Shifting the Search Paradigm: How Branch Discovery Powers On-Device Intent for Half a Billion Users with Harish Thimmappa

Speaker A: For an ad tech company to succeed, the uh, supply and demand, which is where you get your inventory from and where you get ad order from, they need to be very closely managed. The product and engineering should also be very closely integrated with the go to market side. It's not always possible in all companies. There are large companies with big teams on the demand sales side versus business development side and they both really have exactly aligned incentives. You might onboard supply, you cannot monetize, you can sell things which cannot really deliver stuff. So, so that happens until you become a behemoth like Google where the system takes care of it. It's so big that everybody gets fed. So if you're a small, um, to medium sized company, aligning these two things are super important.

Speaker B: Welcome to How I Grew this where we dive into the world of digital marketing and connect with leaders in the space about how they're scaling, evolving and growing. Hello, Hello. Welcome everyone to How I Grew this. To today is very exciting. We have a great guest here. This is Harish Tumalpa, SVP and GM of Discovery Ads at Branch. This is a totally standalone business unit at Branch, so it's very fun for us to get to talk a little bit, dig under the hood with him. Hareesh has spent more than two decades helping grow technology and advertising businesses and he's been instrumental in building Branch's discovery ads business into what it is today. So welcome to How I Grew this.

Speaker A: Thank you Amanda.

Speaker C: Hi Adam, thanks for joining us. So we obviously know you but help our listeners learn a little bit about your background. So like tell us about yourself. Where did you come from? How did you come to your SVP role of branch discovery?

Speaker A: All right. I've been in adtech for over two decades now, almost 23, 24 years. Love it, love the technology aspect of it, love working with small companies, large companies, working scale businesses, build new revenue models. This is who I am. I'm an ad tech person through and through and I have been working with startups in the bay area from 2010. Around that time and just before Branch, I was a CRO at a, uh, small company called Unboxed, which was on search, e commerce search and before that I've had pretty extensive experiences with companies such as InMobi where I built their entire North American branch business team on the demand side, then supersonic ads, then Kenshu, so on. So really like dyed wool ad tech guy in the Bay Area. So I came aboard Branch to help run um, the discovery business unit which from the very beginning has been treated as a Startup within the company.

Speaker C: So Amanda mentioned it's standalone business unit. So a lot of these questions I'm going to ask, uh, I'm learning about these for the first time because I have very little experience, completely separate business. Tell us about what is branch discovery.

Speaker A: So branch discovery is many things today, but at its core Discovery is a product. It's a product that makes phones feel more intuitive. In a nutshell, that's the reason for Discovery to exist. How do we do it? What format? It's in the form of a uh, SDK, the discovery product put there by the OEMs, the phone manufacturers, to help learn the user, learn his or uh, her usage pattern and make the phone get them to where they want to go in a more intuitive way, in a quicker way. And today this product is on half a billion devices around the world. So incredibly proud of this. So that's the core aspect of what Discovery is. On top of this we have built a platform which helps developers and agencies reach audience and um, which also helps end um, users discover new things. So that's actually one of the reasons why we named it Discovery.

Speaker B: And for those in the audience that wonder, do I engage with Discovery? Like what are some of the OEMs that you work with? What type of device and operating system might they have that they've interacted with it?

Speaker A: That's an important question. So on the OEM side, so Discovery has multiple businesses. One is on the OEM piece, which is our fundamental business. Next one is on the apps, directly with the apps. So I'll talk more about the OEMPs on the OEMPs. For many reasons, we are on Android only. Also, I'll go out on a limb and say we launched Discovery before Apple launched Spotlight. So Discovery is really.

Speaker B: Yeah, we were the original Spotlight.

