Last year, I joined VTS as the first data science hire, helping to make data science a reality at the company. I was super excited by VTS’ unprecedented picture of leasing activity in the industry (we capture >75% of office building activity in top US markets). With this data, we have the potential to drive decisions across the commercial real estate lifecycle, from identifying emerging markets to invest in, to optimizing the layout in a building to maximize NOI.

16 months in, VTS Data Science is a full-fledged squad incorporating predictive insights across many of our offerings! In this blog post, we will walk through the key decisions that made such a transformation possible.

Getting the Data Foundations Right

When I joined VTS in the fall of 2020, I noticed that while plenty of valuable data existed:

• A lot of querying, transformation, and cleaning was necessary to make the data useful to interpret, analyze and build models off of
• There were shortcomings in the tools used in VTS to perform the above operations, in terms of speed, reliability and security

To unlock our data potential, we realized that the first order of action was to uplevel our data infrastructure. This involved working with our talented analytics engineers to:

• Migrate our data warehouse from Postgres to an MPP Warehouse (Snowflake)
• De-standardize our data model and create clean, rich data tables
• Find a secure workflow that gives data scientists the ability to create production machine learning models with minimal engineering assistance

Our company created a data access scorecard to track our progress, and within six months we were able to make tremendous progress on all of these fronts! (Special thanks to Jason, Yuriy, Jin and Jess for making this happen)

Prioritizing the Projects with Highest Impact

When I first met my colleagues, many of them were very excited and came to us with many suggestions on where data science can make an impact. With limited resources at hand and data science being a new function, we had to be prudent and deliberate in selecting the first projects to tackle, out of many good ones. To help prioritize various projects, we considered several dimensions:

What’s the impact of the project?

While impact can come in many forms, we at VTS focused on whether a project either:

• Directly delivers a solution that our customers want,
• Significantly expedites the process in which VTS delivers solutions to our customers, or
• Is a foundational initiative that unlocks many valuable workstreams.

What’s the feasibility of the project today?

Given the state of our data today, we had to be realistic about what data science projects we can deliver. In evaluating our potential projects, we determined whether:

• We were collecting enough data to generate reliable outcomes, and if so
• We were collecting data of sufficient quality. In defining quality, we evaluated whether our data was complete, accurate, timely, unique and consistent

How much data science effort?

As data science is a discipline that requires in-depth research, our development cycle is naturally longer than some other functions. At the same time, we want to develop trust among our stakeholders that we are a reliable team that regularly contributes to business outcomes. As a result, we examined our potential projects and prioritized those that can be completed within a 6 month outcome. This ensured that we had early wins and contributions that give us credibility to focus on even longer-term, step-function types of initiatives.

With this framework, we were able to identify a few initiatives that satisfied all three impact criteria, that utilized data that was high quality and proprietary, and that could be delivered within the first few quarters. This ensured that our team was set up for future success.

Hiring the right founding members

Recruiting the right talent is clearly one of the most important aspects of building an effective data science team. To find the talent that will thrive in VTS in its current state, we identified several key traits:

• Can wear multiple hats: Given that our data foundations and product development processes are not fully mature, we looked for “generalist” talent that could function in various roles, from cleaning and transforming our data (typically a data engineering role), to building ad-hoc analyses and dashboards to drive decision making (typically an analyst role), to helping guide roadmaps for their product squad (typically a product manager role)
• Is solutions-oriented: We sought for data scientists with an eye towards building practical solutions to our problems. We wanted teammates that would iterate towards finding a “good enough” product as quickly as possible, rather than spending time finding the perfect solution
• Has strong communication skills: This trait is especially important for us at VTS, as data scientists not only collaborate closely with our internal stakeholders, but also share findings and prototypes with our clients in the commercial real estate industry. Being able to explain concepts and demonstrate our value in a way that real estate leaders can understand is crucial to our success
• Possesses industry knowledge: While this is not a requirement, we value data scientists with real estate backgrounds. The real estate industry not only has extensive domain knowledge, but also has many players that interact with each other. Having that background knowledge gives us an advantage in user empathy, allowing us to create products and solutions that suit the needs of the industry

Fifteen months into this journey, we have hired multiple data scientists, all of whom come from the real estate world with extensive data science backgrounds! They have been the backbone of our team, and we look forward to transforming the industry with our data-driven products.

Fostering the right mindset for the team

At the beginning of the year, we, as a company, aligned on four attributes that exemplified our mindset: resilience, teamwork, action and boldness. We took those attributes further and outlined what it meant for data scientists to exemplify the VTS mindset:

• Resilience: Commercial real estate is an area where data quality is often not ideal. In spite of that, we find a way to generate value, through our scrappiness and creativity
• Teamwork: We engage deeply with our product, engineering, design and sales counterpart to build solutions that directly improve outcomes
• Action: We strive to be proactive, rather than reactive thought partners. We find opportunities for us to be impactful, instead of waiting for opportunities to come to us
• Boldness: We challenge the notion that commercial real estate cannot be quantified with data science. Our goal is to uplevel the science in decision making across the industry (as opposed to just using art)

Having this mindset at heart empowers our data scientists to be fearless leaders across the company, and to push the boundaries of applying our data to our products and to our commercial real estate customers.

What’s next?

While our team now consists of talented data scientists, who have delivered multiple productionized models, and are operating on sound machine learning infrastructure, we see a couple of areas that we can improve on:

• Collaborating deeply with product squads to define the user problem our models can solve will help us drive more value. In other words, it’s not enough for us to predict the future; we must clarify how our predictions help our real estate stakeholders drive better outcomes. This involves conducting user research, designing prototypes, and analyzing user behavior early on in the development process.
• Explaining our outputs in non-technical, real estate terms will help build trust with our product partners, as well as our end users, leading to higher adoption of our product and higher quality feedback
• As we build more complex predictive capabilities, so does the need for us to have robust infrastructure to surface these capabilities in real time. This involves working in tandem with our data platform and foundation houses to realize this vision.

We are hiring!

We have made tremendous progress so far, but we are just getting started. We are hiring additional data scientists and data engineers to join our team in 2022 — come help us bring commercial real estate into the information age!