Strategic Decision Making, The Role of the Data Scientist, and Building Uber | Sundar Swaminathan

Speaker A: My gut feeling is that AI makes the best data analyst even better. Like, I think for the first time in the history, you can have a 10x analyst. Just, you could have a 10x engineer. You can now have a 10x analyst because they can automate away all of that work that used to be sort of monotonous, repetitive, etc. What that also does though is you can outsource the again, the 90% of analysts that have been sold that just knowing SQL without communication skills is enough. You've now got a tool that can replace literally every single one of your analysts. You have 10 of those analysts that are contractors. Every single one of those is the push of a button, doesn't get tired, and is 90% of the cost. But what it can do is sequence the right context of questions at the right time in a way that like actually pushes the needle and makes marketing or product and growth move along.

Speaker B: Sundar, welcome to the show.

Speaker A: Thank you for having me.

Speaker B: I'm excited to chat with you today about all things data analytics, AI, growth marketing, Uber, maybe even some department of the treasury stuff, which will be, which will be fun. I think just to uh, kind of kick things off.

Speaker A: Would love if you could give a

Speaker B: brief background into, into your experience, who you are, what you do.

Speaker A: Yeah, uh, yeah, well, again, thank you for having me. Excited to be here and um, you kind of already touched on some of my background, but I've had an interesting career where I, I started my career as a software developer, didn't love it. I went into finance, which is where I went to the US treasury and loved that, but just didn't really love what, you know, what my manager's role was. So I was like, okay, this is not for me. And then finally got into marketing data science and like truly found what I'm passionate about. And I've stayed there for over a decade. So most of my career was at Uber where I spent five years across the entire funnel. So I, when I left, I actually built out their global brand data science team. So really, really top of funnel. Before that I was in performance marketing, before that in lifecycle marketing. To really have seen the full stack and then went into a few different product data science roles, Product growth data science roles, if you want to call them. I was consulting for two years and then as of, yeah, Monday, I'm now the head of data at Balance, which is a series B consumer marketplace scale up, which is really exciting just to bring back a lot of the stuff I loved at Uber. I'm now doing something Very similar. Which is awesome.

Speaker B: That's awesome. Yeah. And it's. You have a really unique experience kind of running the gamut from different stages of companies. I love chatting with folks that got into marketing and growth in startups after kind of testing, experimenting with different careers because it also informs how you approach your work today and how you approach your work as a, as a data analyst, marketer even Myself I started my career in sales and then kind of pivoted into marketing and growth marketing and a lot of those elements I still kind of use on a day to day in my role and also just in, in real world and in life. If you kind of think about, you know, some of those foundational elements of starting as a software engineer at Bloomberg and working at the Department of the treasury, is there anything that kind of like sticks out to you that like informs how you've kind of grown in your career in the startup world?

Speaker A: Yeah, that's, that's a really good question. I think, I think just the fact that I was able to bounce between those roles like I mean so uh, two, two questions answer. I think one, I think just learning to be a software developer was just learning to be very independent. Something I was really surprised by is like how much they softer just repeat like you're not like they're basically the rule is like if you're rewriting or if you're building new code like you're doing it wrong, like you should just be basically rewriting or borrowing from someone else and just packaging it for whatever your app needs. And so I became really good at googling as a software developer and it's something like you don't think about but like it's really important to be able to just be self sufficient like I can if I get stuck on a problem and now AI makes it a lot easier but like it used to be like you would have to learn how to Google and go through stack overflow. And that's also true as a data scientist. So I think learning how to Google is something I've become really good at and it's still something like that makes me easily like if I'm curious about something I just go Google away at it. So it's a weird skill set to say but it is something I think I'm quite good at as a, at the U.S. treasury. I think a lot of it on like just understanding how important your decisions can be and learning to make really important decisions at a very early age. So I was at the U.S. department of treasury and so we were managing $19 trillion of U.S. debt. We had to predict when we thought the government would run out of money during government showdowns and sorry, government shutdowns. And so you learn very quickly, like it's not about the complexity of the work you're doing, it's just like this is really making sure you've thought through it, like making complex decisions, like having the validation behind it. I took those into Uber. And so being independent was really useful at Uber because you had a lot of autonomy. But then also making like pretty big decisions at a relatively early stage career was also something that was true at Uber. So yeah, I think those are the two biggest ones. I would say that's something I hear

