Back to Case Studies

Improve user experience by reducing fraud & scams

  • Industry
    Real Estate
  • Location
    Hong Kong
  • Company Size

Client

Yondir was a real estate startup backed by a Hong Kong innovation lab investment. At its peak, it operated the largest and most comprehensive rental listing website in Hong Kong.

Objective

Improve user experience by reducing fraud & scams

As the only real estate listing aggregator in Hong Kong, at any given time Yondir had over 100,000 rental listings from half a dozen direct listing and agency websites. While this alone saved Yondir users hours of time for a search, it did not solve another problem for their users - namely the wide number of fraudulent “bait and switch” listings, unreliable real estate agents, and scams prevalent in the local market.

Client Challenges

Reboot AI Challenges

Solution

In order to detect fraud and repeat bad actors, the Reboot AI and Yondir team devised a proprietary strategy to triangulate multiple data sources and use machine learning, image recognition, and other techniques to then weed out culprits.

Triangulating multiple data sources first required establishing a rigorous data ingestion, cleaning, preparation, and standardization pipeline. Once processed and triangulated, probabilistic machine learning techniques could combine duplicates of the same listing posted on multiple sites, and then model a “source of truth” based on historical and current listings about which listings represented actual rental opportunities and which were merely “bait and switch” attempts for agents to win business.

As a further step, machine learning and AI valued each rental listing based on that apartments historical listing information, current aggregated listing information, images, and market conditions.

Results

Yondir saw a drastic reduction in fraudulent listings at a time when many other listing websites and real estate agencies drew public warnings for publishing so many scams and low quality information. Yondir users were able to avoid fraud and scams, find the best agent to work with for a given listing, and search for the best relative deals on the market with an automated impartial valuation.

  • 100k+ daily listings processed by data triangulation pipeline
  • 330% increase in user experience quality over alternative listing sites
  • 65% reduction in duplicates, frauds, and scams
  • 90% accuracy in rental valuation
Back to Case Studies