Using data to make better real estate investment decisions [podcast]

Investor Mindset Podcast, Let's talk residential real estate investing - Episode 3

David Garrard, Chief Data Scientist at Konfidis joins us today to discuss how data can answer your key questions when thinking about residential real estate investing and help you make better investment decisions.

Tune in to this episode on your favorite podcast platform:

Spotify

Apple

Google Podcasts

RadioPublic

In This Episode We Cover:

• Sources and data applications in the real estate industry

• Predicting rental rates for various real estate asset types

• Challenges in real estate data quality and how it’s normalized

• Using data to identify where you should buy your next investment property

• Leveraging data and trends to find discounts and premiums in the marketplace

The live video version of this episode can be found here. View all Investor Mindset show episodes here.

Full transcript:

John Asher: Hello and welcome to The Investor Mindset, today, we are talking data. Ten years ago, we used to talk about how big data was going to become the new oil of the future. My name is John Asher, Co-Founder and President of Konfidis. At the time, the ambition was clear, let's take the largest amounts of data possible. Let's clean it, let's analyze it, and let's look for answers to our big questions like, “Where are the best neighborhoods to invest in and which properties are the best to own?” A lot has happened in 10 years, and it feels like we're starting to answer some of those questions. Today, please welcome David Garrard, Chief Data Scientist of Konfidis to join me to talk a little bit deeper about how we can use data to answer these questions and to really make it better for investors. Welcome, David.

David Garrard: Good to be here, John. Pleasure to talk with you about data.

John Asher: So we have a very special treat. Obviously, our backgrounds are totally different this time, and it's a very data-inspired kind of theme. So, David, please, can you give us a little bit about your background and how you came to join us at Konfidis?

David Garrard: Absolutely. Well, I've been in software engineering and data science for more than 20 years, and I started way back in the early days of web development. Back when programming and web were still finding their feet. Well before the web 2.0, 3.0, and machine learning was widespread. I worked in a business process automation, e-commerce, publishing industry positions, and after I've built up the book of business, I created my own production company gradually built larger and larger projects. As a result of that, I was able to work with much larger data over time, and when you start larger datasets, you start having to worry about data pipelines, big reporting frameworks, performance, and analytics outcomes over and above what you will be able to do with a traditional Excel-based tool for example. Eventually extending all the way out into a terabyte scale data and, things with billions of rows. The really interesting thing is I've worked in many different industries related to data science and software, but the client-side of the equation or the user side, which is I've seen similar problems solved in different industries as people approach you know, the patterns repeat themselves in certain ways and includes things like energy, loyalty, healthcare, marketing, communications, these kinds of things. Prior to joining Konfidis, I was at Last Call Analytics. I was a CTO [Chief Technology Officer] there, that's a SaaS business serving the beverage alcohol and cannabis industry. We'll talk more about that later, but that data sets like transactional data, retail information, very different patterns in real estate.

John Asher: So let me ask you the first question then, it's clear your background has not been in real estate, but it's certainly been in data. So, when you start to look at and think about data applications for real estate, are there practical applications for data in real estate?

David Garrard: Oh, yeah, absolutely. I mean it’s hard to imagine an industry that wouldn't be able to apply data but real estate is rich with data. It's looking at the data in terms of its different characteristics. So, real estate data got a lot of quantitative data, a number of bedrooms and square footage plus a lot of qualitative data that more recent technologies are now coming to bear things like the text descriptions, the categorization of photographs, or GIS [geographic information system] data that we could bring to bear. So there's a lot of interesting things in real estate that allows us an enriched data landscape that I think there's a lot of tech that can come to bear in that area.

John Asher: Alright, so let's talk about how we actually make it work. So let's take a step back. And let's think about, when we think about data, data applications, what data is available to us, when we want to start thinking about mining it for better answers?

