like something that needs to be included in the high school curriculum.

Bill: Wait, wait, cleanliness in high schoolers? I was about to say I have like a 90% improvement. Man, I want to see that on that forum. You know what I'm saying? I've got a couple of high school boys, so that's what makes me think about it.

Shay Thalek: I will say after a year, it dropped to about 40% of running the experiment. And it's still amazing. That's really compelling results. That's amazing. Wow.

Overcoming Status Quo Bias in CRE

Drew: All right. Well, okay. So let's think about like culture and collaboration. How do you get a property team or even an executive team to embrace change when they've been doing, like we see it all the time. I'm sure you see this all the time. They've been doing things the same way for 20 or more years, just forever. They've been doing the same way. So how do you bring them along for the ride and compel them?

Shay Thalek: Yeah, that's a great question. I will go back to my first answer that the top down approach has really helped in Portland and that having the entire C-suite bought in on kind of this more streamlining of operations and technology and data has been incredibly helpful there. I would say in terms of my interactions with the teams in general, I think pulling people into types of like design thinking workshops can actually be really intuitive. One example I'm thinking of, one of like the main flagship products I work on, which I can't really go into the specifics of, but about like six months ago, for the first time really with this product, because it was very early on before, kind of like very scrappy, we decided to do like a year planning process. We said, hey, what's like the next year of this product going to look like? We brought in a lot of stakeholders who hadn't really been a part of this type of process before. I think this type of thinking is just very intuitive in a lot of ways. So for instance, in this case, we had an executive kind of divine decision for what we wanted this product to look like. From there, we kind of started asking how might we actually achieve this vision, right? So I started asking a lot of questions and folks naturally have so many questions. This isn't like an unintuitive thing to folks who haven't been a part of design tech process before, kind of goes in well. And then from there, we could really go into like future ideation ideas, how we could actually improve. And one VP in particular, you know, it's been in this business for a long time, been at Cortland for 10 years, just took to it like a fish in water. It was like incredible. Honestly, I was working with the UX designer. She was like, I just love, I love this person. She just had so many good ideas. She's so excited. She's so engaged coming into it. So I think just like in general, bringing people in, educating them, not talking down and really coming with a collaborative lens. Like I think a lot of this kind of like design process, which ends up leading to good data solutions is really like intuitive. I think we're all kind of designers in our ways.

Bill: Yeah. No, that's good. I feel like so much of this does go back to like garnering that buy-in by proving the process. This is not just a box to be checked. It's not just a project that happened to be funded. So we have to do it, you know, inspiring. I mean, I guess that's true for many, many types of projects, but especially here, it feels like just getting that buy-in because it can seem so daunting. It can seem so ethereal and distant, like, ah, what are we really doing? So that's good. That's good. It's kind of bringing some tangibility to it.

Drew: Okay. So this might tie back into your earlier answer, which is really cool about sort of storying, showing potential designs and inputs and, you know, modeled outputs and things like you were talking about in terms of having an actual session with someone before real data starts to flow. So this might feed a little bit into there, but in and around that topic, are there any tools or frameworks that you rely on to make data more accessible to non-analysts?

Shay Thalek: So frankly, Excel is amazing. Prototypes, I would say. Yeah, see, I'm really just, if it's tabular data, you can just put in a table, you can put in like fake data, whatever you want. If you wanted to add a couple charts to kind of show what that would look like on a dashboard, you can do that really easily. So that's really, I think our team's main go-to comes to some of like the more lower lift items or like V1s type things. Some of our more flagship products, I've been exploring a lot of like the generative AI coding tools that are coming out more. So some popular ones, they're like Cursor, Cloud Code, Bolt. There's a few examples of ones I've used. And these essentially allow you to actually build like real front ends that people can kind of just like play around with as web applications. So for instance, there's a big analytic feature I want to push in Q1, which is going to be a big lift. It's going to be like a custom build, essentially, from scratch. And I've been working with the team main stakeholders for the last like couple months or so. I started with just like interviewing, kind of shadowing, mapping out their user journey of how they're generally making the decisions that this dashboard got. And then from there, I was able to really just start going into, hey, here's a tool, like here's me walking through a demo of the tool, like let's talk about it. We've had a few sessions since then, folks have played around with it. I've been able to kind of in real time change the design or like within like a couple of weeks and things like that. We're at a point now where we're essentially design ready because folks have had that opportunity to really see it. And ideally, when the thing's actually released, it's going to look exactly like the prototype design, essentially. Folks are going to get exactly what they expect and a lot of the ambiguity is taken away. And then to your point earlier, like it's not like this ethereal thing. It's very tangible, people can play around with it.