Speaker B: over and over again when, you know, listening to interviews from executives, founders, CEOs, you know, they even talk about Jeff Bezos talking about this, like the, the magnitude of decisions and how you don't need to make a lot of decisions, you have to make a few really, really good ones. As you think about maybe decision making, your decision making framework, do you think that's kind of like a, uh, muscle that you've, that you've grown over time to be able to differentiate between what's a micro decision that I need to make quickly versus what's a long term decision that I have to go on a walk and I have to journal and reflect and kind of think about like what advice would you have for marketers, analysts as they think about that making decisions?

Speaker A: Yeah. Oh, I love it. This is really like, uh, these are some hard hitting questions right off the bat, which I love. But I think the only way you learn what is a micro versus a macro decision is a bit of experience. I mean there's some of them of course, that you have a gut feel for. But if you look at it from a business context, from a marketing context, if you're afraid to make the micro decisions, it becomes nearly impossible to make the big ones. But if you approach the micro ones as macro decisions early on in your career, because let's be honest, when you first join, you're not making macro decisions. Like you're not being put into situations where you have to make macro, but treat the micro decisions like macro ones, right? Like just because it's a micro decision, it's micro in that the impact might be small and therefore you don't have to think about it a lot, but you still should think about it in an intelligent way. And so treating the micro as like, uh, okay, what if my career depended on this? Just as a hypothetical early on I would overthink it, but that made me really good because over time those micro decisions, I would think quicker because I've already spent a lot of this time overthinking it. And you like at first you're going to overthink, like you shouldn't try to avoid it. It's actually probably good for you to overthink. And then you learn, well I didn't have to overthink this. And the next I was like, well I definitely didn't have to overthink that. And oversaw all of a sudden by process of elimination, you think exactly the amount you're supposed to for a microdecision. But as your decisions get bigger, that framework of I know this isn't overthinking, I know this is you start to calibrate. And so I would say in the beginning, treat microdecisions m like macro, build up that muscle. Because unfortunately if you f up a micro decision, depending on the place that you're at culturally, it could be like, it could not go well for you. Like you also have to be in a place that supports you. So that's the other thing too that you learn over time is it's less about the decisions you're making and um, where the place that you're at that you're making them that actually will have the bigger impact. But when you don't know that in the beginning of your career, just try to think a little overthink and then just start to calibrate over time. And it's the only thing I've, I've personally worked, has worked for me.

Speaker B: Right. You know, I, I think about that as, I think about a lot of things as existing in a spectrum, maybe decisions existing on a spectrum of like you know, maybe easy, simple, small to large, complex, longer term impact, whatever. And then within any given role there's going to be that spectrum. So an entry level job, you know, the small decision might be like going uh, like doing a training or I don't know, uh, you know, attending a meeting, like very like micro, micro kind of things. And then the, the larger decision for that entry level role like you said would be something seemingly simple to a more you know, someone in their mid career or something like, like that. So now as you exist in you uh, you kind of like grow your role over time. I guess the idea is that you want to make sure you have a balance between like spending time in those small decisions and automating that and making that easy but also making sure you're allocating time towards Thinking through some of those more complex kind of things and like how do you know what the appropriate balance of that was?

