Property Intelligence Principles and the Art of Possible

Property Intelligence is a ground-breaking technology enabling Matterport customers to derive powerful insights from their spaces and use this information to save both time and money. More than 90% of all the data that has ever been created in the world has been created in the last two years¹. But despite the massive amount of data generated, 95% or more goes unused² and more than half of all project rework is caused by poor data management and miscommunication³. Property Intelligence eliminates the need for multiple site visits or automates previously lengthy or manual workflows, enabling Matterport customers to make better decisions more quickly from anywhere in the world. And it isn’t just about saving time and money but also saving the planet. By reducing the need for business travel, Matterport digital twins have the potential to greatly support our customers’ sustainability efforts. 

Inferencing, Accuracy, and Confidence

Matterport is adding capabilities to enrich digital twins with high value semantic data that Matterport’s CortexAI is able to infer from those spaces. Matterport brings a rigorous process to development and productization of this semantic data based on a set of carefully developed principles.

For example Matterport’s CortexAI is able to detect and help blur human faces that may have been captured as part of scanning a space. This is done through a neural network trained to detect human faces either on people and in photographs that may be on display in home.  Each detected occurrence of a face in the photographic imagery from a capture is an inference. Matterport carefully characterizes the accuracy of face detection to ensure we meet quality criteria for both precision and recall in face detections. Further, for each inference face detection the associated confidence score is provided for that detection to provide further context on the relative confidence between detected faces.

Matterport aims to maximize the accuracy of all inferencing over time but it must be acknowledged that inferencing is imperfect. Therefore, the next principle to call out is Editability.

Visualization, Editability, and Engagement

For important spatial data Matterport believes it critical that visualization drives understanding and that mode owners must be able to edit and personalize spatial data. It is this engagement with spatial data that helps Matterport to improve the product over time. Consider again the example of detecting faces in scanned spaces. It is important that Matterport provides model owners with a visualization for the detected faces in a space, allowing owners to confirm these detections individually or in bulk. In some cases, faces may be detected in artworks that should not be blurred, or alternately if a face was not detected it can be marked for blurring. This ability to curate the inference results gives agency to model owners to engage with this data and to make minor adjustments as they choose, and then to publish the curated set of blurs through into the model assets.  Editability improves the user experience, streamlines customer service engagements, and also helps Matterport understand how it can improve the product further.

Generalization of Principles

These principles are applied not just to face detection, but rather broadly to new insights in development and planned for release in 2023. Whether it is understanding the number of bedrooms and their dimensions, the size of kitchens, the location of plumbing or HVAC fixtures, or simply identifying faces that should be blurred, these principles guide our best practices. These powerful insights help model owners and visitors better understand Matterport spaces, making it possible to make better decisions, faster, and from anywhere.

Later this year, Matterport customers will be able to view a personalized report with data extracted from a digital twin within their scan portfolio. In just a few clicks, Matterport makes it possible to quantify, for the first time, attributes that drive the value and speed of selling a home; attributes such as room size and boundaries, room type, or detection of common household objects like appliances. Previously, this information was only available through in person inspection to complete manual measurements plus in-depth analysis of similar property transactions (also known as “comps”) in an area.

By connecting digital twins to rich property data, Matterport aims to empower our customers with unprecedented attribute visibility based on their portfolio of digital twins. As Matterport continues to expand the Property Intelligence portfolio, insights from digital twins could be used to optimize a property listing or plan renovations that have the greatest potential to increase value. Similarly, Matterport wants to help our customers to eliminate hours of manual work required to enter MLS data, to generate insurance quotes, create property comparisons, and more. 

This and more is possible because Matterport has been a leader in 3D capture technology and insights for more than a decade. While similar services might use 2D for property analysis, Matterport’s AI engine is powered by more than nearly 500 million 3D scans from across the globe. And each of our scans is multi-pass, meaning that we’re capturing details from a variety of angles, giving our users more flexibility to extra data and insights to suit their unique business needs. 

Property Intelligence is a fully automated way to access, analyze, and action aggregated insights from a space empowering our customers to compare and collaborate on any space, anywhere in the world. Offering Property Intelligence as part of our industry-leading digital twin platform marks a significant milestone in the industry and is a capability only Matterport can provide. We’re currently piloting Property Intelligence with select customers and expect to reach general availability in the second half of 2023. Customers who are interested in learning more or participating in the pilot program can contact us at [email protected].

¹Source: Mandy Mock - Vice President and GM of Desktop, Workstation and Channel Group at Intel Corporation - Autodesk University 2021 ²FMI ³Autodesk