 If I am to tell you that I like Decathlon and ask you to guess what else I like, would you be able to guess? You can guess but you will guess everything under the sun. But if I tell you I like Decathlon, I read Times of India, I shop at Fab India and I listen to Ed Sheeran and I am a fan of RCB, you get a better personality of me, isn't it? You will be able to guess more correctly what I really like. That's exactly what cookies contain. Cookies contain deterministic data of what you are and cookies essentially are personalities of people. So what do we do at Affinity Answers? At Affinity Answers what we do is to enrich these cookies at scale by looking at the deterministic attributes and enriching them with probabilistic attributes. And this is done at scale at around 1.5 to 2 billion cookies every day. Given attributes A, B and C of a cookie, we add X, Y and Z which are the most likely. How do we do this? Unrelated to the cookies, we have a social recommender system built on social data and we have a catalog of around 75,000 entities across brands, celebrities, movies, sports stars and we use that to create a recommender system and that recommender system fuels this cookie enrichment. Actually it just happens to be that cookie is the example that it took today but it could enrich any data for that matter. This is something which we call as an act-alike model instead of a look-alike model. Where is this, how is this useful? This helps in advertisers to find mid-funnel consumers at scale which is very difficult otherwise. And what about people like you and me who are consumers? We start seeing advertisements which are more discovery in nature rather than a lot of retargeting and look-alike modeling. If I like Decathlon, I start seeing ads of Nike. That may not really reflect my personality. So at the end, what we really do is use social data and transfer that learning to completely different set of different type of data sets. Can you talk a little bit about how you do it as opposed to just what is done? I think that would be more useful for the audience. What exactly do you do which is different from look-alike or the recommendation system? Take a minute to explain that. What we do is collect engagements of brand-initiated contents, engagement of consumers to brand-initiated contents and apply collaborative filtering to get a recommendation model out of that. So that recommendation model can be applied to different phases like the way I explained to you, the cookie enrichment. Okay. Questions? Okay. One question in the back there. Just trying to understand the business model better. Your clients give you a bunch of cookie IDs that you track for them or how does this work? No, it doesn't work like that. So what we enrich is the third-party cookies. Cookies that belong to various websites are called first-party cookies. We do not enrich them. What we do is to enrich cookies through something known as data management platforms which for example, Oracle Data Cloud, LiveRamp, and they have cookies, aggregated cookies when we do not even know it is completely masked. As far as we are concerned, it's just a data with a set of attributes. Got it. And how does somebody leverage this insight and how do they tie it to their data? And advertisers through DSPs are able to buy audiences that we put on the data store at the audience store at these DMPs. And once they get bought, a small fraction of the money percolates to affinity answers. Okay. I think these are very good questions. Third-party cookies and some of the data privacy things that we were talking about are also linked to it. I mean, and that's like a much broader question. Maybe Vivek is probably not inclined to answer right now. But thank you Vivek so much for sharing what you're doing.