Drew: Yeah, that's fantastic. Yep. I love the thought and the planning that goes into it, Shay. I think that that is an approach we don't see a lot of. Your data group is so mature there at Cortland. If you could wave a wand and fix one data problem across commercial real estate, what would it be? It's just a fun question. It could be persons, it could be technology, it could be anything. And that's meant more for the industry, like you see it across the industry, not just in Cortland.

Shay Thalek: I think if I could wave a magic wand, it would probably just be kind of like an open-mindedness and willingness to change. Coming from behavioral science backgrounds, there's a big concept there called status quo bias, which is what it sounds like, basically. Folks are generally really inclined to do what they've already been doing before, which is incredibly fair. We get used to what we're doing. It works, maybe to build some identity around how we do our work. So these are really tough psychological challenges, not downplaying that. But yeah, I think in general, in commercial real estate, there is an opportunity among all firms. I'm not really comparing Cortland to anyone else. I think it's just kind of universal. It's really embraced this emerging technology. We're in a very disruptive period. So a lot of the work that you're doing, there's so much opportunity to really bridge this gap between the physical and the digital by utilizing sensors, IRT, et cetera, actually capturing this data. And then outside of that, kind of property management, financial data, et cetera. So much to utilize to drive NOI, make better decision making. And I think you can just be really intimidating to folks a lot of times. So kind of just slowing down, feeling like taking the time and opportunity to actually embrace these new technologies and figure out how it could help folks. It's probably the biggest driver of change in the industry from what I can see.

Agentic AI and Future Data Science Trends

Bill: We all know status quo bias is real, but I like your approach to slowing down and explaining the why. So that's of great value. Thank you. Okay. So looking ahead, what do you think is next for data science and real estate, or what are you most excited about over the next three years?

Shay Thalek: So I think agentic AI has a lot of use cases, which at a high level is essentially letting tools like ChatshubbyT interact with tools like your calendar, make updates in databases, et cetera. There's a lot of opportunity here because I think there's a lot of manual work in real estate, a lot of different processes, types of flows. We can essentially automate these or at least like speed up folks a lot, or potentially provide real time information that could really help with either in the leasing process, if a resident has an issue, things like that. AI could, I think, generate some really interesting insights there. I think agentic AI, just because it's such a new technology, is probably the most untapped among real estate and like all industries in general. So that's very exciting to me. I have a machine learning background. That's a lot of my focus. So I'm also very excited about the capabilities there. It's a technology that's been around for a long time, but I think we utilize a lot more within real estate environments, just being able to basically predict the future. I mean, it's kind of magic. And there's a lot of need for that in real estate, from rent growth to figuring out what's going to break the building to anything, really.

Drew: Yeah. Well, it has a lot of applications in commercial real estate, assuming you have the data. It all goes back to where's the data and where's the digital infrastructure, as you well know. So what's one small step any owner or operator can take today to start building better data habits?

Shay Thalek: That's a great question. I think in general, like find one problem that has been like really difficult to solve with the current information that you have. And I would start there and then kind of figure out what data would actually help here. Do I own this data to your point? If not, how could I own it? And then really just like introduce one project. I think case studies are such a powerful thing in organizations, because if you can really show that you made one

Drew: It really inspires the rest of the organization to kind of adopt that type of methodology moving forward, I think. I've seen that time and time again. So I think really just starting small, being problem first, so focusing on a problem that you think would be really high impact, that you theorize the data could solve, identifying the data you actually need for that, and then working with the appropriate teams or consultants to actually build out that product.

Bill: So start small and go slow. I like it. Start small, go slow, but in that small, like pick a high impact thing. Ideally, a high impact effort, yeah.

Drew: Well, in the book, we call it the pick your play. So we're using sports analogies, of course, but pick your first play. Don't go there with a whole book and try and do them all at once, pick a play. Break it down, yeah. Yeah. Awesome.

Personal Insights and Career Advice from Shay

Bill: Okay. All right, so before we wrap up, Shay, we always like to take our guests to what we call the extra floor. It's just a quick set of five questions off the cuff, gut level, about randomness here. Short answers. So here we go. What's a book or a podcast that has shaped how you think? Besides, obviously, besides this one.

