Overview

Machine Learning and Artificial Intelligence (AI) are oft-cited buzzwords in our contemporary conversations about property operations and analytics -- but what are their meanings and, more importantly, how can their application benefit multifamily business operations? 

If you are encountering these concepts in internal meetings, vendor presentations, and sessions at conferences -- but wish to take a deeper dive, you've come to the right place.  This article aims to demystify the concepts of Machine Learning and discuss their role in the broader context of Marketing Analytics. 

So, first, the terminology.....

What is Machine Learning?  How is it different from Artificial Intelligence? 

Machine learning is a method of data analysis used in computer software to identify patterns in data and make recommendations for optimization based on those patterns. 

Artificial Intelligence is a broader discipline, that includes Machine Learning -- but also represents a higher level of cognition. An example of Artificial Intelligence technology would be in certain types of robotics, where a more human-like sense of reasoning would be applied -- such as with self driving cars -- whereas Machine Learning is more focused on analysis and logic. 

In this article, we will focus on Machine Learning, as a fundamental concept for building Marketing Analytics systems. 

How does Machine Learning work? 

The basic principle of Machine Learning is that it takes a set of assigned data; finds patterns within that data; builds a model around those patterns; then tests and refines that model to make recommendations based on the patterns in that data. 

What are some examples of Machine Learning in our everyday lives?

"Hey Siri....?"

Most of us are already using machine learning, and training machine learning models, in our lives as consumers.  

Call Centers and Voice Search

One example of Machine Learning in action are call centers that use voice recognition for their navigation prompts.  Although these systems are not yet perfected, you may have noticed that their voice recognition capabilities have improved immensely in recent years.  That's because we, as consumers, have been training these systems -- and, as a byproduct, their machine learning models.  Consider the different ways that the same word can sound from different people -- based on dialect, inflection, the pitch of our voices -- so, essentially, you have an infinite number of ways that the same word being pronounced.  Machine Learning technology in call centers has enabled all of those variations of the same word to be associated with a single word or voice command. 

The same principle applies to voice recognition systems such as Alexa and Siri.  Mobile devices are capturing and transmitting vast amounts of data on a continuous basis, we train these models as a natural byproduct of our ordinary use as consumers. 

CAPTCHAs

CAPTCHAs, the photos you select on websites for security verification, are also an example of machine learning.  For example, through our input, the CAPTCHA software learns to associate the word traffic light with photos of traffic lights -- thus making the model more accurate. 

Amazon

The product recommendations on e-commerce platforms like Amazon.com also are an example of machine learning.  In this case, Amazon uses data from your former purchases and purchases from other consumers to recommend products that you may be interested in. Furthermore, recommended products that are purchased also contribute to the model. 

Netflix

Machine Learning is used by streaming services like Netflix to make personalized recommendations to its users based on shows that they have watched, genres they are interested in, and even the duration of shows and movies they have viewed in the past. 

In all of the above examples the same basic principles of Machine Learning apply -- data is collected from users, correlations are found with certain behaviors, the model is then tested and refined over time. 

What are some other real-world applications of machine learning? 

One of the main benefits of machine learning is its ability to quickly identify commonalities among large amounts of data.  

In the medical field, vast amounts of data can be analyzed to determine warning signs based on a patient's medical history or through image analysis. The larger pool of data that you have available to access, the more correlations can be identified. 

Machine Learning is able to help drive this by analyzing vast amounts of data much more quickly that the human mind is capable of, as well as making its conclusions available to a wider audience. 

How does Machine Learning work for me (and my company)?

So, the ultimate question for many Multifamily Insiders readers, how does this all apply to the Multifamily space? 

Predictive analytics has been used in multifamily for building economic forecasts for decades.  Consider all of the factors we use to market and predict the performance of a property -- local economic conditions, seasonality, market rate, area comps, etc. 

In Marketing Analytics, the same principles apply -- there are numerous factors that contribute to the strategy we employ for our property marketing campaigns:

• Occupancy goals
• Seasonality 
• Resident retention
• Lease expiration 

Google PPC (pay-per-click) campaigns represent one of the best opportunities to leverage Machine Learning, which factor in business goals, market demand and budget to make recommendations on Google PPC strategy and spend.  One of the major reasons for this is that both the data and the function of the Google PPC software are contained within the same ecosystem -- which means we have a great amount of actionable performance data.  Some of the factors that a Machine Learning algorithm analyzes are:

• Floorplan availability 
• Lead generation
• Marketing budget 

Based on these needs, Machine Learning software can make recommendations on the content, timing and spend for Google PPC ads necessary to meet your business goals -- e.g. number of leads in a given timeframe. 

Ultimately, Machine Learning systems allow for more impactful interpretation of data to make better, more informed decisions on marketing spend and strategy.  By identifying patterns and building models based on a broad amount of data points -- Machine Learning technology can provide predictive models to optimize your marketing campaigns.  

Other Considerations

Granted, there are layers of complexity to Machine Learning models that we did not cover.  For example, if a model draws a certain conclusion based on the data presented does not mean the model or data are accurate.  If you're familiar with statistics and analytics modeling, you know that there are scores of different types of models -- each with its strengths and weaknesses.  Programming and analytics modeling, testing and validation is part of the backend process in building and refining a Machine Learning system. 

Conclusion

Machine Learning technology complements your Marketing Analytics strategy by quickly analyzing a wide range of data points that enable multifamily marketers to make more informed decisions on marketing strategy and spend.  

As with any tool, human experience and objective reasoning should also be employed. The role of data is to help inform and guide our decision making, but not to supplant it. Hopefully these basic principles of Machine Learning covered in this article help provide more context on how this technology works and how it can impact your organization.