Speaker A: It's like cars having four wheels, right? So all cars have four wheels. I'm not going to say that somebody copied us. It's very intuitive. It's such an amazing idea and the way we deliver it is one of the best. So yeah, Discovery is on Android devices and thus the distribution follows Android distribution. Europe is pretty big, Latam is pretty big. Southeast Asia, including India. So these are the regions where Android is pretty big and thus Discovery is pretty big. So if you are in one of those regions, or anywhere in the world for that matter, Xiaomi, OnePlus, Oppo, Realme, Honor, Transient, some Samsung devices, these are all the devices where we deliver the experience, user experience. If you go into your settings, if you play around there, you will see Branch Discovery ads as one of the providers who customizes your interface for you. Again, if you are one of the lucky half billion people in the world.

Speaker C: So let me see if I can read this back to you and tell me if I'm right. Webranch or the branch discovery products puts ads alongside search results on the phone.

Speaker A: It's one of the many things it does. So let me restate this. Branch discovery customizes your search for you and pre search as well.

Speaker C: Does it do the search pre search?

Speaker A: Pre search as in zero state. So before you even start searching or you type. Yeah, before you type. That's a better way. Thanks, Amanda. A, uh, user swipes up, so there's no other interaction. Okay, so you just swiped up. Based on what we know about Adam, what we know about Harish, what we know about Amanda, depending on whose device it is, it's going to throw the right thing on the screen. It could be a, uh, motorcycle website for me. It could be book a cab for your teenage son to come home. It could be, hey, complete this shopping, school supplies for your elementary school daughter. It could be for Adam, spare part for his Porsche. It could be Amanda, uh, trying to look up real estate prices somewhere else with the Redfin. So all of these use cases are built upon how each one of us use our phone. What apps do we open, how frequently do we open them, what's the pattern of usage between days of week, time of day, so on and so forth. By putting all this together on machine learning, which we have built over a period of time, we are able to predict where Adam wants to go next. Thus we bubble that up. Not only that, we get you a deep link into that app. So it could be your favorite restaurant. Right. On a food delivery app, we'll get you a link into that because we know every Friday evening Adam orders this

Speaker B: stuff he can't resist.

Speaker C: Yeah, Sushi every Friday. So from an end user's perspective of the 10 to 15% of all of humans that have us on their device, if they run a search, branch is powering that search. And we can not only search on their phone and outside the phone, but we could also give them customized search results that are advertisements for the rest of the world. So it's like Spotlight, Apple Spotlight on steroids.

Speaker A: Yeah. It is also similar to Google search if you think about it, because a whole bunch of real estate is. Or it's actually better than Google in a sense of how the real estate

Speaker C: is closed, but, well, it's also on device. I can find my Uber app On my device. So it's like Google search. If Google also searched my computer.

Speaker A: Correct. And the way in which results are shown today, a whole bunch of Google results are, uh, promoted. Whereas organic, you have to go below the fold and scroll and whatnot. Right. For us, we have consistently maintained at no point in time should ad load be more than 10%. So 90% of what Adam sees will be organic stuff from your device, from your usage pattern.

Speaker C: Way less than Google now.

Speaker A: Exactly.

Speaker B: We're anywhere.

Speaker C: Yeah.

Speaker A: And it's also the half a billion devices. It's also a responsibility. Right. So what we don't want to be is a product which makes a user complain about the phone. Because we are integrated so natively with the device, what we do, you feel like it's your phone doing it. That's how much OEMs trust us. That also means we are responsible to not screw this up. Uh, right. Not make the end user complain to the phone maker. Hey, dude, what's happening? Why is my phone behaving this way? And we have managed to do this over several trillion interactions, actually. Trillion. That's how many interactions we have had with the user.

Speaker C: That's wild.

Speaker B: I, uh, want to dig into something wiki because you mentioned how it, uh, compares to Google or compares to Spotlight. This has been worked on for many years, and obviously the landscape of how people search and discover apps and content and information has changed a lot over the last few years. So can you talk about that evolution and how that's impacted discovery as product?

