While many businesses struggled to keep pace with the changing economics of a global pandemic, the real estate industry was booming.
The housing market reached record-breaking heights last month, with median existing-price homes rising 17.2% over the prior year. This increase in the average cost of a house was compounded by accelerated closing times, as the average house sold in 18 days, a record low. Higher home prices, declining inventory, and competitive offers have made the home-buying experience a stressful one.
Using their advanced algorithms built on the cloud, Opendoor has been able to quickly provide customers cash offers that are competitive and guaranteed. In one of the most difficult times to buy a house in America, Opendoor is leveraging big data and cloud computing to provide an effective solution.
Ian Wong, CTO and co-founder of Opendoor, joined Cindi Howson on The Data Chief to discuss how Opendoor is using data analytics and machine learning to disrupt the housing market and streamline the home buying and selling process. Read on for excerpts from their discussion.
Cindi Howson: Now, for people who are not familiar with Opendoor, tell us a little bit about it.
Ian Wong: Opendoor is the best way to buy or sell a home. If you want to sell a home, there's a lot of pain points surrounding the sale process today. It's going to take you roughly three months to sell a home, and in those three months, you're doing a lot of work. You're paying a 5 to 6% fee, and worst of all, you don't know how much you can sell your home for and how long it's going to take you to sell that home.
With Opendoor, you can simply come to our website, submit your information about the house, get an offer within the same day, and choose your move-out date. That unlocks liquidity for homeowners. On the buyer side, we've also done a lot of work to make it easier to tour, shop, and close a home. We most recently launched a cashback offer for buyers that makes it a lot more competitive in the market today.
We are trying to innovate on ways to help homeowners and home shoppers sell and buy, and doing both at the same time.
Cindi: It seems to me you're just getting started leveraging the whole buy/sell real estate ecosystem. Can you tell us about that?
Ian: If you're shopping for a home while you're looking for financing, that's one of the first data points that you need to understand how much home you can buy. We realized that was a common thing for a lot of our buyers who visit an Opendoor home or use our services to tour any of the homes on the market. So we said, "If that's a customer pain point, let's go and build Opendoor Home Loans." Now that's integrated with the buying experience.
One of the key approaches we've taken is to not be afraid of different disciplines. Let's figure out how to use that difficulty to our advantage because if it's hard for us, it's hard for everybody. Homes are getting multiple offers very quickly, so we helped buyers be more competitive by coming up with a cash offer. That is intricate from a technology, pricing, and finance perspective, but we might as well be the ones to solve the hard problems to better enable our customers.
Cindi: Most of your models would have to be operationalized because that's what drives the pricing or cashback offer, is that right?
Ian: Our algorithms run the business in a real and meaningful way. When someone submits information about their home, we take that information and combine that with a data platform that we've procured. We've ingested data from a variety of sources, we've normalized it, and made it available in real-time.
We combine their home information with our data set and we run it through our algorithms to understand what is the competitive price that we should be offering for that specific home. We have a cutting-edge, deep learning-based valuation model that, from my perspective, is the most advanced one out there. We can see that in the accuracy of the model, and that is what's driving the offers that we extend to customers.
Cindi: You're taking both first-party data, maybe what I provide you as the seller, and combining it with third-party data. Can you give us some examples of this third-party data?
Ian: All great algorithms start with great data. If you have garbage in you're going to have garbage out. Data, especially high fidelity data, is one of the key differentiators for any high-performing model.
We are so detail-oriented when it comes to data that we even created our own inspector app. We design the data interface that customers use so they can be customized to each home, and we can ask exactly what the algorithm needs to generate a good offer. That's the level of detail that we get to on the first-party side.
On the third-party side, it's a matter of being able to ingest all sorts of data. We're talking about transaction history, we're talking about public record data. But we're also talking about heterogeneous data types: geospatial, free text, images, videos. The magic sauce is: how do you take all these disparate types of data, normalize it, and make that readily available to both your product folks and your researchers so we can iterate very quickly?
Cindi: How do you measure if a model had the desired impact?
Ian: There are all sorts of failure modes for machine learning to make a dent in business. We've tried to design the platform so that it can ingest new data sources quickly so people can ideate and test out the ideas as fast as possible. Being a data scientist or data professional is hard because you have to combine technical skills, coding, algorithms, mathematics, statistics, you name it. You have to combine that with commercial instincts.
We built our first valuation model at Opendoor in the first month, and that was powering some of our initial offers. We've had to improve that a lot over time, but at least there's a baseline from which we're improving.
Get more insights from modern data and analytics leaders like Ian by subscribing to The Data Chief today.