Can this algorithm determine what a “bad” property is?
A new app, realAs, claims to be able to predict how much a home will sell for, with a margin of error of just 5%.
The app, which is said to learn as it goes and to become more accurate as people use it, is based on this core algorithm. Consumers are asked to include their views about homes they have seen, and prices are input into the system.
It has some lofty credentials as well, being cofounded by David Morrell, the first buyer’s agent in Australia, who also sits on the board with Andrew Newbold, Jeremy Press and Josh Rowe. It was developed by computer scientists at RMIT University over a three year period and claims to be more accurate than overseas counterparts Trulia and Zillow.
The Melbourne-based company claim that the predictions for 64% of sales are within 5% of the sale price, 89% are within 10% of the sale price and 99% are within 20% of the sale price.
Morrell says that the unique propreitary algorithm listings to buyers and “understands what’s a good property, it understands what’s a bad property and it understands what the property value is”.
In a much publicised claim, it also aims to stamp out underquoting completely, as consumers use the realAs app to determine the price independently of the real estate agent.
“realAs is by buyers, for buyers to research and make the right property purchase decisions. Saving buyers time, money and heartache,” said Rowe.
While they’re not giving any secrets away about the algorithm itself, which he says does prompt a discussion around the level of trust, he does note that it goes beyond past price sales.
“Traditional predictors of property prices usually only take into account what is happening in past sales,” says Rowe, criticising that they do not take into account the individual property’s features or the current listings."
This algorithm uses current market data with specific relevance to the individual property and the features that property has.
Currently in the form of an iPhone app and responsive website, Morrell believes it will be disruptive for the real estate industry where he says overquoting and underquoting is rife.
He explains that current sellers being told by an agent to take their property to auction and to expect over $2 million may look at the app and realise it’s worth 30% less.
As a knock-on effect, the seller won’t go to auction, there will be less advertising for the real estate agent and, effectively, newspapers also won’t get the spend on real estate advertising, he says.
“For the first time they’re able to see what other buyers are thinking,” Morrell explains.
However, that’s not expected to be the only effect on the market. Rowe says that there is another functionality beyond the price predictions – an open dialogue for sellers and buyers.
Currently, the majority of buyers speak to the seller “through” the real estate agent, Rowe explains. Now, they can speak “directly with sellers about the quality of properties” from quantitative questions to qualitative questions, without a filter.
Property Observer notes that in the majority of cases looked at, for instance in Sydney's west, it appears as though the agents quote and the realAs prediction are quite similar.