Speaker A: You're right, it's changed a lot, Amara. But at the same time, the fundamentals have remained the same. So there are a couple of things. One, where a user searches has evolved. There used to be a time where users searched in one place on the device looking for apps, um, they've installed on the device what we today call app drawer search, which is, hey, I open my phone, I'm going to go either to the list or the drawer of where my apps are shown, and I'm going to search for the app, which I want to go open right now. Okay. And that itself felt like, uh, a evolution on top of having to scroll and pick the app from the icons you see. Right. So today on an Apple Watch, you have to scroll to pick an app you want. That used to be the case on the phones. And then came Search, and that's the app drawer search, what we call app drawer search. That still exists. There's another location today on these devices, almost all devices called Global search. So spotlight on iOS is similar to what you call as global search on Android devices. Many Android devices wherein you are searching not just for the names of the apps, you're searching for activities within those apps. So if I search Fitness plus, I might start your run out run, outdoor run or start your workout. That's a deep link to an action within the app. And that's brand new. That's in the last few years that's changed a lot. That's changed user behavior a lot. And also users are more open. I mean this is an interesting thought which most of us know information, which most of us know. But it's still weird. Many countries outside us have much better Internet cell phone, Internet situation than a us does. So what that has done, it's made those users open to installing an app as long as it's not like multiple gigabits worth of an app. Right. So that also has made them open to suggestions when they're looking for an app. So if I'm looking for Uber as an example, if a recommendation, uh, comes to me saying hey, get $10 off on your next Lyft ride, the user there is totally okay doing that. It's a recommendation. Yeah. I mean Uber is obviously there at the top, but you also get a competitive recommendation or suggestion. So use. And the app load is not heavy enough to prevent them from installing and using it. So couple of things. One, number of places where users search has increased. Two, their expectations has increased in terms of what they want. They just don't want app results. They also want links within those apps. Three, people are open to installing apps even if it is for infrequent usage. And then there are obvious common things, themes which you I'm sure would have heard, which is there's clutter on the app Store. There are way too many apps. People are getting confused. Many apps do the same thing. Names are getting confusing. So building trust is super important. For somebody like us who exists on the device, we have consistently built trust. Why do we want to build trust? Right. We want Adam to trust the recommendations we make so that he continues to click on the recommendations we make.

Speaker C: I want to drill into that for a second. So I mean it seems like you have this device that's the most intimate device in a person's life. They have everything they want on it. It seems like an incredibly signal rich environment to be able to give them relevant advertising. What kind of signals are the most important that you actually use for targeting and delivering ads? Obviously someone says Uber love Lyft would be a likely candidate to try to advertise. But what other signals do you look for?

Speaker A: Yeah, I mean you called out one of the most important one, which is the search intent. Because that search intent is massive.

Speaker C: Super powerful.

Speaker A: Right, super powerful. Beyond that, the other signals that are uh, relevant to us is patterns of behavior which are intricate in the sense we are able to identify when a person is holding the phone upside down, when you're lying down on your couch. We're able to understand when a device is moving in a train. We're able to understand your behavior pattern on a weekday, which is a weekend. And also we're able to understand what you typically do after you spend time on this app for whatever weird reason.

Speaker B: Maybe after Redfin, I check my financials.

Speaker A: Yes, that's a good example. I was like, maybe after Slack app, a, uh, user opens up ebay to look at whatever he's been bidding on because he's just like spent some time working. So stuff like this. Okay. And there are very non intuitive patterns that emerge once you observe these things. Like what you just called out Amanda is a super intuitive stuff. But it might be weird to see that after you surf or uh, check your real estate apps for uh, whatever reason you are looking at soccer app consistently or there is a sports correlation there. We don't know why it is, but it's spotted. And Adam, to answer your question, yeah, those things like those, the user pattern and anticipated behavior.

Speaker C: And so how do you take that signal? So let's use your shocker example. Like everybody, every time they close Slack, they end up opening up a soccer score app. How does that turn into an advertisement?

