reduced error (MAE) by 9.2% with Restb.AI´s property condition scores on their AVM (automated valuation model)


From publishing listings to helping site visitors find inspiration when remodelling their home, have spent the last few years at the forefront of innovation in the world of property portals. A recent focus has been to provide accurate home valuations to the thousands of buyers and sellers in the NZ market.

New Zealand’s real estate market has transparent and structured data by law, but were the first to see it’s potential to better serve home buyers. So, developed a proprietary valuation model to arm home buyers and sellers with extra valuable information.

Their AVM not only increases transparency and accurately evaluates homes, but it also raises traffic and leads to more visitors to their website. Always looking to break new ground, they asked themselves: ‘how can we make this better?

Laptop with landing page
Two property listings scored as excellent and average


Imagine an industry leading technology that can accurately determine the condition of a kitchen or a bathroom, just using images and in less than a second. With funding from the EU’s Horizon 2020 vision fund and a team of industry leading computer vision specialists, this is what created. knew that most AVM’s have a blind spot when it comes to home valuations: the house’s interior! When they found out about and their room condition technology, they had to test it to understand the impact it could have on their AVM’s accuracy.


Using the turnkey models, ran a pilot of over ten thousand images to determine how influential’s proprietary room condition analysis model is. All tests included images of kitchens and bathrooms and they received results based on the condition of the rooms.

“Not only has helped us with our AVM accuracy, but their tools have helped us develop new product solutions to ensure we continue disrupting the New Zealand property landscape.” 

Tom Lintern – Chief Data Scientist, 


After several months, the results came in and noticed significant improvements in the classification of the houses’ interior quality. They found that their AVM has a greater ability to account for the relative difference in a home’s interior quality in a particular area, like a postcode.

In certain regions, the AVM’s median error rate dropped by up to 9.2% and expects the accuracy to increase even further. Prices are more reflective of the house’s true value and buyers can see more clearly the difference between good and bad condition properties within the same area. is continuing to explore how different bleeding edge technologies can help them better serve their clients, users and stakeholders and are proud to call them partners!