Robot real estate texts 2.0 — driving subscription sales through relevance
Schibsted title Bergens Tidende have just gone live with some of the most sophisticated robot written real estate texts United Robots have ever developed. The strategy behind the publisher’s first foray into automated content is to make it attractive enough and geographically targeted enough to drive digital subscription sales. Says BT Project Lead Jan Stian Vold: “Like all media houses, BT need to look for new revenue opportunities. Automated content is such an opportunity if it’s made relevant enough for readers.”
UPDATE Aug 21: The real estate texts are now behind a hard paywall. After the first month, these robot written texts have converted 47 readers into paying subscribers.
Bergens Tidende have worked with United Robots’ developers for some months now to fine tune the structure and content of the automated articles the real estate robot produces about house sales in this city on the west coast of Norway. The team pressed the “publish button” at the end of June, and new articles now appear on the BT Homes sales vertical every weekday. If you’re curious to check out what Bergens Tidende’s automated articles look like, do it soon. The texts are currently behind a five articles a week metered paywall, but once the big launch happens in a few weeks, this will be content exclusively available to paying subscribers. More text examples here and here.
So what development work has been done to enrich these automated real estate texts? The primary data source (from a data warehouse) is the one that describes the sale that’s taken place — the “news”. It includes street address, type of property, price, buyer, seller and geographic coordinates. To that has been added a second data source which provides additional information about the property, such as size, number of floors / what floor it’s on, whether there’s an elevator, what neighbourhood it’s in etc. All this information makes up the first couple of paragraphs in the text. The robot also has access to historic data, so compares the sales price with the most recent previous sale and calculates the increase/decrease.
The next subhead in the article covers sales in the neighbourhood, how many and at what price. Here, the robot also calculates and includes price per square meter. This section has a list of the top five most expensive homes sold in the city area (i e several neighbourhoods combined, e g Bergen West) so far this year. The final subhead is Prices in Bergen Now, which is based on a monthly update, and compares prices with the previous month as well as the same month last year, including a bar graph comparing prices with other Norwegian cities. This data is synched daily so the updates happen automatically.
The image of the property sold comes from Google Streetview (United Robots have a global agreement with Google). A new feature for BT is the frame from automatically generated Google Earth “drone” videos. These frames have the property pinpointed and labelled (see image).
Geographic targeting will drive relevance for readers. According to Jan Stian Vold the relevance the real estate content drives is twofold. One aspect is the depth of information generated. “Information about trends in the real estate market is key for anyone owning or aspiring to own a home. By combining specific information about individual sales with general trends in the local region, BT provide readers with a better overview of the market, which means we put them in a better position to make educated choices in that market.”
The other relevance piece is to do with geography. “In the first version everyone will see the same articles. However, we have a well-developed tag structure which means readers will get a good overview of their neighbourhood and street. In coming versions we’ll add geographic search and targeting. And as we put this content behind the paywall, we hope our real estate bot will contribute to recruiting new subscribers.”
For BT, deploying robots is not only about the additional content generated. Says Jan Stian Vold, “By automating information generation, we free up journalists’ time. And that can be used to dig deeper — about the real estate market, as well as on other relevant aspects of society. Robot journalism is not about replacing human journalists, but about making it possible for them to do an even better job.”