Speaker A: Okay, I'll answer your question in a slightly different way. Not as how does it turn into an advertisement? How does it turn into a, uh, user experience? Because again, remember what I said earlier, everything leads to 90% of it being organic and maybe less than 10% of it being an ad.

Speaker C: And you guys are supplying both the search results, the organic search results and the advertisement.

Speaker A: Exactly right. So what do we do with this intelligence? Right, so we know this happens. So as soon as you close your Redfin app, your home screen, we can determine what's on the top of your home screen. So we will recommend your most frequent soccer app out there at the top, the one you regularly use. We also have something called suggested link which is check out Switzerland versus Qatar's scores. Right. That again opens the same app, but it's deep link to the specific match or the highlights on the ESPN app, or a YouTube link which is highlighted there. Again, these, uh, Are not direct app related. Right? Not direct soccer app related. We just know you open FIFA because you do that FIFA app often. We are now able to recommend ESPN and YouTube and a bunch of other things to keep you engaged in your line of thought along the way. Could we throw you a book? Tickets through Ticketmaster at Levi Stadium for the next match? That could be one of the 20 results you see on the screen. 10% off on your next ticket purchase on Ticketmaster for your upcoming World cup match. Does that answer your question?

Speaker C: It does, and I missed that the first time around is that you guys are powering both the organic and the ad search. It's not just an ad platform. So I understand what you say that. I also understand that if you understand that someone's a soccer fan, you can then use that for better targeting. One of the things I found curious is that on the website, you guys say that the bidding for the ad placement is on the device. That's fascinating. It doesn't happen in the cloud. How does that work?

Speaker A: It's a real cool innovation which we. I wouldn't say stumbled into, but we evolved into. I'll tell you why.

Speaker C: Good spin.

Speaker A: Right?

Speaker C: We didn't follow those vaults. Grew into.

Speaker A: Yeah, you know me, Adam. So it's not the classical bidding part, it's the selection part. So at any point in time, when Adam swipes up on his device, There are typically 200 to 300 different things we could show on the screen. And the screen can only receive 10 things. Could be eight apps and four links and something else. Right. So we have 200 to 300 different things. Not all of them are unique apps. Could be 80 apps and two and a half links per app. Something like that. Right. So we have these. And we also know the pattern, Adam's behavior pattern over the last whatever period of time. Right. All we need is 12 days sometimes. We know the user for years together. Right. And people keep the phones for two, three years. And even when they migrate from one device to the next, there is a continuity aspect there. All according to the local regulations that allow us to track a user from device to device. What do we do in terms of on device decision making? The reason we invested energy into this is we don't know when Adam's device is next going to be connected to Internet. Okay. So you might be connected to Internet right now, but you might text. Moment you swipe up, you might be going through a tunnel. Okay. And we don't want your experience to be interrupted. Remember, we also drive your organic results. So Organic selection is on the device. So it's an algorithm which we have perfected over a period of time. The ads part of it is sent to the device in what we call bundle updates fairly periodically. So Adam's device, based on the pattern it's exhibited, fits a few categories which a uh, potential advertiser is targeting. Based on the categories that Adam falls into, we get see he qualifies in the next six hours for seven to eight ads. We send you all those eight ads in very lightweight setup which we have again invested a lot of time, energy, effort because half a billion devices, we don't want to inundate the Internet or cause them to use their bandwidth or

Speaker C: pay for a bunch of ads that never show up. Which is expensive, right?

Speaker A: Correct. Exactly. So we send you a device is 7 to 8 ads as well. Now when you swipe up next time or when you start searching, the logic is built onto the device which then considers the current time. What was Adam doing just before this? What is Adam likely to do next? And looks at these seven ads and say while he's eligible to receive all these seven ads, any one of these seven ads, these two are the right ones for this movement. So that decision making, local device search lds and also the other aspects of ad selection are on the device. By putting this on the device we are also overcoming bunch of other problems. Privacy related issues.

