AI-enabled portal UrbanZoom aims to improve housing market data transparency

Mon, Mar 12, 2018

Lee Meixian


TECHPRENEUR Michael Cho believes that the man in the street should have equal access to housing market intelligence as the real estate agent - and for free.

That was the motivation that drove him to create UrbanZoom - an online artificial intelligence (AI) enabled research portal that includes an auto-valuation tool and other useful transaction data designed to be user-friendly for home buyers and sellers in Singapore.

UrbanZoom is decidedly not a listing portal, since that space is over-saturated, with even the likes of Carousell Pro and Facebook Marketplace having entered the space. This makes it difficult to innovate, value-add or compete on that front, he says.

Rather, one of the main things that UrbanZoom focuses on is its AI engine that automatically spins up a price prediction for a specific home based on its address by relying on more than two decades worth of transactional data, and other macro and micro input variables.

In fact, so confident is the engine in the accuracy of its predictions that it also includes a confidence level with each prediction. The median error rate for HDB and condominium homes is 2.4 per cent - which is "considered close to the theoretical limit because we are valuing the homes without actually going to see them", Mr Cho says.

The error rate would be higher, however, for less frequent transactions such as penthouses. It is obtained by UrbanZoom first creating a prediction model based on earlier data, and then using it to make predictions for more recent transactions and measuring the discrepancy compared to the actual transaction prices, he explains.

Besides auto-valuation, the site will soon also include "big data" sourcing capabilities that help property buyers and sellers answer questions like how long their neighbours took to sell their units, which units in a condominium are facing South for feng shui purposes, and which is the best month in a year to put their homes on the market.

His team is also working on geo-tagging news articles to homes, so that property buyers can find out if gruesome accidents have occurred in a particular unit before, and the frequency of traffic accidents in a particular area. Other upcoming features include an accessibility score that ranks every location in Singapore on convenience, as well as an en-bloc potential prediction engine.

The portal is the result of six months of data mining and combining various data sources with advanced machine learning techniques. Already in its two-month private beta period, it has chalked up more than S$5 trillion worth of property searches.

Asked if he thinks tools such as his signal that valuers and agents are no longer needed for property transactions, Mr Cho's reply implies "yes" for valuers and "no" for agents.

Home property valuation has become a thorny issue since the launch of the new HDB online resale portal this year, which the industry read as a disruptive move by the government to render HDB valuation firms redundant.

Meanwhile, the legal battle between StreetSine Singapore, the Singapore Institute of Surveyors and Valuers, and various property consultancies rages on, after StreetSine alleged "conspiracies" among the defendants to injure its business and reputation by intentionally misrepresenting to the public and financial institutions that its valuations are "not considered valuations" according to SISV standards.

The defendants are finalising their defences and the case is moving into another round of submission of documentary evidence. The Court is expected to assign a trial judge in April.

Mr Cho says: "I won't say valuers are redundant, but I do think that auto-valuation has a big benefit because it is real-time." That has implications, for instance, when banks are competing among themselves to issue mortgage products.

"If the fellow is in your office already, you can straightaway key in his apartment information and generate a fairly accurate valuation and offer him his mortgage."

Asked if human and AI valuers still arrive at similar numbers at the end of the day, he says it's "apple and orange" and not comparable.

"I'd say that there is still use for these guys (human valuers) for a fair amount of time, but I also feel that it's a matter of time before they go. While the auto-valuer is figuring its way out, we will still need human valuers. I don't want to speak like a dystopian geek, but having done AI and machine learning these past two years or so, my opinion is that people in Singapore are very under-equipped to deal with the new world."

The story is slightly different for real estate agents in his opinion. In fact, Mr Cho went to lengths to understand the psyche of an agent - by becoming a licensed PropNex agent himself. What he found was that data is a poor substitute for agents because the emotional roller-coaster experience of buying or selling a home, probably one of the biggest transaction anyone would make in their lives, is still best counselled by an agent.

What he feels indignant about however is the asymmetry in information between the agent and customer - for instance, when an agent tells a customer that this is the highest price his or her home can fetch when it may not be true.

"I feel that there's a lot that can be done to make the transaction a lot more transparent . . . Sometimes, it's also in the agent's interest to not show too many things. So the angle for UrbanZoom is to be the best research tool, a special type of search engine for property-related queries for home buyers, sellers, tenants, landlords, agents, anyone."

It was also his experience standing in the shoes of a property agent that gave him the idea for the business model that UrbanZoom can adopt.

He realised that agents spend much more money marketing themselves through flyer distribution and cold-calling than marketing actual properties on listing portals, because lead generation - the seeking out of their next customer - is crucial to their work.

"So the business model is that we have this free research tool that ideally draws eyeballs of people interested in buying or selling homes. Guess what? These are the eyeballs that agents want when they distribute flyers and do cold-calling."

In other words, UrbanZoom can become a lead generation tool for agents. "If we have all the eyeballs, the agencies will come knocking on our door. That's why in my view right now, what we are really focused on as a first step is to ensure that we make a damn good research tool that is free and something that people will use alongside all these listing portals.

"Unlike listing portals which are supply-driven, this will be demand-driven. I'm giving the end consumer a useful tool so that they can do their research, and these are the people in demand of service."

Somewhere down the road, UrbanZoom plans to add a function allowing users to contact an agent should they be interested in buying or selling a unit in a certain area.

The company now has an eight-strong team, including data scientists, software developers, product designers and real estate professionals. It has received its first seed round of funding from angel and institutional investors.

Mr Cho, who quit his investment career with an Abu Dhabi-based investment firm after a near-death experience in a plane crash, said he stepped into the uncertain world of entrepreneurship because the close shave became a stark reminder to him of what is really important.

"Almost everything - all external expectations, all pride, all fear of embarrassment or failure - these things just fall away in the face of death, leaving only what is truly important," he says. And using technology to create something that helps people is where he thinks his calling might be for now.