Artificial Intelligence (AI) Driven Real Estate Underwriting with Marc Rutzen

Podcasts

Feb 08

Our discussion is on the use of artificial intelligence (AI) in real estate underwriting. How does an individual investor or mid-sized operator, compete for deals against big companies? By leveling the playing field with the use of technology that allows real estate investors to see and use the best and most current data.

John Wilhoit: My guest is Marc Rutzen. Marc is the CEO and co-founder of Enodo Inc. Enodo is an automated analysis platform for multifamily real estate. Our discussion is on the use of artificial intelligence (AI) in real estate underwriting. How does an individual investor or mid-sized operator compete for deals against big companies? By leveling the playing field with the use of technology that allows real estate investors to see and use the best and most current data. Enodo Inc. offers one such solution that is cost effective and races past the fluff- it delves directly into the core financial concepts an investor needs to have for making investment decisions in real time.

Marc directs the development of the platform including user interface and research & development for new product features on the platform. Marc is a licensed broker. He has a Master of Science Degree in Real Estate Development from Columbia University. Welcome Marc.

I think AI (Artificial Intelligence) is something that everyone is familiar with. What they are not familiar with is that AI is having an impact on real estate and particularly multifamily. What can you share with us about that?

Marc Rutzen: To give some context, people tend to conflate real estate data and data analysis and AI and they’ll call it CRE tech. Over the past 20 years there has been a renaissance in real estate data; twenty years ago, you didn’t have data. There was no transparency into transactions. You didn’t know who owners were easily. You didn’t know what property trading prices. There was no Costar, Axiometrics, Real Capital Analytics. These platforms have brought data to the fore and made it more transparent so we know what’s going on in the market.

The problem today is that there is just so much to analyze that no one person can make sense of all the data. Analyst pay for data sources and then devote further resources to confirm the data is accurate, trying to analyze it, make sense of it and then fill a narrative for a particular asset.

What I see AI doing and what Enodo is trying to do, our mission; is to quantify the drivers of real value. Not just provide more data to analyze but tell you what that means in terms of investment potential. I see a lot of disruptive potential not just from our company but a lot of companies in the space that are pushing for the adoption of AI and it’s getting some traction.

What Enodo is trying to do, our mission; is to quantify the drivers of real value

John Wilhoit: What I like about Enodo as a user is that it’s built for real estate professionals and more specifically for multifamily which is my area of expertise. Marc, you say that Enodo automates the underwriting platform for the multifamily industry. Talk to us about that.

Marc Rutzen: Think about what you go through on a typical deal analysis. If it’s a marketed deal you’re going to have a rent roll and T12 to analyze. You’re going to look at the market for your comps (comparable assets) and you’re going to compare rents to your comps and see if what they’re achieving is reasonable.

Enodo automates the underwriting platform for the multifamily industry

You’re going to put everything into your underwriting template and put in your deal assumptions. Then export everything into a finalized product and then use this to indicate to an analyst or your superior to do the deal or not do the deal. In the case where you’re the owner underwriting the deal; you are pursuing the transaction to syndicate equity for the deal and get financing from a lender.

What our platform enables you to do, I know this is important to you John, with your writings about rent roll analysis, we have a rent roll parser on our site that it’s free right now. You can load any rent roll. This will work with most property management software. Export your rent roll from property management software and load a POD into Enodo. This will parse floor plans, percent of market rent achieved by in-place leases and lease turnover exposure (when leases will expire). It will take out cost-to-lease so you can see what the effective rent is. All of this happens in a few seconds. That is something that helps you to tee up the data and not waste time manually parsing the rent roll.

John Wilhoit: What you’re saying is that utilizing the Enodo platform compresses the amount of time required to perform underwriting?

Marc Rutzen: Yes. Absolutely. It’s a huge time-saver. Once you have uploaded the rent roll and T12 into the system you’re going to be able to seamlessly compare your rents with the market and see if they are reasonable and gain some insight into if a deal is performing as compared to similar assets in terms of year built, number of units and location and see if your operating expenses are in line with the market. That’s a quick way to use AI to underwrite deals and assess potential upside.

John Wilhoit: Split off for us how to utilize Enodo from an operator’s perspective and then from a developer/builder perspective.

Marc Rutzen: Say the scenario is this: you have a property built in the 1980s that has not been renovated and in need of some TLC. You think that there’s some potential upside if you were to invest some money in renovating units and bringing them up to market. What Enodo does is enables you to see which amenities (amenity by amenity) are going to add the biggest rent premium. This folds in directly with our mission: to quantify the drivers of real estate value.

