redfin - data, not enough science
Redfin is an online real estate search site like Trulia and Zillow. Unlike Trulia and Zillow, Redfin is also a real estate brokerage with a network of agents around the US. Trulia and Zillow tend to see agents as their customers and home buyers as their product (that they sell to agents). Redfin's model moves the home buyer from product to customer and the agent from customer to partner. Direct contact with agents and home buyers gives them a steady stream of data about actual homes and home transactions. Redfin home search has two features that are making use of that data.
Agents tour a lot of homes and make notes about what they see. Their training and experience helps them notice the most important details. Since May of 2011, Redfin has been publishing some of these notes in their home search products as 'Agent Tour Insights'. The notes are typically well-composed and of what I have seen manage to avoid the lame tropes that are associated with real estate listings.
Similarly, agents are involved in a lot of successful and unsuccessful bids on homes. Redfin has just announced the launch of 'Offer Insights' - semi-structured details of bids and bidding processes recorded by their network of agents. They present the war stories in aggregate as 'what it takes to win an offer' in a neighborhood.
New kinds of data put into the hands of consumers smells of innovation to me and I applaud Redfin's work on both of these features. Home buyers are voracious for data and these kind of human-language field notes are a welcome addition to the slew of purely numeric information on a typical home search results page.
Redfin concludes their Offer Insights launch announcement with:
"This is why Offer Insights isn’t just a feature, it’s a new way for Redfin — and perhaps other companies that combine technology with real-world products or services — to be: big and small, juicy and dry, national and local, data-driven and personal, computer and human."
I agree that information presented as text is more 'human' that charts and graphs, but it is not clear to me that these features do that much to humanize the process of home search. A typical Redfin (or Zillow or Trulia) results page is very long and very detailed. Text is a welcome change from yet another chart, it is all too easy for an individual to draw incorrect conclusions by trying to extract a trend from snippets of human language. I think this is a place where coldly inhuman computers could actually HELP to humanize the presentation of this data.
For example, Redfin could calculate a single 'volatility index' (taking into account numeric items like number of bidders and fuzzy contributors like natural language processing on agent free-text descriptions) that is comparable between places and trend-able over time. Some home hunters will want to exclude areas or housing types that are likely to get them involved in a long and difficult bidding process or one where they are likely to be forced to move well above their opening price to be successful. Other buyers are willing to tolerate this kind of process, but could use the index to set their expectations.
It is good to have more data on the table, but it's better to have data
that is easy to act on. The techniques for really humanizing the blizzard of home buying data out there and trends in home search will favor those that apply them. The needs of SEO incent Redfin to present more original text on their pages, but the needs of users should lead them to boil that text into usable information. There is a reason why data scientist is one of the in-demand jobs of our times. Data is good. Data + science is better.