How Wahi Built a Revenue-driving Realtor Recommendation Engine with Vector FastLane
June 19, 2023
June 19, 2023
Putting more control in the hands of real estate buyers and sellers by giving them better information: that’s the goal of Toronto-based Wahi, a digital real estate platform.
“The problem users had before was they needed to have as much relevant information and insight as possible at their fingertips to make their own informed decisions. We want to give consumers agent-level insights so that they can be in control of the decision-making process and look out for their own best interests.”
The company had already built several tools to this end, including an AI-powered Home Value Estimator and an online marketplace where users can auction or list their homes. But one feature idea had the potential to be the startup’s key, unique offering: an AI-powered realtor recommendation system that would automatically gauge local realtor quality and surface the top 10 per cent to customers. The tool could also be a key element of Wahi’s monetization strategy, generating revenue through commission-sharing with the realtors who got leads through the platform and proceeded to close sales. But developing such a feature is a tall order for a startup. To accelerate their efforts, Wahi joined Vector’s FastLane program, where they were able to collaborate with Vector’s AI Engineering team on industry projects and get exposure to the latest AI techniques.
Vector’s FastLane is designed to help small and medium companies in Canada accelerate AI commercialization in their business. Vector packaged AI industry projects, technical bootcamps, and professional development courses together into one program specifically for companies with clear plans for AI but limited resources to carry them out. For these companies, FastLane provides a catalyst to help turn their AI use case ideas into a reality.
For Wahi Head of Data Science Eman Nejad, one specific program offering caught his eye. “I found FastLane had an industry project for AI recommendation systems, and it sparked in my mind that maybe we should join them to focus on our specific problem statement,” he said.
The problem was how to accurately evaluate realtor quality and then recommend the best to users in any location and for any type of home. The AI Recommendation Systems project featured emerging techniques for developing AI models that can identify and suggest relevant items to users. Like all of Vector’s industry projects, it also provided the opportunity to work closely with Vector’s AI Engineering and Applied AI Projects teams as well as its community of researchers to discuss which techniques may work best and then to experiment with them.
“Because we have a small data science team, we lack the resources to dig into the research. So, the idea of getting insight from the instructors and mentors was very helpful.”
Head of Data Science, Wahi
During the six-month project, the Wahi team, with training from Vector experts, created a system that could evaluate a key dimension of realtor quality: the level of care and attention a realtor devotes to their work. Embedded in realtor listings, the Wahi team believed, were clues that served as reliable proxies for realtor commitment and attention to detail. These proxies include the quality of writing, the correctness of the home descriptions, and the completeness of the listing. A natural language processing model could help gauge these aspects of the text used in home listings.
“The mentors pushed us through some specific embedded systems to encode these texts and differentiate good from bad quality,” Nejad said. “This was the problematic block that Vector helped us to overcome. We worked with the researchers to get a clearer idea about the data, the problem, which direction to go, and whether academic paper-based research confirms this. The machine learning model we found helped us determine the best quality descriptions.”
The team compared the recommendations from their matching algorithm – which could now consider listing quality along with historical transaction data like sale prices and average time-on-market – with real client reviews on common realtor review sites. The results were consistent.
Wahi had created the AI secret sauce that could serve and inform real estate buyers and sellers in a novel way.
Katchen recounts how he felt seeing this product come to life with Vector: “In the middle of the project, Eman told me, ‘I’m confident it works, I’m confident this shows a better realtor.’ I attended the final FastLane meeting and when I realized, ok, this thing is real, it isn’t just a lab experiment, I started to get excited. I said, ‘What if this could be the heart of the marketplace?’ So, we started building all the other elements around it.”
Now, just six months after the project’s completion, Wahi is launching their realtor matching algorithm, unlocking a new revenue model, and earning the startup recognition as 2023’s Best Real Estate Innovator at the Canadian Business Innovation Awards.
“This [recommendation engine] is the heart of [our marketplace]. The heart of it is this algorithm, and the FastLane program enabled us to get it across the goal line so that we could get it to market and scale up quickly.”
“I’m proud of the model we built, and proud to be part of Vector’s FastLane program,” he continued. “What I tell people is that they’re one of the top AI institutes in the world, and we’re lucky to have them in Canada, and that Wahi is privileged to have done this FastLane program with them.” Now that Wahi has launched this new algorithm and user experience, Katchen says his aspiration is that they can empower home buyers and sellers and transform how they approach the real estate market. “As this takes off,” Katchen said, “it’ll be one additional propeller that pushes the property-tech industry in Canada forward to allow consumers to do their transactions in a digital-first way.”
Sign up for Vector’s FastLane today and collaborate with Canada’s best.