Artificial Intelligence Multifamily

From downturn to upside: using AI to help retain renters

First off, a little bad news. 

According to Freddie Mac’s multifamily Midyear outlook, apartment rental growth is slowing across the country while vacancies are on the rise.

In the report, Freddie Mac speculates that apartments could weather a 9% drop in collections before they breach any covenant headroom on DSCR (that’s assuming 1.25x). 

Okay, that’s not so bad. 

What’s more, falling interest rates are creating a double whammy on both higher cashflows (through lower debt payments) as well as tighter cap rates on real estate.

But this begs the question: in this market, with a wildly uncertain future, does retaining customers even make sense? 

Absolutely, for these three reasons.

(1) Applicant credit quality: In previous recessions (take 2008 for example), consumer credit quality was massively impacted. In fact, FICO data shows that consumers with higher scores saw their scores drop for up to five years after the recession. So it pays to keep heads in beds.

(II)Concessions on new leases: These are only going to rise as we’ve already seen in Q2 due to poor demand and oversupply.

(III) Turn costs: With new amenities (e.g. smarthomes) and increased sanitisation, “turning” a unit will require both a higher capital expenditure and time than in years past.

Current Residents - Renew Your Lease by September 30th! — College Station  Apartments | Student Apartments in Normal, IL

With a recession in mind, Beekin developed Leasemax.

Utilizing a multivariate model, Leasemax allows apartment operators to accurately predict who will stay, and helps retain your best customers at the right price.

Over the past 6 months, in the midst of the COVID-19 crisis, Leasemax has picked longer-staying renters – and retained them!

Check out these results for yourself:

Expected Beekin renewal probability band% of residents who actually renewed No. of happy residents
Retention Study: How 1,000 leases performed in the midst of the COVID-19 outbreak. NB: The 75% renewals in the 0-19% band are statistically skewed by only 4 leases.
Cartoon Move House Stock Illustrations – 1,139 Cartoon Move House Stock  Illustrations, Vectors & Clipart - Dreamstime

As with all machine learning approaches, stability and objectivity matter.

That’s why our data science team spent time ensuring that our model didn’t violate fair housing. That meant eliminating variables which were considered “protected class.” Of course, when we did this, a fair amount of our model accuracy was lost. Maybe having kids does make people move, we’ll never know!

To recoup some of this accuracy loss, our team feature-engineered variables while augmenting the datasets with internal data that was purchased, curated and normalised. This included neighborhood, mobility, and demographic data across thousands of neighborhoods in the US. 

The Beekin datastore, painstakingly constructed over the past 18 months, began to deliver big results, with our feature-engineered model performing flawlessly across MSAs!

Play offense – or defense.

Think of Leasemax as the ultimate data strategy play. It’s built to optimise pricing by overlaying strategies, much like stock market strategies.

Or for fans of Moneyball, we empower data to help transform you from a regular “scout” to a pennant-winning, come-from-behind underdog. In fact, our resident node which predicts whether a resident will stay or pay, is codenamed “Billy” for that very reason!

Get closer than ever to your customers. So close, in fact, that you tell them what they need, well before they realize it themselves.

Steve Jobs

Use Leasemax to find the perfect customer for your assets. It’s a natural win-win for landlord and resident, alike. Thanks for reading.

To learn more about what Beekin can do for you, we invite you to reach out personally at for a free portfolio analysis. 

Artificial Intelligence Data Engineering Multifamily

Are Americans moving more?

Average Length of stay for Renters across apartments and homes in 11 cities in the United States, and 5 year change

It is a known fact that renters move. Some more than others, and sometimes it’s a happy move (moving for making more $, larger family, better weather) and sometimes, well, it’s not.

Landlords, in their search for utopia, want renters who (1) keep paying them higher rent every year than the year before (2) pay their rent on time (3) don’t wreck the place and (4) stay for extended periods of time. Utopia may not exist, but thankfully data science does. It can not make you young, taller, but it can find patterns about why people move, and once known, can predict when someone is likely to move.

In our quest for THAT answer, Beekin data science analysed 5 million renters using their mobility data, rent data sourced from 12 different public and private datasets. A few months of hard work, but some cool results that we feel pretty excited about.

We found resoundingly similar patterns across cities, and an understanding of when people are likely to move which depends on both their demographics, the property as well as the area they live in. Duh, but wait, this is a quantitative model versus a judgement.

In the first part of a 3 part series, we output the- renter tenure or length of stay across a few cities in the United States. This helps guide both investment and asset management decisions.

Overall – Los Angeles did well compared to most cities, with renters staying for an average of 5.7 years and an increasing rental tenure. We’ll get back to why that could be in a bit.

In contrast, most cities saw a marked drop in rental tenure over the past 5 years. This is proven and probable: things like lifestyle changes, job transitions and new rental housing stock in these cities all contribute to people moving a lot more.

Interestingly, all of the cities with falling tenure have seen strong net migration as well as net new construction starts. Specifically, people moved a lot more into – Dallas, Denver, Atlanta, Washington D.C.

The more red a state, the higher the net migration into it. No pun intended.

A snapshot of 2018 Census data below, shows us in-state moves and moves into the state (net migration %). With some obvious skews (D.C. as a state is small, but people live in VA, MD), there is a visible trend toward the sun belt states.

The graph below shows you how many people lived in the same house 1 year ago, how many moved in-state and a % of out-of-state and international migrants in as a proportion of state population.

US census 2018, net migration into the 11 states for the 11 cities

In the spirit of compliance, while understanding these relationships, Beekin data science “muted” variables which could be considered discriminatory including marital or family status, ethnicity or race. The results are robust, fair and scalable. And if done right, can lead to a happy renter and happy landlord.

Hope you have fun using them.