> Before starting this company, my co-founder and I had tried to invest in homes.
Were you guys successful?
Were you guys successful?
(This will sound like we are very lazy...) We essentially wanted to buy an affordable home that would just grow in price over the next few years without us doing anything. At the time, there were no tools to help us find homes like this, which ultimately led us to starting this company.
On the plus side, our customers can choose to buy us out after 3 years, and see 2 more years of growth after, and they wouldn't have to share that profit with anyone.
> We were inexperienced and didn't really know what to look for. Homes that required extensive rehab seemed too daunting and a turnkey property in a nice neighborhood was too expensive.
1. what in your experience are the top 3 things to look for especially for a market like Bay Area ?
2. Follow up to 1. longer term do you think the model will do well for markets outside of Bay Area like Phoenix, Vegas etc.
2. How did you overcome the cold start problem for your models ?
(A while ago there was a podcast (IIRC from NPR) where Zillow ran a contest to reduce the error % of their price prediction model and the team that won probably ended up also using factors like direction of the sunlight etc.) 3. Given your competitors use these kind of data points, will your model rely heavily on satellite imagery etc. to infer these kind of data points in the future ?
1. A few of the top and more intuitive things we found in the bay area to be indicative of an upswing include an increase in food trucks and vietnamese restaurants. As well as increases in the number of social media postings about pets.
2. Yes, we believe it will. We have backtested on a ton of different markets and been tracking our models predictions in these markets over the past year and it seems to apply for most cities where these alternative data sources are present. Obviously (and perhaps interestingly) the things that seem to drive revitalization do have some constants between cities but they do also vary a decent amount by geographic area.
3. This was somewhat tricky. Obviously some of the sources we use like home prices and sales are more readily available and have existed for a long time. Others, not so much, especially for alternative data sources. We tried to choose sources that had been around longer (around a decade was a proxy) and had historical data that could be accessed via an api or scraping. This limited the list of sources we could use but we are quite happy with the list we ended up having that met this requirement.
And yes, we had looked at that competition Zillow ran and drew some inspiration from it. We do currently use satellite and street view data and are actively adding more uses for it, although we currently do not have the sunlight measurement per property integrated. I noticed that (i think?) as a new feature for homes when you look at them on zillow which was cool!
Additionally, most of our customers still visit the properties we recommend before they buy, so these types of problems are usually spotted during inspection before the deal closes. This has filtered out any bad quality deals due to home conditions.
Hope this answered your question!
We don't pretend that we know more than others about real estate, because we are still mainly a tech company. Our expertise is in finding properties, using data, that can perform well, which has been the case both in back-testing as well as walk forward predictions.
This is why we don't handle the transaction process for our customers. Instead, we rely on partners that have more experience in these areas than we do.
To further ease people's mind, we offer the downside protection. So, if we mess up, we pay the price, not our customers.