Speaker A: Let's just go with the examples you provided because I actually really like them. Right. So if I were to think about like a m, like a training decision for a micro, as a micro thing at uh, the earlier set of my career, I was overthinking like what is the audience going to ask as I presented it, right? Like let me go. And I would spend hours rehearsing a uh, uh 15 minute training. And then over time I'm not rehearsing for hours, I'm rehearsing for a minute, but I'm still practicing. As I've gone in my career I've really realized like your manager is probably the most important person in your career. Like the whole classic saying if you don't leave jobs, you leave managers has been very true for me. Like I've stayed because of managers and I've left them. And so it. Part of wisdom I think is also realizing you're not in it alone. And so when you go through and you become more mid career, I would lean on my manager even more because the decisions become bigger. So I'm like, hey, you know, I can either do these 10 trainings for example, or I could do this one meeting, but it has like a senior stakeholder in it, which one should I do? And it's like, well you go to the senior stakeholder one because you need that visibility. Why do I need the visibility? It's not going to help me today. But like, yeah, but come, come this time where you need buy in or you need promotion or you need something, you know. You know, my manager's telling me like that visibility is important. That's not something you know yourself. So part of it is learning from people that have already done it. On what are those micro decisions versus macro. Having the faith that, listen, I've already made these type of micro decisions a hundred times. I don't need to spend as much time, I need to get better at this stuff that I'm more uncomfortable with. And there's this natural pull I think of um, if something is uncomfortable for you, at uh, least for me, I know it's bigger. I'm prepared for. And so I like collaborate. I mean I corroborate that with someone who has done it and I'm like, hey, is this really as big of a decision as I'm thinking? Yes. No. And then you, you sort of steer the appropriate time towards it. So yeah, a lot of it is just, it's, this is why I think like even marketing data science is such a human function. Like it's so much of it is like gut feeling. It's talking to your manager. It's like, hey, predicting like what I think people are going to react and then leveraging the experience of stuff you've already done and be like, okay, that wasn't as big a deal. I saw it play out the way I thought it would. This makes me feel uncomfortable. Let me go towards that. That's kind of how I thought about it. I don't know if that's satisfactory because this is the first time I've had to think through how I think about these decisions. But it's a really, it's a really good question. So. Yeah, let me know if that's helpful.

Speaker B: Yeah, yeah, for sure. Uh, I mean I'm also, I'm also thinking about, you know, I want to, I want to get into, to Uber, but we're, we're jumping. I want to stay on this for, for a second. There are, there's also, there's, there's decisions, big decisions, small decisions you have to make in your role. And then there's also the skill set that you have and you tend to lean towards making the tough decisions using this. You want to leverage the skillset that, that you have. Right. When it comes to analysts, one of the, the things that, that you kind of mentioned when we chatted last time is that 90% of them are pretty shit. Right. And I think what we were talking about is that certain elements of an analyst role have kind of been commoditized to a certain extent. And so it becomes more about some of these soft skills of communication. I was reading an article that, that you wrote on experimental about how to have a very simple dashboard and how we can over engineered dashboards and create too much work. And I see this as something that's like very prevalent specifically in the kind of data analyst function of really smart folks that maybe don't. Aren't. It's not. The human element doesn't come as easy. So I wonder how you think about that.

Speaker A: Right.

Speaker B: Like that kind of the skill set that somebody technical should be thinking about building.

Speaker A: Yeah. So an important not even clarification on the 90% or shit is it's not because the analysts.

Speaker B: Right.

Speaker A: Like that's something I want to Clarify is like 90% of analysts are shit because they are trained to be shit and, or that is the expectation that they've been sold. But it's not like a I, uh, don't think there's like a unique talent that you need to have to be an all star analyst. Right. This is not like an athletic thing where like kind of like you're born with the ability to be 6ft 5 to go play NBA. It's uh, it's not like that in a professional career. And so what's important for me is how do you actually train analysts to understand what is important in the role and then how do you fit their skill set into that? That's how you make a good analyst. Right? Is here are the expectations for high quality analysts and here's how we can morph your innate skill set to uh, do that. So just as an example, I've never learned Python and I think if you were to tell any data analyst, if I were to just like throw my hat in the ring, like, I don't know, like, I think 99% of data analysts would laugh at me. But I feel really strong in my career. Like I am an above average analyst because I've got this ability to admit when I'm wrong and just this is going to be a weird thing to say and just like smile like I'm pretty, like I'm often like the cheeriest person in the room. And that's my advantage because like, I'll admit when I've made a mistake, but then I'm like, hey, let's like work together. Like how do we. Like, like how do I. And, and I'll go and be like, I am sorry. Like all these little things that we don't really do a good job of coaching analysts to be like, hey, at the end of the day it's just a relationship. Like a stakeholder is making a decision and evaluating you. Yes, of course on the data, but they're always willing to forgive that if there's a relationship behind it. Right. I'm not a perfect analyst. And so for me, like the thing that I would coach analysts to like a lot on is remember, it's like a human relationship first. And then like, like you don't have to be extroverted, not introverted. It's not about that. It's just about again, it's, it's a human thing. So like, that's how I like recommend analysts sort of go beyond what they're just reading and be like, I got another certificate on SQL. It's like, well, cool. But like, when's the last time you've actually sat down and, and even just had lunch with your stakeholder or you know, like Go, go do that. And I bet you you get much further in your career than getting that next certificate. And uh, a lot of it is because I had mentors and coaches tell me that they're like, hey, you're, you know, you're not good at saying no, you're not good at, ah, you know, making sure that the stakeholder has like the right information. Like all of this stuff. Like, I was so buried behind, like how I would look at the analysis versus positioning for how the stakeholder wanted it. And so, yeah, that whole framing is something I recommend.