Speaker C: Privacy issues. All right, I was going to say I see three things that this does. It allows you to do everything on device which is privacy protected. Number two is you have better information about how this ad will target than the advertiser would. So you can target and then also you don't need them to bid the most. If you know about the user, you can place the ad that's going to be the most effective, not necessarily is paid the most. And then, and that's essentially how Google works, ranks ads as well. Right. Google ranks the ads are most likely to be clicks, not necessarily the one that's going to earn them the most money.

Speaker A: Exactly.

Speaker C: So it's more performant. Yeah.

Speaker A: It's not local yield optimization, it's global yield optimization. You've got to keep the population consistently happy with the things that they're seeing and expressing this happiness by clicking on them. And we are also measured, we measure ourselves and we show this to OEMs, the CTR on the organic side, Adam. So not just the ad side because it's a nice metric, a barometer to show is Adam enjoying his phone? Is he using more of it? Is he clicking on more recommendations here or does he not click here and goes to the app drawer and manually searches for things and goes clicks there. Which means you are failing in a recommendation. Yeah.

Speaker C: What is the CTR on the organics?

Speaker A: Varies. It varies device by device. It varies time of day, day of week on average. Couple of things that will help you get the picture. Uh, a user swipes up on their Device anywhere between 40 to 80 times on average a day. 40 to 80 times a day or

Speaker B: doing it ticker on yourself every time

Speaker A: you do that on average. And every interaction typically gets on average leads to 0.7 clicks. Again, depends a lot. So imagine a user clicking 30 to 50 times a day. Right. And out of them maybe 2, 3% of them are ads. So we get massive. That's why the numbers scale astronomically. Half a billion devices 40 times a

Speaker C: day, 70% click through rate. That's crazy.

Speaker A: Yeah, yeah. And um, those 70% organic plus ads click through rate Ads obviously is going to be much smaller than that, which is totally fine.

Speaker B: Okay, so I want you to talk a little bit about how your advertising background and understanding of some of the limitations of advertising today that's more traditional that maybe Discovery helps overcome or that you've brought into this experience to strengthen the product overall.

Speaker A: That's a great question. From the very early uh, stages of my career I always felt for an adtech company to succeed the supply and demand which is where you get your inventory from and where you get add all from, they need to be very closely managed. The product and engineering should also be very closely integrated with the go to market side. It's not always possible in all companies. There are large companies with big teams on the demand sales side versus business development side. And they both really have exactly aligned uh, incentives. You might onboard supply. You cannot monetize, you can sell things, you cannot really deliver stuff. So that happens until you become a behemoth like Google where the system takes care of it. It's so big that everybody gets fed. Right. So if you're a small to medium sized company, aligning these two things are super important. And that's something I'm able to do because of the. One of the best things about Discovery is the team. Right. So by far the best team I've ever worked with. So we have the supply team and the demand team. They work very closely, hand in hand, they fight often and get stuff done.

Speaker B: That's conversation.

Speaker A: Exactly. And the product and engineering piece. Amanda, uh, the fundamental reason Discovery has succeeded is at any point in time we are Ready to fail on the 50% of things we work on because we have to be nimble. We right now have eight projects live. We expect to fail on seven of them or six of them. We expect to because. Or even if we don't feel we'll abandon them because we know that things fail. We know that we can't plan for things we don't know of.

Speaker B: Very healthy attitude.

Speaker A: Yeah, yeah, absolutely. I'm the biggest fool in the team, by the way. I make the most mistakes and I ask the dumbest questions. So these guys, the product and engineering team, are extremely nimble. And they are able to do that because we give them the confidence. I give them, them confidence. It's okay. It's okay to fail. Like, we don't know. We don't have all the answers. You're as much as responsible to generate revenue for this as a sales guy is. So let's work together. Right? And at the end of the day, nobody's going to pat us on the back. If we build cool tech, we need to build a business which adds value to the end user. In our case, half a billion install base and advertisers who pay us money. So keeping supply and demand aligned, keeping product and engineering nimble, setting expectations that many things fail. So these are the things which have really helped discovery.