Part of that is delving into determining the rent delivery value of granite countertops versus hardwood floors. Asking the question: what’s going to bring the biggest impact on rents: stainless steel appliances versus a rooftop deck. Which one will bring the highest rent? Enodo does that with each individual amenity allowing you to build your amenity package in the platform and see how it will perform if you were to invest those dollars and do those renovations.

John Wilhoit: That’s from an acquisitions perspective? If you’re looking at it a deal that’s on your plate for potential acquisition, utilizing the Enodo platform can help you not only analyze the deal but analyze where redevelopment dollars should go in the deal?

Marc Rutzen: Yes. If you already own the property, you can see, am I charging the best possible rent, am I optimizing my rent for this asset based on everything that’s going on in the market? It doesn’t have to be a value-add deal. You could just look to see if you’re in line with the market and how you should adjust your rents to maximize revenue.

John Wilhoit: Let’s talk about a new development deal, brand new coming out of the ground. How does Enodo assist with underwriting that transaction?

Marc Rutzen: Enodo is collecting data from about two million properties nationwide on a daily basis. This data is coming from Property Management Software, property web sites, user uploads of rent rolls and T12s. What we’re doing is building all the components of real estate value into our data. We assess and quantify buildings individually in each market so that you can hypothetically build a building from the ground up having this comparable data.

Enodo is collecting data from about two million properties nationwide on a daily basis.

That new building is just the sum of its parts; put in the number of three-bedroom units and the square footage of those units then the number of bathrooms in those units and continue this for the entire unit mix and you are building the build-out virtually. Then add amenities to each one of those different floor plans. Enodo allows you, as the developer, to virtually build out your amenity package at the community level and at the individual unit level.

Using Enodo, You Can build-out a new development – virtually.

You can instantly see as soon as you’ve crafted it (and you can do all of this in the platform) a virtual model of the asset. You see what rents the new development it can generate as if it were to exist today. You see what the comps are (the closest comps). For comps, if you’re doing development in the suburbs if there’s no new development near by the Enodo platform will go to the next similar suburb to your building and find comps. This way you always have comparable properties to compare.

And then finally the Enodo platform will show you the operation as stabilized. You can then represent to a lender or an investor this is what stabilized operations look like were the asset built.

John Wilhoit: Tell me how artificial intelligence is assisting in making better decisions about those structures that you just mentioned.

Marc Rutzen: On our platform, type in an address for your new development. You’re able to see, if you’re crafting a new development from the ground up, average square footage for units of each type in the market and what frequency of properties have different amenities. For example, I go into a market and I can see that ninety percent of buildings in this market have a pool that means I need a pool or I’m not going to be competitive. And then in Enodo you can click on that amenity and see what you’re going to get in terms of rent for that particular amenity. What’s great is sorting by the most frequent amenities in the market. You could sort by the highest premium and craft your amenities packages to select the highest premium and the most frequently listed amenities.

John Wilhoit: The “AI” part is having all of that on one page at one time?

Marc Rutzen: Yes. The “AI” is being able to individually quantify what the variables are doing. Otherwise you can’t do it. You know that Building A has X Y and Z and Building B has X Y Z and one generates more income but you don’t know exactly why. The AI analyzes a hundred different variables recognizing walkability in one location is slightly higher and the median income slightly higher than a similar location. Therefore, even with similar amenities one location is going to have a higher premium than another similar location.

John Wilhoit: So, with Enodo I can look at comps and review the amenities in each of those comps and determine which ones are adding the most value?

Marc Rutzen: Exactly.

John Wilhoit: This tells me that Enodo is amenities-centric and it allows me to add value based on what’s occurring in a particular market and not the market in general is that correct?

Marc Rutzen: Correct.

John Wilhoit: Tell us the difference between AI and machine learning. What’s the difference between those two?

Marc Rutzen: There is no difference. Machine Learning is AI when you’re training a model. What is commonly understood to be AI is a little bit different when you’re talking about it from a non-data science perspective.

John Wilhoit: For users of the platform they represent the same thing?

Marc Rutzen: In terms of what they see in the platform it represents the exact same thing. The only difference is that in AI the way people see, the way people view AI, the machine takes data and determines exactly what to do with it and that result is the output. I’s very nebulous and frightening, right? You don’t know what it’s doing exactly. But the way we craft these models is we select the features. Because we’ve surveyed hundreds of real people we know which variables are important. Then, we feed those variables into our model.