Speaker B: One of the things that, uh, I've noticed and I see kind of maybe changing within a lot of analyst teams, you'll have, uh, a lot of subcontractors or functions that are outsourced to different countries. I mean, a lot of time, India, right? Different. Different places where there is maybe, uh, time zone differences, works being done off hours and then presented in the future, you know, being presented without really being communicated as much. Maybe there is communication, cultural disconnect sometimes. And it's a challenge, right, because some of these, so soft skills, you know, it's, it's. You can't take a course on it. Right. And I wonder, like, you know, you see the ability to take these boot camps. A lot of this education has become commoditized, it's being hired out internationally. And then you have AI on this other side of things, which might automate a lot of it too. So I'm curious, like, when you talk about the future of, of analytics and we talk about how these soft skills are super important, does that even become more important in a world that I kind of just structured, right. AI happening. And just the challenge of maybe outsourcing some of this stuff to save a few dollars too, because you have to pay someone more in America.

Speaker A: Uh, yeah, this is, this is a topic that comes up often. And so what I'll say is my gut feeling is that AI makes the best data analyst even better. Like, I think for the first time in the history, you can have a 10x analyst. Just, you could have a 10x engineer. You can now have a 10x analyst because they can automate away all of that work that used to be sort of monotonous, repetitive, et cetera. What that also does though is you can outsource the. Again, the 90% of analysts that have been sold that just knowing SQL without communication skills is enough because you've got some, you've got a, you've now got a tool that can replace literally every single one of Your analysts, you have 10 of those analysts that are contractors. Every single one of those at the push of a button doesn't get tired and is 90% of the cost. But what it can do is, is sequence the right context of questions at the right time in a way that like, actually pushes the needle and makes marketing or uh, product and growth move along. Like, I have yet to see like an agentic AI that's like, okay, cool. Like, hey, have you thought about answering this churn question first and then this and then this and then. And so that sequencing and sort of human thinking is not there yet. And I'm not sure if you can ever completely encapsulate that because it's, it's uh, yeah, like again, people need like AI is just is essentially taking a bunch of information and bringing it to the average so that it's, you know, you're, you're, you've got the average information at your fingertips, which is awesome. That means it leaves a lot of room for above average. But anyway, my point there is like, it will change, and I hope it does change the analytics and data science field in, in that it removes the need for like, you know, just being able to have SQL. But if you are really good and you have really strong communication skills, I think you become even better. Like, I imagine if I had an analysis now, it would take me 20% of the time it used to take me, which now means I can churn out five times more like analyses. Each one of those then makes me more impactful because not only am I doing the same volume of analyses, but I'm doing in a shorter time. But you have to have been a good enough analyst to understand to ask the right questions in the first place. And so even a lot of the conversations I hear around, okay, we're going to give everybody access to a prompt that'll write SQL for you. Well, cool. I mean, people have not been able to ask analysts the right questions for two decades. I don't think you're all of a sudden going to be able to just because you have this AI tool, right? It's always, were you asking the right questions, even understanding stuff as simple as like cac? Like, people still have like a thousand different definitions of cac and like, not everyone's computes the right way. Like, I could go down to so many different examples of. It's never, it was never the, the, the answers that were wrong. It was always the questions that were wrong. And I don't think until people learn how to ask better questions that it, it completely eliminates them, but it does make, it makes people. If you're a bad analyst, you should be scared. If you're a really good analyst, you should be happy because you can become a 10x analyst or data scientist. By the way, I use those terms relatively interchangeably here, right?