Speaker C: I have a question about supply and demand. So supply is obviously provided by the OEMs. So that's your customer on that side of the marketplace. Then you're going out, uh, directly to advertisers. What is your ICP for advertisers? How are you growing those together?

Speaker A: Yeah, so we are going directly to advertisers, both direct brands, developers and agencies, large agencies like WPP and Publicis and OMG and so on. Right.

Speaker C: And are they selling or I guess they're advertising, managing advertisers on their side. Are these mostly like cpg? Are these mostly apps? Like what kind of advertisers?

Speaker A: Four of our top categories are E commerce, finance, travel, entertainment and offlay technology, like OpenAI and Gemini. They are a couple of, uh, our large advertisers.

Speaker C: And they're pitching the app, right? They're not saying come to the website, correct?

Speaker A: Yes, predominantly like, oh, more than 90% of our business today is apps, with about 10% of it being a click to a web page. Okay, then that's expanding. And what's a usp? That was the question. Right. So how do we present ourselves in this market where there are multiple options to an advertiser?

Speaker C: Right.

Speaker A: The fundamental USP is our first party data, having a space on the user's device independent of the apps they have installed. Which is, come to think of it, this is incredible Superpower. And again, I'm the first to put my hand up and say we can do cooler things than we have done so far, which we will.

Speaker C: You're the only one who can say, do they have this app or not? Besides the advertisers themselves.

Speaker A: Exactly. Even without an mmp, we can say that. Right. So that's the level of the APK being installed is clear to us when it gets installed. So, okay, when we go to a large advertiser or, uh, an agency, we are able to showcase that we have this larger install base, this deeper data, and this solid set of targeting that's available to you. And at the end of the day, we also deliver performance. So the best way, after having convinced them that this is such a unique opportunity, best way to scale, is to drive the first campaign to the moon.

Speaker C: So basically, succeed through performance. Yeah, I like that answer. I think that's the best answer of any advertisers, is just perform.

Speaker B: It does seem like partnerships make up a huge part of your role and what makes you successful. Can you share with our audience some of your tips and tricks for driving successful strategic partnerships?

Speaker A: You're right again. We have failed a lot. We have failed a lot. But we have succeeded where it matters by being very transparent with our intended outcomes. Every discussion, every proposal, the team today has reached the maturity to articulate, hey, this is the vision we have. This is where we see ourselves. We see you as a partner in 12 months, in three months, in six months. These are the references we can give you of how we got there. And we promise you one thing, which is transparency. If things don't go the way we planned, we will articulate this to you in a month, as soon as we get going. And also, we are confident enough to say that we are the best at this. There's nobody in the world who does this business as well as we do the on device search.

Speaker B: It's nice to have in your back pocket.

Speaker A: Yes, we really are. And also, this is not a surprise, right? Every large publisher ever always wants to sell their own ads. They always want to think about it. The thought just crosses their mind. Uh, it's our user. Why don't we sell them OpenAI?

Speaker C: That's what they're doing. They're moving ad business.

Speaker A: Exactly.

Speaker C: Where does this go? So you are on 10 to 15% of the world's devices today. Obviously, increasing your footprint's great. Where do you take this business from here?

Speaker A: So the key thing here, we also understand there is not an infinite ceiling on this, on being on the device per se. Okay. So the maximum ceiling is all the devices in the world, every fridge.

Speaker B: Are we talking like to that point too?

Speaker A: There is this one small company in Cupertino which might not agree. So that's a big add to.

Speaker C: Well, let's throw that out there and answer this if you're comfortable. Will you ever be able to get on Apple devices? Now, it's obviously up to Apple, but do you think.