We control the range of influence that variables can have so we don’t get anything crazy excluding things like seeing a rent premium because it rains slightly less in one area. If you just fed variables randomly into a model you could come up with some weird stuff like that. We put a lot of human intelligence into our model on the front end so that you don’t get crazy results.

John Wilhoit: Tell us more about how Enodo provides locational characteristics and demographic characteristics. How can users of the Enodo platform implement this information into their decision-making processes?

Marc Rutzen: What our platform does is it takes median income, population density, average family size, the proportion of one-beds, two-beds, three-beds in the market, the walk score, transit access and negative externalities – all the things that real estate people intuitively know have an impact. If you were to put all this data into an excel sheet and try to tell me the impact of each variable it would be impossible. There’s is not enough data- you can’t fit enough data into Excel. What the Enodo platform does is quantify each variable. Then it tells you if you land on a certain place on the map the market is contributing this much (X dollars). Thus, anything you do to this property the market is contributing exactly this much. In other words, just being in this area generates X dollars towards rents.

The Enodo platform quantifies each variable

As you look at the different amenities and add them to your virtual model, each one is a component and tells you what it adds to rents. The amenities package is going to be the sum of the parts from the community and from the unit level. Our data scientists are looking to bring into the platform a tool that allows our customers to click on a map and see what the key drivers of value are in a market. Imagine being able to see that an area has an 80% college educated population and median income of $110,000 a year- that’s what’s driving value the most. You know in this area that’s what’s driving value the most. So, it doesn’t matter if you do anything else. It’s that kind of insight that helps people craft your plan for an asset, your plan for an investment to maximize value.

John Wilhoit: We’ve spent a lot of time talking about the income side. How does a developer or an analyst review the operating expense side of a deal using Enodo?

Marc Rutzen: On the operating expense side, what we do, we pull in data from all the T12s that are uploaded from users (anonymously of course). We’re taking that data and keying it into a location and then loading it into the algorithm to train. We’ve loaded about 3900 CMBS (Commercial Mortgage Backed Securities) deals nationwide. These are CMBS deals that are not distressed. We control for those that have too low of an DSCR (Debt Service Coverage Ratio) or had notes on the deal that it was under water. We use CMBS data and we use benchmark data from NAA (National Association of Apartments and IREM (Institute of Real Estate Management).

We also have partnerships with several lenders that contribute T12 operating statements from closed transactions nationwide. We feed all this data in and we control for each variable the same way we do on the income side; we do this on the expense side. We control for each variable and determine, based on the number of units, the year build, the market, net rentable square footage and income profile for this property, those being the biggest drivers, that you should be at this level for salaries and personnel or that you should be at this amount per unit to budget for insurance, taxes et cetera. We calculate that as a series of market benchmarks and then compare the outcomes to your uploaded T12. This calculation compares your T12 to market benchmarks delivering actionable information about your property and if it is in line with the market in terms of each individual line item.

John Wilhoit: When you say “market” what do you mean by that?

Marc Rutzen: One of the things that Enodo has created is an algorithm that will dynamically pull in census tracts that are similar in terms of supply and demand fundamentals; similar types of properties and percentages of different unit types and rent that they’re achieving and similar types of people seeking those assets.

What we do is dynamically build a market area instead of drawing a rectangle or circle or some oblong shape to statistically determine what your effective market area is by bringing in adjacent census tracks until we’ve got enough data to analyze statistics. Going into an urban area the “market” could be as small as a few blocks. Going into a suburban area the market could be wide yet still represents your effective market area with just less density.

John Wilhoit: We understand that markets are not round or square. Markets are certainly property specific. I think what they’ve done within this product can outline competitive market areas versus something that is often esoteric. Marc, you don’t use the term submarket?

Marc Rutzen: Yeah, we don’t use the term submarket because what is equivalent in size to a submarket is determined algorithmically.

John Wilhoit: And that offends the senses for those of us that have been working in submarkets for years and years. That’s all we know, right? We know that we like this submarket and we don’t like that submarket. What you’re doing with AI is redrawing those lines where the submarkets are more distinctive around a specific asset or group of assets versus that oblong or square or circle that you were talking about earlier. That does throw people I think, originally, to be able to look at the outputs from Enodo and see something different than submarkets. Your response to that is?