David Garrard: So, we bring a lot of datasets together and then the key here is to align them together if you will. That needs to be aligned geographically, in time, as well as in definitions. So the kind of the building blocks for that in Konfidis as a brokerage, we get MLS [multi listing service] data feeds of property listings that you consider to be familiar with. Those are very wide fields with very rich data, it comes to us in various different formats. We bring that in and make it all the same so we can analyze properties in a comparative fashion. Then we try to do is join that data with a whole bunch of different public data sources to enhance our understanding of either the property, the neighborhood, it's in the city it’s in, or whatever trend characteristics we can bring to bear. Some examples of that kind of thing are all of the statistics of demographics we might have, population growth trends, a lot of neighborhood level information, like walk scores, school locations, the scores of those schools at different grade levels, building on that activity in the areas, transit locations. Then a lot of sort of macro statistics regarding the overall real estate, resale or rental market behaviors in those localities and all of this stuff occurs of kind of different levels. Some of that stuff, like, whole counties or townships or cities or some of it’s more localized and the definitions will differ. So we do a lot of work to make it all smooth and comparable so we can have a more accurate picture to compare property A to property B, even if A and B are in different cities or even provinces. And if I may draw another major area that we bring to bear is our rental data. One thing that's extremely relevant to our investor clientele is predicting rents and the trend of those rents for different kinds of units. I'm sure talk about that more today, and that is an ongoing activity for us and we gather data from our clients, from our partners, and from private, public listings that we bring into bear and to build up a bespoke dataset and that's our main dataset supporting our particular activities in this area.

John Asher: A lot of times when I'm talking to clients or investors, I always get the question of, “Well, that all sounds well and good. But how do you really know if rent is going to be a true rent prediction for a property?” Can you just talk to me about what do you have to do to the data so that you can get good predictive analytics?

David Garrard: Well, I mean, the biggest determiner, of course, is just large sample sizes, so we do a lot of work to collect a large amount of rental listings to address any sort of statistical concerns you might have regarding how many things go into a given decision. And over and above kind of the baseline information, so that would include things like obviously where it is, but the number of bedrooms and baths, the type of unit it might be located in, things like that. We also know information from our MLS relationships in terms of the prior history of a given property for example, and we build a level of knowledge beyond the listing to classify it and one of the biggest areas of classifications is unit types. So you can say, well, what is a three-bedroom rent for? Well, that varies quite a lot between, is that a multi-storey residential three-unit bedroom? Is it a house with a backyard? Is it like a condo? And obviously, the bedroom counts will vary in terms of the housing stock, but you have to understand that next level what kind of thing is this? And then we take that and we align it with the actual investment plan that we're aiming for. So in many cases and SFR [single family rented] plan, you know, renting it as a whole home, but conceivably, you would maybe want to do a duplex conversion to a legal duplex. See why no one has an upper unit or a lower unit call for a triplex and we have answers to those different flavors separately, they're not just a blended number, like two-bedroom equals dollar X.

John Asher: I like blended flavors. Sounds like a fun melting pot of data. Okay. I can truly appreciate the data but let's go back one more step. What are the challenges you have and even just faced by you when it comes to making sure that things make sense? I often and I have, unfortunately, one of these people that does these things where you know if I have a spreadsheet and I'm trying to collect and record data, I might not do it consistently. If I think about even MLS, there are 90,000 realtors in Ontario and that means 90,000 possible different ways in which someone's going to input data. How do you start to tackle those problems?

David Garrard: You're quite right, real estate data is interesting. And this is something I learned when I sort of got into it, is that it's very wide. So you might discover that we know like 500 things about properties in general. It's an extremely wide spreadsheet, if you think about it in Excel terms, 500 columns, but it's often extremely sparse. So you will discover that there are lots of optional things that may or may not have been filled out or are relevant. You consider like there's information about a swimming pool. Well, if you don't have a swimming pool, a bunch of things are just blank. Right? And while we may not care a great deal about swimming pools on the margin this creates a great deal of kind of extra noise. And we do a couple of things with this problem. One is we need to understand where we can do gap filling, what kind of information can be substituted, where there is missing information. Or we understand where we can use broad location averages and in place of things. Now we don't look at all 500 fields equally of course, some things are more important things like prices and fees. The quantitative values tend to be things we care a lot about, and to answer a question like “How do we normalize it?” Well, another big challenge is definitions, and not just between properties, but data sources, between feeds and boards. So we discover things like what is the type of the basement? Well, actually, there are probably 30 different ways a basement is described. But we don't necessarily want 30 different flavors of what are probably only five or six major classes of basement. So we have a lot of logic that converts different source definitions into a Konfidis definition of what we describe the property as and that's applied across lots and lots of dimensions to normalize the information prior to us doing any effort to compare things.

