For many people, the process of buying and selling a home will undoubtedly be the most difficult decisions they will make in their lifetime. Is the price you’re paying for your home fair? Is the price you’re selling your home for an adequate sale price? For a long time, realtors have been oracles with the answers to those questions, but times are changing. Today, Opendoor, a data-driven real estate start-up, is putting data to work to answer those questions for you.
“Opendoor from a technical perspective, is the hands down single most fascinating technical challenge I’ve ever come across. If you look at the questions that we have to answer, they’re just fascinating and you have to answer them in the right ways. Otherwise they just don’t really work. You think about the prediction problem, you think about the optimization, probably you think about the portfolio optimization problem. There’s so many different pieces to the problem and you just have to do it.”
Kushal Chakrabarti is the VP of Research and Data Science at Opendoor, a company that is reimagining the way homes are bought and sold by moving the process online and empowering buyers and sellers to make informed decisions by taking an algorithmic approach and removing the ambiguous nature of the home buying process. On this episode of IT Visionaries, Kushal explains some of the processes Opendoor uses to help make home offers, including the importance of clean and trustworthy data. Plus, Kushal opens up on his personal journey, including how he got into data science and some of the trends he sees in the A.I. and machine learning space.
Main Takeaways
- This is Trendy: There are two popular trends happening in A.I. and machine learning. The first is the democratization of tools, which in return is making it easier for data scientists to quickly test and measure their models. The second is the abundance of third party data that is readily available, which is making it more difficult for data scientists to know which data they are getting is trustworthy data, and which is not.
- Trust in Numbers: All great algorithms start with great data, but having a high fidelity of data is one of the key differentiators for any high-performing model. When you’re mixing first-party data with third-party data, be intentional about how you create strategic data models that fit your business.
- Asking the Right Questions: As a data scientist you must be asking the right questions otherwise, your models will have a tendency bias introduced to them. When models have bias, you might get an answer that is close to your hypothesis, but it won’t be the correct answer.
For a more in-depth look at this episode, check out the article below.
Article
For many people, the process of buying and selling a home will undoubtedly be the most difficult decisions they will make in their lifetime. Is the price you’re paying for your home fair? Is the price you’re selling your home for an adequate sale price? For a long time, realtors have been oracles with the answers to those questions, but times are changing. Today, Opendoor, a data-driven real estate start-up, is putting data to work to answer those questions for you.
“Opendoor from a technical perspective, is the hands down single most fascinating technical challenge I’ve ever come across. If you look at the questions that we have to answer, they’re just fascinating and you have to answer them in the right ways. Otherwise they just don’t really work. You think about the prediction problem, you think about the optimization, probably you think about the portfolio optimization problem. There’s so many different pieces to the problem and you just have to do it.”
Kushal Chakrabarti is the VP of Research and Data Science at Opendoor, a company that is reimagining the way homes are bought and sold by moving the process online and empowering buyers and sellers to make informed decisions by taking an algorithmic approach and removing the ambiguous nature of the home buying process. On this episode of IT Visionaries, Kushal explains some of the processes Opendoor uses to help make home offers, including the importance of clean and trustworthy data. Plus, Kushal opens up on his personal journey, including how he got into data science and some of the trends he sees in the A.I. and machine learning space.
Launched in 2014, Opendoor set out on a mission to accomplish one goal: simplify the way consumers buy and sell homes by offering fair prices. But there was a catch: those offers would not only be dictated by what the housing market was saying a home was worth, but also by what Opendoor’s complex algorithm suggested was a fair price. As a result, Opendoor needs a ton of data.
“This is the hands down single most fascinating technical challenge I’ve ever come across,” Chakrabarti said. “What do you do when the data just doesn’t exist? When you just can’t get the data?”
To gather its data, Opendoor has to account for various factors, including simple things such as the square footage of the house, number of bedrooms and bathrooms, whether the home has a pool, etc. But the more challenging areas include how the company decides if a home is priced competitively based on market, neighborhood, and even the street it sits on. This means the data that Opendoor uses to make these determinations has to be trustworthy.
“There’s lots of examples of bad data, unclean data, and dirty data,” Chakrabarti said. “If we calculate a price based on [bad data], you get the wrong number. Now all of a sudden, we’ve insulted you and this is a bad experience for you because this is the most important transaction in your life and it’s almost certainly the largest financial asset in your life. I think there’s always a balance of how much do we ask you to put in and how much do we use for external data.”
That balancing act of not only gathering your own first party data, but trusting where your third-party datasets come from is an important part of the equation for Chakrabarti in how Opendoor’s algorithm works. But also said one of the important aspects to follow when it comes to data science is the abundance of tools that are now available for data scientists, making it easier to spin up models and get results more quickly.
“I think if you look at the history of data science, machine learning, there’s been two broad trends over the last couple of decades,” Chakrabarti said. “One is that there’s just a broad democratization of tools and two is this massive explosion of data, which means that in most of these situations, you just throw enough data to one of these black box learning tools, and you may not get the perfect answer, but you’ll get a good enough answer in most cases.”
But above all, when using data to draw conclusions it all comes back to one main thing, data scientists must still be asking the proper questions to get the best results.
“There’s this notion of quality and quantitative research,” he said. “ You really need to bring together qualitative research and quantitative research to answer questions in the right way, but before that, you also have to be asking the right questions to even start with.”
To hear more about Chakrabarti’s career journey and the work that he is doing at Opendoor, check out the full episode of IT Visionaries!
To hear the entire discussion, tune into IT Visionaries here.