Marc Rutzen: Our response is to ask how is the submarket determined? What data went into that submarket and is it continually updating and refining? If you look at community areas within the city they were drawn decades ago and they don’t change. Does that neighborhood represent the people in that neighborhood and the properties in that neighborhood or have those boundaries changed? That’s why we’ve built an algorithm that’s dynamic that will show, at any given point in time, the most relevant competitive market area. And that may offend the senses but you want the best answer- you don’t want the answer that’s most comfortable.

John Wilhoit: Which is why we’re talking about artificial intelligence driven decisions versus things that are based on what might have been in a textbook from 1982.

Marc Rutzen: Exactly.

John Wilhoit: There’s always somebody that wants to know how you come up with that number, right? My question to you Marc is about Enodo Rents. How did you come up with that number and how do we know it’s a number we can rely on?

Marc Rutzen: The way we get to Enodo Rents is by utilizing both in-place and advertised rents. This is after removing outliers- so we don’t take anything like six-month leases or something that’s going to be artificially high. We train on that data alongside all the demographic and economic and local demand drivers. We will calculate what we’ll get (what rents will be) if you were to take this unit in this building to market today. What are you going to be able to get for it? The market rent that you see in rent rolls is not necessarily appropriate market rents to use. The trends in the market when analyzed by AI can inform you as to where rents are going and where the optimal price point is more so than just looking at the operations of your own asset. You have a look at the entirety of the market and what’s going on to determine where to price your units.

John Wilhoit: That’s right. And that’s where I come in with my book Rent Roll Triangle: The Ultimate Rental Property Grading System because you’re not just looking at what’s on the rent roll. There’s always going to be a disparity between gross potential rent, effective rent, stated lease rents (what’s in the lease document) and collections. So, there’s four different price points none of which necessarily tells you effective rents on a given day until you analyze what those differentials are and see where you should be from a market rate perspective.

Marc Rutzen: Right.

John Wilhoit: Back to Enodo Rent…

Marc Rutzen: Enodo rent is not biased. There’s no bias from the property managers saying I’ll never hit those numbers or someone saying we’ve got to keep rents artificially low to keep occupancy high. There’s no human bias whatsoever. We’re just looking at pure data in the market and determining what your price (rental pricing) should be. You may have different opinions about it and in some cases, there may not be enough data in the market and Enodo rent may not be the right one to use- I’m not going to say it’s perfect. But in a lot of cases you’d be surprised. You look at the Enodo prediction of rent and you compare it to what you think the rent should be and there could be a pretty sizable Delta. Then maybe you might want to reconsider, wait a second, can I raise rents for this unit type and if I do you I’ll make some more money on the deal, right? So, that’s what Enodo Rent is there to provide; that market context and the upside or downside potential for an asset.

John Wilhoit: It’s good to have a target, and more than one to compare, so that when we go out into the marketplace we’re not relying on just what’s in advertised rental rates because that doesn’t tell us anything about discounts or concessions. Enodo Rent takes that into consideration and tells you, or shares with you, what a unit should be renting for net of concessions. Is that a fair statement?

Marc Rutzen: Absolutely.

John Wilhoit: There have been times when two months free rent was not uncommon in metro Houston, Texas. So, that doesn’t necessarily tell you what stuff rents for, right? If there’s nothing sharing with you, the buyer or the operator, that a market is offering two months free rent as the norm to get occupancy anywhere near normalization. Enodo Rents takes that factor into consideration and tells you what the rents are versus what some conflated or inflated number is that may or may not be real.

Marc Rutzen: Yep. That’s the beauty of it.

John Wilhoit: This article is a transcription of a live recording and podcast with Marc Rutzen about Artificial Intelligence Driven Real Estate Underwriting. We’ve been talking with Marc Rutzen of Enodo Inc. Click here for a discount code to try out Enodo Inc. for free. Give it a try. Marc loves feedback. Hit them up on the Enodo Inc. web site with any questions that you have. They’re always available to you. They’re always improving their product. I endorse Enodo and I think they’re building a great system. I think it’s got legs and a great future. Hopefully you’ll take the time to check them out and to find out and see they are delivering on their promises.

Marc Rutzen: We’re excited to hear from your audience. Tell us what you think. We’re always improving as John said. We look forward to working with you.

To listen to the full podcast of this interview go to podcast.

Join the newsletter

Subscribe to get our latest content by email.
We won't send you spam. Unsubscribe at any time. Powered by ConvertKit

About the Author

Pin It on Pinterest

Share This