John Asher: Okay, so that's all well and good. Thank you very much. You've organized, sorted the data but listen, I just want to know, where's the best neighborhood? Where's the best city? Where's the best property? Okay, and that seems like a rather straightforward answer. And again, I think if you spoke to 10 different realtors in the room they might tell you why the neighborhoods they service could actually be the best property. But is that true? How would you actually go about doing that? Finding those premiums.

David Garrard: There are different answers at each of those levels. So you're quite right like city, then neighborhood, then the currently purchasable property. So I'm gonna start at city level, and there’s going to be more strategic choices like I'm building a portfolio or I'm a new investor, where, what municipality should I be operating in? Before we're talking about specifics beyond this point, and when we're looking at cities, what we're trying to understand is what are the broad long-term differences between cities that allow us to, an apples-to-apples comparison across lots of cities? I want to compete every significant municipality in Canada to each other and understand for different asset types, what are the most virtuous options. And asset type, I mean, single detached homes, condos, typically sometimes we might also look at townhomes and semi-detached as a distinct asset class instead of those buckets of things. And the biggest city determination, dimensions we typically explore are - what are the prevailing long-term home price appreciation HPI and what are the prevailing gross yields? So gross yield is a function of prevailing purchase price for typical properties and prevailing rental rates. To talk about it right, like where do we get those rental rates or the purchase price that's fairly widely available data you see in the news. And were just trying to get macro comparisons between these things, and we do that for each asset type that's widely for sale in those locations like condos are not widely available in every city so we don't bother looking at condos in small centers, for example, we look only in major cities. We discover some interesting things about this. So, taking couple of examples tends to be a good way of describing this. If you look into the Kitchener, Waterloo area, to the west of Toronto, you discover that it's got really excellent historical price appreciation, great price positioning in that city, those cities, for single-detached homes very popular, but pretty average to below average gross yields. But on the other end of the equation, if you were to go to Edmonton and invest in a condo, you would have had terrible home appreciation, but you would have had excellent gross yields. So you can imagine when you're sort of choosing in terms of what is your portfolio balance, or how do you want to balance out different sources of return? You can look at these dimensions together. What we have found is there's kind of a band in the middle which is more of a sweet spot that allows you to get good performance in both of these dimensions at the same time, instead of having to focus just on one or the other exclusively.

John Asher: Right, yeah for an individual investor I mean if you're trying to invest in capital appreciation it might feel a little bit more speculative. Like, it's all capital gain. There's tons of literature out there around cash flow, cash flow is king. And that's probably the predominant strategy in the U.S., but you might not see those houses actually appreciate. So, I guess that those are the dynamics between the two ends of the barbell. But now what you're beginning to talk about is like a total return, style, investing, and thinking about particular neighborhoods. Now, let's say you're right, and you're able to calculate those kinds of returns for a city. Now let's go one dimension deeper. Okay, so let's say I found the city, it's got the right balance between yield and capital appreciation. But now how do I start to think about, “Should I buy a whole home? Should I buy a townhouse? Should I buy a condo? Should I buy an apartment?” How do I start to think about that?

David Garrard: In terms of asset types, we are able to separately look at rents as well as price and in most cases of appreciation, differently by asset type. So it's absolutely the case that, townhomes might outperform single detached homes in certain localities or vice versa, that does happen. And so we can certainly talk about asset type. We can also look at trends, and I think that's really the key aspect when we talk about looking at these performance trends differently from each other. And we start looking at long-term trends and pricing by those asset types. And then we go a level deeper into the neighborhood versions of those. Once I've made a selection of a city for example, “Okay, great. I want to go to London, Ontario.” But like, London is not that small, where do you want to be in London? And what we start breaking the city down in is to understand for rents, for prices, what are the long-term trends and how can we start correlating those to other things that lets us try and seek leading indicators? And when I say earlier the data we joined into properties to understand them, those are the kinds of examples of things that help us so we want to look at. Alright, what happens when new transit stops are created? Like in the case of the GO transit network. What happens when you see school scores rising? What happens when you see an increase in new building permits or renovations or different kinds of building permits? What we're trying to do there is identify trends that we can isolate within neighborhoods to compare them within cities we've already identified to see where can we outperform in a neighborhood versus the city as a whole, and that, again, might vary by asset type. It also may just vary by neighborhood. You move a kilometer West and it's different.