Speaker B: Yeah. If AI is very good at answering, uh, questions, you need to become really good at asking them and then also kind of understanding what is a realistic answer and what to expect and learn how to massage that. And the skill set of learning how to ask the right questions is something that only comes with experience of working with an organization, seeing the growth challenges, being able to identify like here are the areas to prioritize. Um, I think that's a good transition to Uber. The, the headline at, at Uber is that you built the brand data science function and measured 1 billion in in spend. That's, that's a lot of spend and a lot of experience in understanding what are the elements that led to the growth of that organization. What I want to get your thoughts on to, to start is thinking through how to how, what your perspective is on the ways to deploy media, budget to grow an organization. Do you have a framework? Do you have a philosophy on that after seeing so much Ben go through Uber?

Speaker A: So what's cool is that I was able to use data science to verify that like marketing has not changed in a hundred years and the principles of that have not changed. And I don't think will ever change. And what I mean by that is at the end of the day, you need to communicate in the most effective manner how your solution solves a customer's problem. Which means in a very simple sense, you have to understand what their problem is and then you have to build a solution for that. The best example I have is not even from brand, but I'll talk a little bit about performance marketing here. The best performing ad for Uber for a very long time, and I'm not sure if it's still true, was push a button, get a ride. We could not beat that, like just could not, no matter what you did. And at the end of the day it is so simple. It is so fundamentally simple. And I'm probably sure, I'm pretty sure if you research a bunch of like, talk to customers and they're like, why do you love Uber? It's like, man, like I pushed a button and a ride showed up. There you go, there's your language. Now you don't even have to write the copyright. So what I, what I was able to See is, yeah, it's just fitting that solution. And so you could not like, you know, at the beginning of Uber, the best performing market M of China was referrals. Okay, well it makes sense because you've essentially productized word of mouth. But like, was it incremental? Nobody ever asked. But it was still there to prove that word of mouth was going. Then eventually we like we're on paid social. Well, eventually we were so saturated that we didn't need paid social for rider acquisition in the US So we turned it off. Okay, well then what became the issue later? The issue later was that there was in every country there's many competitors. We're losing to DoorDash on UberEats in the US so what do you see Uber do in the last four years? You see a lot of push towards brand awareness, which is where the billion dollars of spend came in. All of those principles were true from the same. And in fact, if Uber had been brand building from the beginning, there would be zero room for any of the competition. Like you could not have gotten a Lyft to succeed. Right. If people had loved Uber from the beginning, Lyft couldn't come in as a counter position to be like, we're better for drivers. And at the end of the day now Lyft and Uber did the exact same thing. So drivers are arguably better or worse off in the same way. But what I learned was like, you know, nothing has really changed. We just have gotten better at the science to back it up and prove it. But even now, like balance is, this is a, is a, I will say a commoditized business in that it's luggage storage. And so the same principles are there, right? Like you can only really win on trust and maybe you have more luggage storage available to you than in competition. And so it becomes another brand play. And so yeah, that's really what I learned and that the metrics and frameworks are all about that. Like if you don't, if you want to measure the impact of brand awareness on business metrics, well, first you have to prove that your campaign actually moves brand awareness. Pretty simple, logical thought pattern. Then you do a brand lift study. Okay, great, we're moving awareness. Well then you set up a six month GEO holdout and you prove that it's moving the business metrics. But how do you know that your campaign is moving brand awareness, Brand lift study. But then the briefs, the campaign, like they were all the same. Like, you know, you still had to write a campaign brief, you still had to draw from an insight, you know, you still had to think through what was a problem that I was solving for a customer. And so, you know, for me it's, it was really cool to just validate a lot of the literature and research around marketing. M is still the same.

