Dilip Kumar was in a hurry. He had just one week to shift into a new house in bustling Bengaluru. The easiest way, he thought, was to meet a broker. Two days went by but the brokers couldn’t find him a place. Kumar logged on to online sites like Quikr, MagicBricks and 99acres. That, too, didn’t help.
It was a Thursday afternoon (the deadline was Sunday). At the lunch table in Kumar’s American Express office in Bengaluru, his friend mentioned NoBroker, an online platform that helps individuals find residences without the help of broker.
Kumar downloaded the app. “Soon a relationship manager called me… I searched through a list of nine houses free of cost, contacted five-six of them, and by Sunday finalised a house,” he says.
Underneath NoBroker’s search is a layer of technology powered by artificial intelligence and machine learning — with minimal, if any, human intervention.
For the 1BHK, Kumar didn’t pay a single penny. “The pictures (of the house on the app) were real… The search was better than any other online platform. Even if the broker would have found me a house, he would have taken a month’s rent as commission,” Kumar tells FactorDaily.
But, Kumar was lucky to find a house for free. If he didn’t find one, he would have to pay Rs 999 to choose from 25 more houses; Rs 1,999 to get a relationship manager to find the right house, make a selection and do the negotiation; and Rs 5,999 to find a house within 45 days on the guarantee that if NoBroker isn’t able to find him a place, it would refund the money. All these subscriptions are valid for three months.
Underneath NoBroker’s search is a layer of technology powered by artificial intelligence and machine learning — with minimal, if any, human intervention. Akhil Gupta and his cofounders Amit Kumar Agarwal and Saurabh Garg have their reasons to make humans — brokers, in particular — irrelevant with the use of technology. More on that later.
AI or nothing
First, the layer of technology.
With the rise in the gig economy, high internet speeds and smartphones, as also the uptake of services such as Uber and Ola, mobile payments, and ecommerce in general, Gupta and cofounders thought they could find a way for landlords to list houses on a online platform to seek renters without the intervention of a broker, and vice-versa.
The pitch, when they started out in 2014, was simple: dear landlords and tenants, you don’t have to worry about paying commission to the broker.
Also read: This AI pathologist could be a life-saver for India’s ailing diagnostics sector
But, as NoBroker started listing properties, Gupta realised that it was important to make search better, understand customer needs, and weed out brokers from the platform to stick to the brand promise “NoBroker”. All this wouldn’t be possible without some high-tech solutions, if you wanted to keep headcount and costs in check.
(NoBroker has raised $20 million from BEENEXT, Beenos Partners, SAIF Partners, among others — of which $17 million was raised in two rounds in 2016.)
So, Gupta decided to use Google’s artificial intelligence (AI) and machine learning platforms to sort, select and list properties better, and in turn help users find the right property.
Analysts, too, believe that machine learning has its own benefits in real estate. “You can customise and curate search, and drive very targeted results. It helps in understanding consumer needs and behaviour,” says Kiran Kumar, programme manager, Techvision, at Frost & Sullivan.
One of the biggest challenges was how to show the most relevant property. Gupta had Amazon on his mind. “They have a pretty good recommendation-and-collaborative filtering model,” he says. But there were unique challenges. “Unlike ecommerce, we don’t get to know when a property is out of stock,” he points out.
So, the team built a machine learning-based product called “property age calculator”, which figures out what is the optimal time for a property to stay vacant, depending on location, interactions with rent seekers, construction quality, and price. Depending on this calculation, the list is put in front of the customer.
The team built a machine learning-based product called “property age calculator”, which figures out what is the optimal time for a property to stay vacant, depending on location, interactions with rent seekers, construction quality, and price. Depending on this calculation, the list is put in front of the customer
Gupta also realised that every day, hundreds of people clicked on the “pay to subscribe” tab but didn’t conclude the deal. “We had to figure out the propensity of the customer to pay, for which we used machine learning. We created a log book and calculated scores… We reached out only to people whose score was close to 100,” he says.
Also read: Will AI eat the fund manager’s job in India? AccuraCap shows it will
As a result, NoBroker’s monthly property transactions have doubled to 6,000. It adds 1.3 lakh customers every month and claims it is saving brokerage of about $3 million on a business of $21 million a month. On Google Play Store, it has crossed half a million downloads.
Machine learning also allows to sort the right property and sort rooms according to images, without any human intervention. The tool is even used to weed out porn images in listings. Yes, some advertisers use porn to get attention!
“A company of our size would need 1,500 people. We work with 270, and won’t need to increase our headcount for listing and sorting even if we reach out to 25 more cities,” Gupta says.
NoBroker’s presence currently is in five cities, of which Gurgaon was launched only in February. “All three founders are Baniyas… we are extremely frugal, and now we are proud about it. After raising Series B, we didn’t expand in any city. We launched Gurgaon only after a year,” says Gupta.
Vandalised: To rise again
NoBrokers uses technology to track brokers on its platform looking for new vacant properties. A proprietary product searches the internet to do that. “If user is a broker, we will ban him from the system. A broker has a very different way of accessing the site — as an individual you will search for 1BHK or 2BHK, a broker will search for 1BHK as well as 4BHK,” says Gupta.
One day in September 2015, a group of 40 to 50 brokers came to the office and started banging the doors, and eventually broke in and thrashed the furniture. “One of us got badly injured,” says Gupta
Frost & Sullivan’s Kumar believes that figuring out a broker is the toughest part in the game. In the past, travel portals like MakeMyTrip, ClearTrip, Airbnb and others have made brokers irrelevant. “With the help of human intervention and machine learning, the same can be done in real estate,” he says.
Brokers know this and have tried strongarm tactics. When NoBroker was launched it had a rough start. It was a bootstrapped company, operating out of a small residential complex in Bengaluru’s HSR Layout. One day in September 2015, a group of 40 to 50 people came to the office and started banging the doors, and eventually broke in and thrashed the furniture. “One of us got badly injured,” says Gupta.
It was a group of brokers whose business was threatened by NoBroker. The founders complained to the police. The mob didn’t return, but they started getting threat calls. Women employees were stalked. Finally, they shifted to a commercial complex. “We still get threat calls; one came just a couple of weeks ago,” Gupta says.
But the founders are more determined now. NoBroker wants to make its customer experience even better with technology. At its call centres, Gupta wants to record and auto-transcribe all inbound and outbound calls and do a sentiment analysis of the callers. “We can gather a lot of customer understanding. If 25 people on a daily basis are asking for similar things, then we can build those products. That will help in evolving the product,” says Gupta.
Also read: The AI secret sauce: Skill sets that will make you ready for the post-AI world
FactorDaily’s journalism is produced by some of the best brains in the story-telling business. If you like our body of work – deep reportage, domain specialist write-ups, data stories, podcasts and the like – consider supporting the FactorDaily journey.