John Asher: I met a lot of realtors, and a lot of great realtors, and the very best ones are the ones that have been in neighborhoods for a long time. They understand the quality of homes, they understand the demand notes. And you know, I've been challenged quite a few times by great realtors who know their local markets by saying, you know, and being challenged like well, “Could you truly understand that the north side of the street is going to be a better rental property than the South Side?” And the real answer behind it is you know, there's a nice green space behind the northside of the street and on the southside, there's, let's say hydro lines. Can you truly see that and replicate that kind of experience? And maybe it's not a question for today but thinking about tomorrow, is that possible?

David Garrard: I think that in the future. It's possible. It's a matter of scale. Right? So well, a given realtor might have a great understanding of the change in a particular neighborhood. We're looking to compare hundreds or thousands of neighborhoods to each other, across hundreds of millions of transactions in real estate. And that's really what we're talking about, you know, big data as it’s referring to machine learning techniques, and try to regress out those particular characteristics. So you know, which direction it's facing in orientation with a property. The proximity of green space are great examples of property-level information that lets us understand across a very large sample, and this is really critical for this kind of thing because you got to look at a lot of this stuff to distinguish between other characteristics. For example, like, you know, did that have a nice bathroom versus a not so nice bathroom? You have to separate out those effects from having a park or not having a park. And so that's what scale is really the critical aspect. And I think, yes, there's a lot of work to be done. I mean, this is, relatively early days in terms of identifying, you know, really, really specific, but you could certainly do it in general.

John Asher: Big question, because this is the one I'm always excited about. What excites you about the future of data science and real estate investing?

David Garrard: Well, the future of data science is really fun because there's been such essential democratization of machine learning and NLP techniques, natural language processing techniques in the last five or so years. And what that really means is the accessibility to take these technologies and apply them to new industries and new problems, is a lot more accessible than it once was. And this lets us answer questions like, what are the effects of orientation on prices you indicated earlier, and so I'm really excited about taking my experience from prior industries and problems I've solved there, as well as technologies, and bringing them to bear within the real estate space specifically for investment audiences. And I think there's a lot of opportunities for us to take those learnings and get tremendous gains.

John Asher: Very cool. Well, I'm here and hear the police sirens behind you Dave, hopefully, it's not the realtor police coming to get you for their insights. But, you know, I want to ask kind of like a final question, which is, does this work allow you to move across the country? Does this work allow you to move across North America? Does this work allow you to move globally? To begin to understand and compare, where are some of the best places to invest? Not just in Ontario, not just in Canada, not just in North America?

David Garrard: Absolutely. I think as we gradually have more and more data available about locations and neighbourhoods, and properties. It's a feast of information available, especially within GIS data, some geographic information, we can apply those to any place allowing you know, I'm an investor, many thousands of kilometers away in different time zones to feel very comfortable that I have a complete understanding of due diligence. And John, I don't think it's just a matter of distance. I think it's about speed as well, so I think a lot of these technologies will allow you to move much, much faster and to feel just as comfortable to make a sight unseen purchase decision, maybe and or including finding tenants and capital improvements as well without feeling like it's an uncomfortable position. Because of the amount of due diligence information that can be brought to bear and the confidence you have in the predictions.

John Asher: Very cool. Thank you, David. I really appreciate your time. As always, and if I could think about kind of a final thought, and I'll go back to the first comment I made. Ten years ago, I remember standing in front of people and saying and preaching that big data would be the way of our future, you know, the new oil. And, I have to admit, 10 years ago, we didn't have the real foundations to be able to analyze it. And, I have to admit again, I can remember saying it's big data and I truly believed it, but I didn't know what the stepping stones were going to be. Flash forward. I now see it. I can see it. I know our team sees it. And it becomes exceptionally powerful in analyzing, “Where should I be investing? What my confidence intervals might be getting to look like?” And it even starts to open up areas and thoughts that I never had before. Around, yes, there are new opportunities. Yes, there are new communities and cities that are pointing in the right direction. David, this has been really exciting. I don't know if people listening or watching, but when it comes to data, and it comes to technology, this is the stuff that I really get excited about and I know it's a big part of our future. David thanks once again, appreciate all your insights. We'll talk again real soon.

David Garrard: Thank you so much.

---

Ready to start the process of buying a real estate investment property?

Try out Konfidis, a real estate investment company that identifies and vets the top investment properties in the market today. We've developed a simple way to invest using big data, cutting-edge technology, plus our team of experts to help you outperform the market.

Talk to us today about our Top Reviewed Investment Properties of the week.

Talk to Us