Speaker B: One of the things that, that I, I love about what we do and I, I love about my experience scaling media mixes for, for different brands over the years is that it, it is very deterministic in a way because there, there really is kind of a clear pathway to getting to scale and the ele you need to have a good product, a great product. You have to go after high demand audiences and then move up to mid M demand and low demand and build that, that brand. And there's really a clear pathway. But it's hard because it wouldn't exist if you didn't have a really good product to start with. I want to hit on the, the Facebook US spend in a second, but before I do that, I think one of the things that a lot of folks listening to this might struggle with right now is that we're in the era of efficiency and trading dollars from really good cpa cac. Looking things into video view, ad recall things is hard to sell to a CEO and a cfo. And so how do you do that?

Speaker A: Um, I don't think they're going to love this answer. I mean, you essentially don't. If that's the argument that you're having. I'm not saying don't try, but you've lost because I don't know how many more examples of this we need to share. Right? Like it's like, okay, pick, like you almost have to get confrontational with your CEO and CFO about this. Like, okay, well, what shoes are you wearing? I'm wearing Allbirds or Nikes.

Speaker B: Why?

Speaker A: Okay. And like, you know how many, like you have to go through this. Like, so my point there is like, I know it's like, it uh, might feel like a cop out answer, but like, it's such a, it's such like annoying question. Not from you, but it's such an annoying question in general in that, I don't know, I just get sort of like fed up of having to be like, why do we have to keep proving this? Like I would almost say then go to some place that actually gets it because that, it's like such a taxing endeavor to do that. By the time you get that buy in, you're almost like so uninspired to. Then go build the marketing strategy because it's going to be as soon as you launch that brand awareness. Well, okay, well, when are we going to see awareness go up? It's like six months, three months. You know, like, okay, fine. Well, okay, uh, three months later. Okay, well, fine. Well, has it gone up? Yeah, it's gone up like 6%. Well, is that a lot? Is that a little? It's like, I don't know, like I can't, like that was, I'm not able to map out and predict. And so you like, the foot is always on, like, this isn't working, let's stop it. And that like attitude is so hard to beat. That. Yeah. The only way I think you can do it is like you go, here's how much budget I need, here's the metrics that I think will move. And then we don't revisit this for six months. Like, unless like the goal for you should be to get like a six month breather. If you get anything less than that, then you're, it's just, it's not worth pursuing. Right. So getting that buy in from your CFO CEO, if you want to look at a heuristic to think about is get a six month Runway and just say, here are 20 case studies. Let me know if you want more. Find case studies within your own business, within your own industry. And yeah, and that's, that's it. And be like, oh, by the way, we have to make sure our uh, product doesn't suck. We have to make sure our uh, brand doesn't change. We have to make sure, yeah, there's not an existential crisis, et cetera. And then cool. Like, I feel pretty confident that over six months I'll show you that brand awareness lifts search volumes, which means we can do higher volume of throughput at the same cac. So like here, here's all the evidence, here's all the case studies. But like, yes, that's, that's kind of my answer. I like, I get this asked all the time and I just, I feel bad. I like, as soon as someone asks like, how do I convince? I'm like, I don't know, dude. It's like kind of like as a data scientist, I'm like, how do I convince my CEO, uh, that like being data driven actually proves anything. Like, look, I don't know, just look at the top 30 companies in the world and they all invest a ton in data scientists, data analysts. Like, is that enough? Not enough evidence? Like, so, yeah, that's, that's kind of the, it's kind of a hard thing to, to, to, to, to push for.

Speaker B: To me, I, I, I feel like it's, you're operating within certain financial constraints. In the aggregate you have a CAC target that you're trying to hit or a roas, uh, target or a revenue target, whatever. And if you're hitting that and you can allocate more room in the budget to doing things that don't have that CAC target because the other things you're doing are getting you there. I think it's a challenge when that I've seen sometimes where you have a CAC target and then every channel is beholden to that same target or we try to think in that manner. Right. Well we need to get a three row as and so everything needs to be around three and if this thing's at one, we cut it kind of thing.