Speaker A: Yeah, I absolutely think we can. We have the technology to make iPhones better, much better than they are today. The Siri AI can be much better with the way to process data. So yes, we can make it better. But I also am a pragmatist to know that this is not anytime soon. Okay, that's one part. Now, second part, what are we trying to do? We asked ourselves a question March last year at MWC, in fact, why OEMs? Why only OEMs? Yeah, because that's the core product about search and whatnot. Can we not help apps like somebody like Reddit? We can definitely do a phenomenal thing with them. We have now started working on with some of the large apps out there as publishers, so ads, but with a product which is tailored to make their experience better. I will not go into full details here, but we have expanded beyond OEMs into apps. That's one thing. Second, you also jumped headfirst into the programmatic business. So once we know so much about this half a billion users, these half a billion users, does it really matter where we see those users? Like once I know Adam does X, Y and Z every single day, if I get an ad request from Adam's device on ESPN.com site which I'm able to get a ad request by being a DSP on any programmatic platform out there, I will know exactly what to bid on an ad for for Adam's

Speaker C: device in a privacy sensitive way. Because bid is not a sensitive information. You're not saying this person loves soccer. You're saying you should bid five bucks because this person's qualified.

Speaker A: Yeah, I mean also the apocryphal connections. Somebody who loves soccer for whatever reason, like buys iPhone, mini whatever reason, like it's random connections like those which we are able to make because of the depth of data we have. By leveraging that data, we are able to be competitive against the behemoths, the TTDs of this world, the metas of this world on the programmatic space as well. So that's the third angle. So OEMs, then apps, uh, then programmatic space which we are pushing really hard into.

Speaker B: I want to go back to something that you said about Siri AI capabilities on Apple now that have just come out and you mentioning that this feature set of discovery is a little bit stronger. Can you compare what discovery is able to do from a context perspective and like the first party data and everything compared to just what like something jacked with AI can do?

Speaker A: Uh, it's a tricky question. I'll take a stab at this, Amanda. Uh, because we also use AI, we also use machine learning, we also try to predict things. What's the user going to do next? So what we have invested the most is the feedback loop. We're going to make a prediction, but we have the privilege to see what the user does. 10 next steps, not just the immediate next step. So we show something to Amanda, that's what we show to Adam. We show something. So we show A to Amanda, B to Adam. We know what happens. Like Amanda might click, Adam might not click. But what happens step two, step three, step four? Where does uh, Adam go? Where does he end up eventually? How long did he use this device? Today versus yesterday. What were the repercussions of showing A versus showing B to the same person on day one versus day two? Things like this. Right. So we capture a lot of post event data, feed that back and um, keep tweaking it all towards making user experience better.

Speaker B: Which is an important North Star that I think is unique.

Speaker A: Exactly, yeah.

Speaker C: Which tries to performance. And if you're trying to drive performance in both the advertiser's perspective and the user's perspective, it makes sense. I've heard this described as horizontally integrated. So you're horizontally integrated from intent discovery all the way through purchase, all the way past purchase to product support, whatever like an LLM is versus a vertical integration like Amazon. Amazon sees discovery purchase and that's it. And often they have no idea why someone bought something. You come on Amazon, you're saying winter coat, but Google knows I'm going to Iceland, I need a winter coat. Does it rain in Iceland? What's. So this horizontal integration allows you a much more holistic targeting environment rather than just like lowest price for a search.

Speaker A: Absolutely right, yes. And um, with the future where we will have a lot of apps working with us directly as well, it becomes both horizontal and vertical. So can we go in depth with an E commerce company let's say the sake of discussion, let's say Amazon. Right. So not only do we know Adam did A, B, C and D, that's why he's looking for this fleece jacket. We'll also know this behavior pattern.

Speaker C: I mean, it's just even retargeting. Like Amazon showing me my cart. My search would be valuable. I want to ask, you're on the ground. You obviously know the space extremely well. AI, uh, is changing the entire world. But from the two things that you know really well, how are you thinking AI is going to change apps? What are you seeing and where's it going? And then how is AI changing ads?

Speaker A: So one not so great thing that's going to happen fairly or is already happening. The barrier to launch an app has come down significantly.

