Rohit Maghnani: what AI changes—and what it does not

Magic Bricks, we are the largest real estate portals in the country. The journey for different people is anything between three months to several years. We compete head on with Facebook and Google because we just get that gut feeling, we just get that sense. How are you seeing trends in data slash AI where you are engaging with the customer in order to understand where that emotion quotient is coming when he's looking to get a place for himself or himself? So I work at Magic Bricks. We are the largest real estate portals in the country. And we along with a couple of other real estate portals have been helping matchmaking in the real estate space, matching buyers with potential sellers or tenants with land loans.

Now a lot of you would have been in the journey at different stages of your life or you start at the PG accommodation, then go to a rental accommodation, then you buy a property. I guess a lot of you would have gone through the property buying journey and raises. And how long was the duration? Anyone? Three months, six months? Several years. Now this is a tough one because the journey for different people is anything between three months to several years because if you find the right house, and very often you may find the right house, the house has got sold out or it's got outpriced. The challenge for us was that in the world of data, how do we assess when is the guy really serious about making the purchase in the next 5, 10, 15, 20 days?

And when is he just casually browsing or she casually browsing? Now, our biggest competitor is actually the two companies which are the best at reading data. So the two competitors are Facebook and Google. So we compete head on with Facebook and Google because they also are advertisers for real estate sellers exactly like us. So this was extremely challenging that how do you get more data than Facebook or Google and figure out when is the intent strong for a person and therefore present him to a potential seller that this is the person who is about to buy versus this person who is going to buy but he is six months away. So we took a very different approach to solving this.

We looked at what people were doing in the real world and we figured that the real estate agent or the broker had a knack of figuring that out very well. So we interviewed with a lot of real estate agents and brokers and said, how do you figure out when the guy is about to buy? He said, I figure out when I talk to him, I figure out I just get a sense of it from what he is talking. How he is talking, I get a sense of it that he is serious or he is not serious. Is he casually browsing, wasting my time? He wants me to show him the whole city and he is going to waste my time or he is dead serious. He is still confused. He says, but how do you figure it out? He said, no, we just get that gut feeling. We just get that sense.

And these people are getting the sense really well because a lot of them advertise on our platform and they would figure out who to talk to. So after the first conversation, they would decide who they should talk to and who they should not continue talking to. So we figured, okay, if they can do it, can we also do it at scale by using exactly the same process that they are doing, which is sentiment analysis. So what we started doing, which is extremely interesting is that we started calling our customers who were seeking properties and made a initial conversation with them on the telephone. That, hey, Saurabh, you are looking for a property. Can I help you? I am a property expert from MagicBridge. Can I help you with your query?

And he would answer the query, help him, give suggestions, etc. Now just this piece, we would record. And then what we started doing was we started using a lot of NLP to, first we used voice to text and then used a lot of NLP to figure out the sentiment on this. Now what we've created as a result of this is a very powerful product where a lot of the developers in the country are now buying this data, which is where we've indexed that this is a serious guy. This is a moderately serious guy. This is a casual browser. And they're actually buying that data because we've got the sentiment analysis done. And we've also got the voice recording to support it. But nobody really looks at the voice recording.

But it's our ability to go through the whole engaging conversation that the customer had with our associate, which helps us figure out that is he serious, not serious. And that's led to, from a yield perspective that we get on the data, about a 6-7x jump on the realizations that we get, which is phenomenal. And all that we did was work on a very simple insight that it's the interaction with the customer is where the person senses how serious he is. And we said, okay, let's use that. That conversation is a verbal conversation, audio conversation. So we just brought that into, we did exactly that, but converted it into standard data metrics. Very interesting.

You mentioned the process to buying a house or renting a home is anywhere between three months to a couple of years, depending on. But you already captured the initial intent. If I'm looking to buy something in Delhi or Gurgaon, we were talking on the cost of real estate in Gurgaon and Bombay. And a prohibitory it is in Bombay now versus Gurgaon. Or whether somebody wants to, how do you use a channel like email to have an AI obviously being the theme, to stay actively engaged with me if I was, if I had an early intent, maybe in a few years or two months ago, to buy a house in Gurgaon? A very interesting question. Like I said, the journey is different for different people. So we are always trying to figure out who is actively engaged.

And those are our cohorts that we target very actively. Now, the equally important is the ones who are inert. Because they've got inert because either the pricing was out of whack for them or they've just postponed the decision because they've got this great place on rent. Now, a key part of how we reach out to them is by looking at, by marrying trends in terms of pricing and their intent. And when there is a match is when different triggers go to them. Now, the way the triggers go to them is very interesting. The triggers don't go to them as standard emails. And we are actually completely opposite of the AMP examples that both my co-panelists mentioned over here. We go to a very conversational kind of an email where it's virtually a pure text email.

Even our logo is not there on the email. So our emails are actually pure, absolute text emails without a logo with a very engaging, conversational kind of… The way I would write an email to you is exactly the way the Magic Bricks emails are written to their customers, which are especially the passive ones. Because we suddenly want to engage with you. So it's not about one more marketing communication which is going to get lost. But it's a one-on-one communication talking about what you are looking for and why you should consider XYZ. And a lot of it is triggered. So when is it triggered for a particular person? I don't know. Because it depends on a lot of parameters fitting in. And if it happens, then a mail gets triggered.

Otherwise, it doesn't get triggered.