Speaker A: Yeah. So I think there, right. What you, the argument becomes like just understanding the three levers that you have to pull with the marketing, which is I can, I forget the three now. Exactly. But like you can keep the same CAC and then therefore keep the same volume. Like I can give you same CAC and higher volume unless I do something more upfront. Um, or I can do, you know, conversion rate optimization. But the point being is like there's three things that you can pull but you only get two things. You can only get like good cacs and decent volumes. But, but then the third thing like has to stay. And so the point there is like listen, if you're happy with our CAC and you're happy with our volume and you're happy with our growth, great, fine, let's not have this battle right now. But just to know like we estimate that we're going to tap out of our demand in six months. So I'm going to put on paper now that in six months your cats are just going to start to like spike but we're going to keep the same volume. So like that's the hard part is like you're not prepping for now, you're prepping for six months and you're also not like, like it is about efficiency, but it's gotta be efficiency in terms of like, like in the context of the business, right? Like if you're a high aov, high order value business, maybe you can get away with not doing brand spend because you just need the one or two contracts and you're good. But if you're like, if you're in this like small AOV business, like you just need that to be able to sustain higher volumes. And then your entire business model is designed off volume. And so again, like, you know, you just have to find a case study. But yeah, again, not an easy conversation.

Speaker B: I think about this world in terms of putting dollars across a demand spectrum. And in any given product or service, there's going to be a large number of people that have low demand for it, a smaller amount of people that have kind of middle of the road demand, and a very small amount of people that have very high demand for this product. Um, and so you spend dollars on the high demand, smaller area and you capture that market and then eventually you get saturated, you move up and you move up and then you're in this place where you're kind of spending across all these different demand stages. Obviously you focus a lot on the lowest demand stage. People, audiences that either didn't know the brand existed, didn't give a shit, maybe had the problem, but weren't aware that this was a real problem for them. At Uber specifically and maybe some other places that you've been at, what have you seen from a messaging perspective work really well in brand or what type of experiments have you run to see, oh, this actually did really well?

Speaker A: Yeah. So part of demand is that a lot of people don't realize inherently that the demand that they have just for example, again, like most people when they think of Uber early on thought of it as like, I need it for the weekends. But then it's like, well, okay, do I really want to go park my car at the airport if I go away on a trip for two weeks, like, and pay, uh, $200 to leave it at the garage? Oh, wait, or I can take like a $30 Uber. So all of a sudden we would start to create demand across use cases. Right. So there's, you know, a lot of it is understanding our customers, right. Which is the primary use case early on was weekend nights. So who are those type of people? They're either, they're not college students because we were a little bit expensive in the beginning. They're probably young professionals, really, you know, solid money. Okay, well, what are the uses that they have? They could take it to and from work, they could take it during the week from like their social outing. And so we would message that like, hey, you know, don't forget, like after like, you know, three drinks at a bar. Uber is a great ride home during the week though, right? During like happy hours. So that's like a very early example. But like, over time it's just introducing more and more use Cases because that's where, that's where the demand is. The demand is like, yes, it might be low demand, but it's low demand, high volume. And if you look at in terms of aggregate demand by just multiplying demand times volume, it's always at this low demand portion that you have just like massive, massive demand chunks. And every company solves the problem of how do we go vertically, like how do we go more upmarket, how do we get them to more use cases. Right. For Airbnb, it might have been a great example is like for the once, once every two week trip with your family. But now like they use it Airbnb for work. So now I'm using Airbnb for work, unlike the daily use case, which makes you think about it for the long term. So, so a lot of the messaging we would do is just how do we introduce new use cases? How do we, you know, make you realize that the solution that worked for this is the same, is the same form, it's the same function of a solution, but mentally we paint it in a different light. That's also why you see more products like comfort versus uberx. It's the same supply, it's the same drivers, just better rides. And so you might be willing to splurge. Like when I'm like, this is either paid by my, my, my employer or listen, I'm on a trip, I'm going to do comfort versus X. Right. It's just taking that same form, function and messaging it. That's how you would go further and further up demand and you would create demand. That's the other thing too. So many people don't realize they, they need Uber not eat, but they want Uber eats or they want UberX. And so it's our job as marketers to sell them on that vision. Why do you even need a car? If I live in, if I live in the city, why do I ever need a car? You know, calculate insurance plus lease costs plus gas and it's, it's, it's more expensive than an Uber. So like right now, all of a sudden you create this demand of yeah, we're just going to replace your entire car. That's like, that was the vision and that's, that's still the, I think the ultimate goal. So that's, that's how you, that's how we thought about it in market.