Speaker B: Run the stores.

Speaker A: We are all going to be inundated. Are, uh, being inundated with pocket load of apps. Not all of them are of good quality. So credibility for apps as a whole has come down a little bit. Two in contradiction, expectation from apps has also increased. So on the one hand, users feel like, hey, apps are crap. On the other hand, the apps I like, I expect them to do a lot more. I expect my Reddit app to behave in a, um, phenomenal way to know exactly which subreddit I might go to. I expect my Uber app to know exactly what I want to do. Stuff like this or where I want to go.

Speaker B: I have come to expect that lately.

Speaker A: Exactly.

Speaker C: That means either flight to quality of name M recognition and a, uh, potential rise for super apps.

Speaker A: Yeah, super apps part. It's only been successful in one large market ever. Right. It's only been successful in China. I love those amps. They're very interesting based on the user patterns which I have seen being a the helm of discovery as well as the end user in different markets, I wouldn't bet on super apps in North America.

Speaker C: So not super apps. So what's the outcome of an app needing to function better?

Speaker A: Just get better. You just deliver. You continue to wow your users in that space you've occupied. Right. So maybe you'll add a few extra steps, make it more than what I expect a, uh, traditional E commerce app to do. Maybe you will. My thermostat app at home is going to be cool at predicting whether at a hotel room, I booked something like that, but it's still in that space.

Speaker C: Yeah. So that's going to be consolidation of quality and big names then.

Speaker A: Um, correct. And from an ad site, ad tech perspective, the algorithms are going to get Better. And that's where a company which has data, real usable data becomes super important.

Speaker B: And first party data primarily, it's gold mine.

Speaker A: It's absolute goldmine. The number of times I'm in a room with agencies, CXOs when they hear the amount of data we have, their eyes pop. It's crazy and it's awesome place to be in. And we are very respectful about the data because we are here for the long run. And we also have a very robust legal team which keeps us all in check.

Speaker C: Right, exactly. We're good with data, that's what we do. So I see what you're saying, which is like as AI uh becomes more prevalent, you're going to be able to serve even more and more relevant ads to make the whole thing run better and AI is going to enable that.

Speaker A: Exactly.

Speaker C: Fascinating.

Speaker B: To close us out. What should people that are perhaps marketers or digital advertisers be doing differently today?

Speaker A: They should experiment. As interesting as these scary times of there being an ad tech company around every corner, they should still experiment. The marketeers still need to diversify their spends from away from just Meta and Google. There are pockets of information asymmetry, there are pockets of consumer delight which you will miss if you don't experiment. And look for tools that allow you to experiment smartly. Look for companies that provide you service beyond just self serve experiments where you will invest way too much time to figure out what exactly happened to your $10,000 test budget. So experimentation is the number one feedback. Number two demand performance. With everybody demand performance. The world is no longer just pay on CPM and pray for the best. So you have to demand performance. Those are the two things if you follow those two things. Um, I mean actually sometimes I wonder how cool it would be to be a marketeer at this time because there's so many options, there's so many cool things to try and so many ways you can call a network's uh, bullshit. In the past anything could fly. Now you have enough tools to call their bullshit. And that's why I'm fascinated when our sales team lands all these big deals and when we go, go back and look at this, it's almost always consistency the data and um, performance that we deliver.

Speaker C: Right or you're wrong. Harish, thank you. This was really fascinating conversation.

Speaker A: Thank you guys.

Speaker B: Absolutely. Thank you for everything you're doing to grow discovery and make user experiences central to that growth. Thanks for being here and chatting it through it all really exciting stuff and always a pleasure.

Speaker A: Thank you.

Speaker C: I'll catch you soon.

Speaker B: Bye. Uh, everyone, thanks for tuning in. Thank you so much for listening. If you like the show, please leave a review wherever you listen to this and share with someone trying to grow their career or their business. Until next time, keep growing.