At Inman Connect (ICNY) 2020, relevant real estate players gathered to discuss the future of the industry. I got the chance to attend the latest edition in NYC.
Since we know machine learning is driving innovation across every industry, real estate included, I took the time to summarize my experience as an attendee of ICNY by highlighting key findings in this blog post. These insights are based on my personal interpretation of the sessions, booth visits, and conversations with industry players (real estate agents, proptech workers, investors, and other conference participants).
My findings are divided into two sections:
- The real estate landscape: a summary of how real estate technology is reshaping the industry.
- Machine learning & data science opportunities: an overview of a few use cases that I discussed and validated with conference participants.
Let’s get started.
1. The real estate landscape
iBuyers: what and how
It’s no longer news that the real estate industry experienced a major revolution over the last few years with the involvement of iBuyers. In short, iBuyers digitize as many housing transactions as possible. Players like Zillow, Opendoor and Redfin offer buyers, sellers, and renters the option to find an agent or search, distribute, finance, and even close a transaction online.
According to statistics shared by real estate tech strategist Mike del Prete, transactions involving iBuyers doubled in 2019 (60k), constituting over USD 8.7 billion purchase value. However, despite this growth, the main similarity among these companies is that they are “sustainably unprofitable”, as Mike pointed out in his January 2019 talk. The overwhelming majority of costs are related to advertising to reach and acquire as many customers as possible, all while competing for market share.
One aspect that stood out during the conference was the interesting relationship between iBuyers and traditional agents. The first group, referred to as the disruptors (or the unproven), currently provides a fair amount of leads to agents, who benefit from the platforms in addition to other channels (personal networking, ads, SEO, etc). However, it’s been widely commented - in several presentations and personal talks - that the quality of these leads has decreased lately. This reinforces the idea that iBuyers are more heavily relying on a closer network of partners, which provides a heightened threat to agents who lack strong relationships and are heavily dependent on said disruptors for leads. In other words, I’d say that the agent <> iBuyers relationship is one of frenemies. It will be interesting to see how it might evolve over the next couple of years.
Other relevant stats on iBuyers include the following:
- iBuyers’ transaction sweet spot is between 260-270k.
- The average market discount obtained via iBuyers is 1.4%.
- The top iBuyer states are Phoenix, Atlanta (doubled in 2019), Dallas, and Houston.
How the incumbent (or proven) players are reacting
In most industries, well-established players develop and integrate technologies to mitigate market risks and meet customer demands. At ICNY, some large players revealed recent developments and how they expect these to increase their own value propositions.
Adi Pavlovic, Director of Innovation at Keller Williams, was a keynote speaker. He outlined their mission to provide technology that delivers higher value to agents. To this end, they recently released a consumer brand app that targets large agents’ needs including branded national search, lifestyle search, and consumer end-to-end transaction capabilities. Keller Williams’s objective is to provide agents with the technical means to compete with large disruptors.
Another important player, Re/Max, presented multiple times throughout the conference. They covered several topics, including marketing, lead generation, and best tech products, among others. In examining their reaction to the market, the most relevant thing to note perhaps is Re/Max’s recent acquisition of first.io, a data driven platform that allows agents to leverage their personal network - as a database - in a more intelligent way using predictive purchase intent. Aside from their own internal improvements (i.e. adding closed captioning to videos), Re/Max saw an opportunity to provide better tools to their agents by acquiring outside tech solutions, making sales more efficient in the process. This is Re/Max’s second tech acquisition in the last two years, following the Booj acquisition in March, 2018.
There’s also a lot going on around search, where the mission is to connect the consumer with a professional in the quickest and most fluid way possible. Key players such as Realtor.com (offering unaffiliated search, which means it’s not tied to an agent) are placing bets by focusing on making online search a positive experience for the user. Refining search capabilities involves using smart filtering, and relevant keyword matching, while at the same time leveraging image search and providing virtual reality tour capabilities.
The main takeaway is that well-established players are developing technology that seeks to improve the buyer/seller experience by enhancing agents’ capabilities and shortening transaction times. This comes from the strong belief on the part of a vast portion of the industry that there’s no technology that would ever fully replace the personal touch that an agent can provide when one is purchasing or selling a house.