Speaker B: Cool. I know we're, we're running on time right now. I have one more question and then we can kind of, kind of wrap up. You mentioned building out different use cases as a way to expand maybe the spectrum of demand that exists. I'm really curious tactically and I'll just give you a very specific example. Like I'm on Facebook, I have a certain amount of budget, I'm doing purchase optimized campaigns, direct response. Right. And for self, the credit building app I work for, have low credit build it with self, get started today, like this type of messaging. And then I want to shift some of my budget, you know, 10, 20% of that budget towards reach optimized. I want to get lower CPMs, I want to get a bigger audience and I want to say something to this larger audience so that they eventually get into that bucket of the performance optimized campaigns. And that's kind of the thing I'm asking about. You kind of mentioned use cases. There's a challenge sometimes where maybe you can't just say here's seven different ads for the seven different use cases. Like I need to have one thing. How do you think about something like that? Right. Like I'm trading higher cpm, more intent based ads for lower cpm, low intent ads.

Speaker A: Yeah, it's still in the service of uh, eventually being able to serve more volume of customers at a reasonable cp, reasonable cac. Because if you never spend the time building the demand or even creating that demand on the low CPM high reach, you just naturally saturate and tap out. And so what this is doing is saying we're going to take a very short term hit on our business right now, but to create a more sustainable funnel, knowing that we're trying to create with this low cpm high reach sort of strategy, someone who will eventually look like this high CPM target audience. The reason like you make that sacrifice, the way you justify it is yeah, otherwise we're just going to run out and the like you could take a 20, 10 to 20% hit now or you could wait completely saturate and then you take a 10 to 20% hit for like four to six months because you have to rebuild that funnel. So we're going to take a 10 to 20% hit now for like a month. But eventually that 10 to 20% funnel catches up versus taking a uh, hit six months down the road. That's the way I think about it. And in terms of picking use cases and all of that, again it just goes back to the roots of knowing your audience. Like what's the sell, what's the value prop that convinces people to join the most? Like you have to know why people are joining now, instead of thinking about, like, well, how do I convince them? Well, don't try to convince someone. Tell them what's already working. Like, there's no like for, like, for Uber. It's like a million people have now taken this trip to the airport. Okay, well, I never thought about it, but if a million people have thought about it, then, okay, cool, maybe I'll think about it the next time I have to. There's no reason to reinvent the wheel for that low CPM audience. Just tell them what's working with the high CPM audience and put imagery that puts them in that situation. Right? Like, like, what is the evocative image that says, oh, yeah, so annoying. Like, you know, maybe it's like the bill at the end of the. The 14 day garage airport, and it's like, oh, my God, I can't believe I paid $200. Well, you could have avoided this if you had just taken a $25 Uber. But, you know, the airport use case is, is good for your business. You know, that's how a lot of people take Uber first. So just take that and use that as your use case. Like, don't reinvent the wheel. Just do what's working, but position it in a way that will eventually work, I think is the best way to think about it.

Speaker B: Awesome. Sundar, thank you so much for coming on the show. Where can folks find you?

Speaker A: You, uh, can find me on LinkedIn. Um, and I've also got a newsletter and soon I will also have a podcast. So, yeah, a lot of exciting stuff coming around.

Speaker B: Cool. Thank you so much, man. Have a good one.

Speaker A: Thanks for having me, Paul. Yeah.