How I Use Data Analytics in Real Estate Investing

[Image generated with Dall-E]

Or, how I failed my way to success

Our business of over 16 years is delivering reliable passive income properties.

A reliable tenant must continuously occupy the property to generate a reliable income. A reliable tenant always pays the rent on schedule, stays many years, and takes good care of the property. Since you will own the property for many years, it will need multiple reliable tenants. The best way to increase the chances of always having a reliable tenant is to purchase properties that attract people from a segment with a high concentration of reliable people. Therefore, my first task was to find a tenant segment with a high concentration of reliable tenants.

As an engineer, I used the standard approach of analyzing data. I tried multiple data sets and wrote a lot of software, but (I would be embarrassed to tell you how long I persisted) I finally decided that classic data analysis would not work. The fundamental problem is that humans do not behave algorithmically. So, I ditched this approach.

Next, I decided to mine historical rental data to understand what people did in the past. I downloaded about ten years of MLS rental history and started over. I tried many things (that failed), and then I plotted monthly rent vs. length of stay. Amazing!

The result was similar to the chart below, showing a strong correlation between how long a tenant stays in the property and the amount of rent. This was the starting point I’d looked for a long time.

I began investigating the tenant segment that stayed over five years. I converted the low and upper rent to approximate gross monthly income using: monthly rent / 30% ≈ gross monthly wage.

I next interviewed property managers and cruised job boards to determine the probable jobs based on gross monthly income. I concluded that people earning below a certain wage tended to have lower-skilled jobs, which made them vulnerable to layoffs during economic downturns. Therefore, I raised the lower income threshold above this wage.

I next looked at the upper-end income range and discovered that the jobs above a certain wage were primarily administrative. I believed these workers would also be laid off during economic downturns. So, I lowered the upper-income threshold below this wage level.

The result was a narrow wage/rent range that I believed to have secure jobs due to the nature of their work, as shown below.

I next had to figure out how to attract this specific segment.

Attracting a Specific Tenant Segment

Each tenant segment has specific housing requirements and is unlikely to rent any property if it does not meet all their requirements. So, if you buy a property that matches the housing requirements of a tenant segment, most of the applicants will be from that segment.

To determine the characteristics of properties that would attract this segment, I used data analytics to determine what and where they rent today. From this, I created what I refer to as a property profile. A property profile is a physical description of the properties that this segment is currently renting. A property profile has at least four elements.

• Location - The locations where significant percentages of the target segment are renting today.
• Property type - What type of properties are they renting today? Condo, high rise, multi-family, single family?
• Rent range - What the segment is willing and able to pay.
• Configuration - Two bedrooms, three-car garage, large back yard, single-story, two stories?

I ran correlations between properties identified by the property profile and actual historical data and found a high correlation. After so long, I thought I had what I needed.

And then reality came crashing down.

The problem was that many new listings could come on the market on any given day, and the properties that appeared to be candidates would often go under contract within two or three days. This left us only 24 to 36 hours after a property was listed to identify it as a potential option, evaluate it, gather analytical information, compile it into a report, and submit an offer, as depicted in the image below.

Doing the above process manually for hundreds of properties each day was impossible. So, once again, I turned to software.

Our Data Mining Engine Today

I’ve worked on data mining engines since about 2007. All the algorithms I tried were similar to what Zillow and Opendoor were using, which was not nearly good enough to make purchase decisions. Finally, in about 2015, I discovered a very different methodology to find good properties. I am still enhancing the software to this day.

The data mining engine architecture is illustrated below.

After years of improvements, the engine can find the small number of potential investment properties from among thousands in less than 30 minutes.

However, data analytics can only go so far because it only deals with data, and we are dealing with humans.

I next put together a team and processes that took the output of the data mining engine and selected properties that matched individual clients’ requirements. These properties are then rigorously evaluated by a team of experts, as illustrated below.

Only if a property matches the client’s requirements and passes the evaluation by multiple people do we send the client a property report (not an MLS data sheet). Due to our software, processes, and team members, we can evaluate a property in about one day and present it to our client.

Do our data analytics and processes work? We’ve delivered over 490 properties to clients worldwide. >90% of our clients buy more than one property. Our largest source of new clients is referrals from existing clients. Of the 180+ clients we have worked with, only 9 or 10 were local; all the rest lived in other states or countries.

Income reliability? Our average tenant stay is over five years. We’ve had six evictions in 15+ years with a tenant population >1,000. During the 2008 financial crash, our clients had no decrease in income and no vacancies. The market value of their properties fell, but no change to their income. We had similar income reliability during COVID and the eviction moratorium.

In my own opinion and the opinions of our clients, our data analytics and processes are effective.

Summary

Data analytics and processes are the cornerstone of our business. Without data analytics, we could not find the properties needed to meet our client's financial goals. Also, we could not evaluate properties fast enough to make offers before they were gone.