Relevant proptech platforms
Many tech products focused on real estate were featured during the conference either on stage, at partner showcases, or in Startup Alley. As an outsider, I’d like to highlight the products that caught my attention and seem to have become increasingly relevant:
HomeLight: This San Francisco based company aims to simplify the process of buying or selling a home by helping the end user achieve the best possible outcome, either by connecting them with the best agent or getting them the best instant offers for their homes. On stage, Drew Uher (CEO & founder of HomeLight) commented on his belief that agents are still essential to any transaction, insisting on how the approach still needs to be more like the entire App Store rather than a single app.
Returning to the product, HomeLight promises to simplify the experience for home buyers and sellers, and is backed by great software tools for agents. These tools help agents find relevant leads (no upfront cost, 25% commission) create custom profiles to highlight their skills, and include a unique messaging platform and financial products such as the recently launched Cash Close. All this pushes the agents to rely solely on the platform. They claim to have served over 365k customers since 2012.
Qualia: A SaaS platform for digitizing real estate deals, including connecting interested parties and automating the close. Qualia’s quick success has demonstrated (at least in part) that there is value in taking the hassle out of the closing process in Real Estate transactions. Multiple stakeholders may benefit from the platform: agents can provide stress-free closing options to their customers online; lenders can easily connect and transact with a network of title agents; vendors can identify leads and transact in the marketplace; and consumers can finally review the closing process online, just as they might do with an Amazon.com purchase. It was surprising to see how Qualia aims to satisfy all stakeholders in the process of providing a seamless experience online for everyone involved.
BoomTown: In the battle for a real estate CRM, BoomTown (BT) appears to be one of the leaders. The product aims to provide an all-in-one lead generation, lead management, and marketing automation suite for agents, brokers, and lenders. During the conference, they hosted a Lunch N’ Learn event where they featured stories of top-performing agents, their strategies, and how they leverage BT! capabilities. Since its launch in 2006, its customer base has grown to over 40,000.
CoreLogic: a company founded in 1991 that provides consumer, financial, and property information, as well as analytics and services for real estate professionals on up to government agents. This data could be beneficial for different use cases such as lead generation, home valuation, risk management, and property insurance.
Matterport: Turns a physical location into a 3D model that is easy to tour in the digital world. In other words, you can “build a space’s digital twin” using compatible cameras and Matterport’s platform. Matterport was one of the shiniest exhibitors at ICNY, showcasing how easily one could build a 3D model of a property using their technology.
The debate over automation
There’s a lot going on with respect to the level of automation the real estate industry could achieve.
On the one hand, the real estate industry is capable of reducing paperwork, leveraging data, and becoming more efficient as a whole. The increased use of e-signatures, transactional technology, online demand generation, search capabilities, the Internet’s massive market, and instant offers together provide a great opportunity for this industry to provide the customer with a better experience in real time. Natalia Karayaneva, CEO of a proptech company using blockchain technology, was featured by Forbes in an article on how venture capital is betting that the real estate industry will become more tech-oriented. This definitely complements Mike del Petre’s onstage comment that real estate is moving slowly, but has never moved faster.
On the other hand, real estate brokers still agree that the transaction process will probably never be fully automated. Personal advisory services and trust are said to be key to all transactions, which speaks to the importance of people being involved throughout the process. Although not a replacement for human participation, the extent to which automation will take root is still hotly debated. What percentage of transactions will real estate professionals fulfill versus iBuyers and/or automation software? That’s the big question.
The main conclusion we can draw is that the transactional process, from home discovery to closing, is being revolutionized by technology trends. This is already changing how we attend to the next generation of potential buyers.
2. Opportunities for machine learning & data science in real estate
There is already a fair number of prop-tech companies that are vocal about how they are leveraging artificial intelligence in their existing processes. Home valuation algorithms such as Zillow’s Zestimate and Opendoor’s pricing model as well as Redfin’s smart Comparative Market Analysis are solid examples of how to gain a competitive edge through the development of proprietary, data-fueled systems.
