Amit Kumar Banka: what AI changes—and what it does not

And obviously AI becomes very important for lot of these growth factors. And we need to cater to the requirement, the understanding of each and every user. To the user with the right message, at the right time, with the right context, with the right content. Personalization for individual users, we see that the performance of each digital marketing campaign or each communication business going out goes up by as high as 2.3. As a marketeer, we need to use AI everywhere. This is not a part of the content. Amit, I'm going to start with you. Brands are using natural language processing to improve accuracy, relevance, and understand sentiment to tailor messaging based on custom emotions.

And food, for somebody like me and Grant seeing math, is like biggest passion, emotion quotient for me. How do you factor that piece to tailor messaging? And you spoke about Briani early. Maybe you can highlight that. How do you think about it? Food at 11.30 I have an answer for your question. Please edit it in the video. One of the points I am going to tell is a little controversial. First thing is I agree with your point. That is what I mentioned. We are trying to segregate now based on machine learning in terms of contextual targeting and based on the user behaviour in the past. Second thing is the area in which you live, the founders also live there. Not only that they also look at whether the messages are coming or not.

This is also part of contextual. What I am saying is that is the same case even with food and every guys or maybe even fashion guys. Again, where I was going with this is that you don't have uniformity. Correct. You are right. Actually I give you one more example. For example, if you have Swiggy1 membership which is like a prime membership. Normally what people do is what I am telling, people will actually share their account with multiple folks. They will be in multiple different cities or they might be in different area. And they might be ordering food in multiple times also. Now what happens is our system will consider only the login account as a user. Because we don't have understanding of any personal information.

Whether you have ordered or he has ordered from your account, we don't have. So for us it is your login details or from that account what all transaction behaviour has happened in the past becomes our primary source of intelligence for any kind of personalization. So to your point, you know what happens, one of the best categories that I enjoy with Swiggy1 is the bakery category. When you order cakes. And I can understand the AI making mistakes but when you write down what is the message that you want on the cake. Usually it's screwed up. Somebody's name will go as something else only and it's hilarious, it's funny. So I discovered this happens always, often. So I have now made it into a prank whenever I send cakes to people. So it goes.

But that's one area for improvement. The other thing that I want to flag as Swiggy user, the data controls and data breach. I have reached years, a CMO colleague of ours who just mentioned about some data scare that they have on a breach and stuff like that. Today, right now, I wouldn't name the CMO. But the fact of the matter is that some of your processes and policies in terms of a single device allowed. Now that two devices are allowed, things like that. Let's assume me and my spouse are sharing an account and she has another Swiggy account. Sometimes it gets confused, kiska account kisko hai. The notification ja rahe. One day I get a call, why do you ice cream kya kha rahe, you got sugar.

Because she had logged out of my account but the notification was still going. So things like that. So I think you are very right, Piyush, that as an industry we are evolving. And as we grow ahead, we are going to fall, trip, learn. But sooner the learning comes in, faster we implement it, we as an industry will mature. I'll just give comfort to him because you're staying in the APR area. All your orders of Swiggy and Instamart will come before time. So that is a plus point also of staying near the founders. But I know currently the great Indian restaurant festival is going on at 50% flat off. Please take advantage of that. I think this is very personal but on a very different point, I think it depends on the mandate as well, right?

You're trying to optimize for the user experience and if Swiggy is competing with Zomato and basically they are trying to push for more orders at that time, they will take that risk. There would be some tipping point at which they see, okay, if I'm sending notifications at the next time and people are putting off the notifications or uninstalls are happening, then they'll probably take that step. But so far you can… So there's a, just to add to your point, there's a metric called unsubscribe rate. What we normally track on the, specifically on the push notification side. So if the unsubscribe rates goes beyond the threshold, so there's a threshold limit what we have defined for each and every channel what we have.

We tend to pull down on the notification or the frequency of the notification which are going on. So that is a trigger which would be said and I think the second thing would be hyper-personalization because AI, no AI, you're talking about a thing where you said a very specific behavior. You're going on to the app to check the prices and not ever ordering. Are they able to take that much of context to see this person opens the app but doesn't order at that time? Maybe then they can experiment once or twice to see if I send a notification will they order? But again that many layers of actual personalization and context. I'll give you one example over there and that is actually…

So AI is one part but there is something called LBS which is location based intelligence which we are trying to experiment. It is not a very evolved thing in India. I'm not sure about the rest of the world. But even how to take advantage of that is very important. For example in RGO's case or Reliance case you will see that on the map, on the Google map, Reliance has actually bought a space of advertisement wherever their stores are there, physical stores are there. In Swiki what we are trying to do is we are trying to take advantage of location based services with respect to if you let's say from Bangor you travel to Goa for example. As soon as you are in Goa you should be able to get personalized message or food or restaurants from Goa directly.

And not of Bangor. So those are some of the LBS things that we are trying to explore. Now let's say if you don't go and open a Swiki app, still I would like to know where your location is as a user. So that I am able to recommend certain products to you in a personalized way. So what we do is we actually also try to partner with different players who might give us that information in real time. Again it is not PII data but it is for user recommendation. For example let's say if you go to MMT, Make My Trip and you have booked some hotel through MMT. MMT might know that this person is going to travel. Right? So that is only a data point or a signal what we get which will help us to actually recommend the right products at the right time to you in Goa.

So these are some of the things which we are trying to do through intelligence, partnership and DRE as well. PII data is a huge round of applause for this final game. I think it was PII data. Thank you PII data.