Shay Thalek: Great question. I guess sticking with the behavioral science front, Influence Expanded Edition by Robert Cialdini. It was over some of the kind of like poor behavioral science philosophies about how humans interact with each other. It was one of the most influential books I've ever read in terms of kind of like how I operate in a business and just my general day-to-day life. Couldn't recommend more, yeah.

Bill: Awesome. Well, what's the best piece of career or life advice that you have ever received and you'd like to share?

Shay Thalek: That's a great question. How do I narrow that down? This, yeah. Shay's Nuggets. Yeah. This is maybe a slightly biased answer because it was somewhat recent, but when I started in this job around a year ago, one of our, the person who onboarded me, worked on the HR team, gave this advice. That was essentially, we're just starting a new job. You're like a sponge. You can only absorb so much at once. And I think that's really helpful advice for onboarding, but in general is really helpful. I think particularly like when you're working with emerging technologies, there's just so much out there. There's so much to learn. It can be pretty overwhelming. And I think people could like just throw their hands up and give up, but just kind of really acknowledging that there's only so much we can learn today. Only so many new things that you can take in. It's also very helpful for me. I think in particular, because I'm definitely someone that could be a little impatient with the learning curve. So that's very valuable.

Drew: That's a fantastic answer, yeah. That brings up like literally images of past experiences that I've had professionally where it's like, man, I think I'm at the threshold. I think this is it for the week or you know what I'm saying? Maybe for the month or something. That's great advice. Great advice.

Bill: Okay, what's a small daily habit or ritual that makes a big difference for you?

Shay Thalek: Hmm, I'm really into breath work. So I like at least for like 10, 30 minutes a day, we'll do like the kind of like four, eight breathing. So four in, eight out. I find this just keeps my nervous system like pretty regulated. Helps a lot with just showing up to meetings, calm and curious, not with like my own agenda or trying to press anything. And I think that type of energy definitely really helps in terms of building trust with stakeholders and advancing my career.

Bill: I love that. I know that you and I have had conversations about breath work and fitness and food and all kinds of things. So I was wondering how you were gonna answer that. So if you weren't in data sciences, what do you think you'd be doing professionally?

Shay Thalek: Woo, in a dream world, I'd be an indie rock musician. I'm actually about to release an album with some friends in a couple of months. I'm pretty excited about it. Oh, awesome. Outside of that, I think I'd probably be a high school math teacher. I'm really passionate about math and education.

Drew: Well, you rockstar math teacher would be really popular in school. Absolutely. That makes me think of the school of rock musical or movie for that matter. Totally.

Bill: All right, last one, Shay. When you're not working, what do you do? Or what do you love to do that recharges you?

Shay Thalek: Woo, so I talk a little bit about this, but I really like rock climbing. It's something I'm not very good at, but enjoy greatly, which I actually really like because it kind of pushes away my perfectionism. It's like, hey, I'm probably gonna get better pretty slowly at this, but while I'm doing it, I'm gonna have a blast. That's a big one. And then, yeah, I guess I just brought this up, but I love playing music. I'm in a weekly jam group, so I do that. And yeah, just try to play whenever I can, basically.

Drew: Right on. Nice, do you like to climb outside or on an indoor rock wall?

Shay Thalek: Pretty much only indoor, but I would like to try outside at some point, yeah.

Bill: Yeah, I'm told it's a big difference. I've only done indoor a couple of times. I'm not good at it, but I'm curious about it. This won't make the recording either, but super quickly, I remember when I first took my first job out of college, and it was in Boulder, Colorado, and it was a bunch of climbers, right? And I was barely 20, and I'm like, oh, I'm gonna show these guys because I thought I was a hot shot at 20. You know, you can do everything when you're 20. And we went climbing, and I was great for about 60 seconds. And they're like, this young sucker doesn't know what he's doing.

Drew: You were Spider-Man for a minute?

Bill: Oh my gosh, it was so funny. I'm like, man, you guys knew. You knew, and they're like, yeah, we did. Oh, that's funny.

Closing Thoughts and Farewell

Drew: Well, Shay, thank you very much, and thank you, everyone, for listening. I appreciate you being here. Everybody listening, be sure to follow up, like, and subscribe, and we look forward to catch you on the next episode.

Shay Thalek: Thank you again, Shay. Thanks, everyone. Thank you, thank you for having me.