After several conversations with industry players (purveyors of tech platforms, agents, and technology experts), I came up with a list of relevant opportunities for implementing machine learning and data science in real estate that I’ll share with you below.
Making the match
Let’s say you already have a buyer and you know their interests and relevant information. You now might need to match them with the most suitable agent, the mortgage with the most appropriate terms, and even find the best home for them. Depending on where you’re at in the process, you would like to begin connecting at least some of these dots.
Instead of doing all this matching manually, you could do this with a smart system. Such a system would leverage the historical information you already possess on successful transactions along with publicly available information, in order to provide the best possible outcome for your buyer and maximize the possibility of a transaction.
Machine learning is well-suited for generating predictive algorithms of this kind since they rely on a large amount of data, examine it for strong correlations, perform clustering, and generally provide the most adequate response for a specific request.
Business-wise, a matching algorithm with these characteristics could save countless hours of manual matching and rule-based logic design, and maximize the chance of a successful transaction. Companies like HomeLight, Rocket Homes, and really every iBuyer could benefit from a predictive solution of this type.
Enhancing search capabilities
During ICNY, the user’s search experience was highlighted as a critical part of the digital journey of the buyer. Computer vision, natural language processing (NLP), and predictive analytics techniques could be used to enhance the current keyword-based search experience. A couple of examples include the following:
- Incorporating image search capabilities: extract information from pictures of the property utilizing object detection and image classification techniques, to be used in search matching.
- Recommendation-engine-powered rankings: search results could be ranked according to the likelihood that the specific user will interact with the results, based on previous searches, profile characteristics, and contextual information.
- Search intent matching: enhance the user experience by adding the ability to write (or dictate) their home preference(s) instead of manually filtering the results. It may be very wise to incorporate such a feature given the rise of home assistants.
- Visual search: perform a search based purely on images of homes. This would enhance the search experience or complement the keyword search and produce more accurate and useful results.
Anticipate the likelihood to sell
One key statistic delivered on stage was that some 5.5 million homes are projected to change hands in the US in 2020. This fact, mind-blowing on its own, should be considered along with the prediction that there will be a supply shortage given this year’s demand.
Although home construction numbers are beyond expectations, there’s still a short-term issue with meeting the demand forecasted. The biggest opportunity here then is to identify homeowners who are more likely to sell and estimate their selling price using publicly available data.
Outside of a learning and evolving algorithm, big data capabilities could be used to find correlations between past sellers and current homeowners, in order to help predict the likelihood that a given owner is willing to sell.
Such an investigation could potentially provide information on the appropriate selling price point and the interest of possible future owners so as to increase the chances of successful outreaches.
Other use cases
- Lead classification: Based on web-based actions executed by a user, understand where they are at in the customer journey and gather accurate information to move them to the next stage in the funnel.
- Risk assessment: Multiple risks should be considered when assessing a real estate transaction. Forecasting models powered by machine learning could complement a traditional risk analysis approach well, especially given their multiple data source analysis capabilities.
- Home valuation: A classic, yet constantly evolving, machine learning task is to set the price of a house based on MLS and alternative data (e.g. satellite imagery). The big iBuyers are all betting heavily on this since it’s key to their business model. An example of the level of detail obtainable is that of how Opendoor analyzed the impact of busy roads in their property valuation model.
My experience at ICNY was super worthwhile. I met with many interesting people, learned from the best in the industry, and got closer to understanding the inborn drivers of one of the most important industries around. It’s now clear to me how technology, especially iBuying, is a major force for change affecting everyone, from buyers to investors.
From a machine learning standpoint, it’s clear to me that quite a few players are starting to move in this direction. There are already players with very sophisticated teams working on it (it’s part of their core competency), while others are slowly adopting these skills internally, driven by impactful use cases.
What’s undeniable is that there is more real estate data available than ever before. This provides an opportunity for certain companies to benefit from it in an intelligent way, all while enhancing the customer experience and improving internal efficiencies.
Like what you read?
Subscribe to our newsletter and get updates on Deep Learning, NLP, Computer Vision & Python.