Michael Kaminsky - modern analytics and attribution, startups and more | E45

Hello listeners, and welcome to the latest episode of the AdQuick Madvertising podcast. A podcast that explores the intersection of technology, marketing content, all of the things that you care about. Today on the show we have Michael Kaminski. I believe I said that right. Michael is the co founder of a company called Recast. Very cool company. We're going to learn a little bit more about today. Um, Michael, thank you for joining us and spending the time.

Adam, thanks for having me. I'm super excited to be here.

Awesome. So, you know, let's, let's dive in and I usually don't lead with this question, but I think in your case it's actually really important for the rest of the analytics discussion is can you give us just a really brief overview not salesy, like sort of like what you're even doing at Recast and like what the problem is and why you even exist, why you even built this company.

Company? Yeah, it's a great question. So what we really care about doing is helping marketers measure the incrementality of their media spend. We think that that is the most important thing in marketing, or at least from like a budget management perspective, at least is to really understand, hey, for every dollar that we invest in each of our different marketing channels, how much return is that giving us? If you can answer that question, then, you know, sort of the world is your oyster in terms of being able to really optimize that marketing mix and drive a lot of profit from the business. So that's what Recast does. Recast does that primarily through a media mix modeling platform that we offer as sort of a software, as a service model, helping lots of different brands run, build these, all of these different models and then use them for planning and budget optimization purposes.

Yeah, and so MMM is a super big buzzword in marketing cycles. Lots of people are hype about MMM right now. Can you give like the, in plain English version of this for any marketers listening who aren't analytics guys just to let them know like what the concept is?

Yeah, so the concept is, is fairly straightforward. Media mix model basically just means top down statistical model. So when I say top down statistical model, I mean working off of aggregate data so there's no tracking of individual users involved at all, which is sort of the primary differentiator with something like an MTA multi touch attribution type platform. The idea is you aggregate all of the data and then you use a statistical or econometric model to find patterns in the data. That is what media mix modeling means at the very Highest level.

Awesome. And I'm curious for you in particular, what do you do to ensure that, you know, Recast has accuracy and reliability in the models that. That you're providing to users? And maybe even more broadly, how can companies, whether they're using you yet or not, can make sure that they are getting, like, a clear picture of what the data says?

Yeah, so this is such a great question, Adam. And I think it's a thing that there are a lot of misconceptions about. A lot of marketers have heard of media mix modeling, and they're like, oh, this sounds amazing, right? Like, I can take my data, I can plug it into this algorithm, and I get incrementality on the other side. And that is very much like the idea. When it works perfectly, that is very much what you get. The issue is that it pretty much never works perfectly. And so the reality is a lot more complex and honestly a lot more dangerous than. Than what I just laid out in that sort of idealized world. And so there's this misconception, right, that media mix models are sort of commoditized. They're like, oh, yeah, like there's this open source package and there's this other open source package and there's other vendor, and it's all a media mix model. It's all just a regression. It's all the same, right? We plug in our data and we'll get results out on the other side. And unfortunately, that's not true. These types of econometric models are very, very sensitive to the structure of the model and the assumptions that go into the model, which is to say that if you tweak the assumption slightly because you're using a different package or because you configure the package differently, you are very likely to get very different results on the other side. And so, again, this might sound like whatever data scientist gobbledygook, but practically what that means is you run the model one time with one set of very reasonable configuration options, and it says that Facebook is your best marketing channel. And then you tweak the assumptions very slightly, you run the model again, and it says Facebook is your worst channel. And which one of those is true? Is it the good one or the bad one? Or neither of them. And that is actually the key problem with doing this sort of media mix modeling is that we don't have any ground truth to compare the results to. We don't know if it was the first one that was right or the second one that was right, because we don't no one knows what the true incrementality of a dollar invested in Facebook is. And we're trying to do what's called causal inference, which is to say, infer things about reality that are actually unknown and unknowable using a statistical model. And so when you're actually thinking about doing media mix modeling and you're evaluating these different tools, you really need to think very hard about how are we evaluating if this thing is right or not. And the work that we do at Recast is actually mostly about thinking about that problem. We put a lot, a lot, a lot of effort into thinking about how do you know if the model is right or not? How do you interrogate it? How do you run different sensitivities and scenarios and see how the results change? How do you choose among those different candidate models to find the one that actually is most likely correct for your business? That's where, if you're thinking about going down this path and evaluating these tools, you really need to put a lot of emphasis on. It's very trivially easy to just take your data and run it through an algorithm and get results on the other side. It's very, very difficult to know if those results are correct or not. And so that's where I think, really. And unfortunately, this is like, you know, again, it's not the sort of sexiest part of these tools. Right. It's not exciting. You don't read a lot of, like, headlines about that. But it's actually the most important part of doing this work because we don't know what the ground truth is. And so you need to spend a lot of time and have a lot of very good tools to evaluate. Hey, did we get to the correct model or not?

Yeah. Yeah, that's really well said. And it's interesting because, you know, I. I just made a purchase for. I bought a gift for a friend the other week, and the way I did it is I didn't know what to buy the person. So I asked my wife for her thoughts, and she looked in the app of, like, this store that she always buys from, this quirky e commerce store, and she found the product and she's like, she texted me a link to it and it was like, hey, this would be perfect. So then I went ahead and didn't click that link. I went and searched in Google for the product and I found, like, a cheaper, you know, version of it online. And so now when I do make that, I made that purchase, that E commerce store that I did buy from, you know, what they're going to attribute is Google Search and but is Google Search the thing that caused me to purchase that product? No, it was, you know, my wife's query on this other app. And so my point was sharing is the bias of all of the analytics tools is still, even now, even though we know last click is not very meaningful. It could be sometimes, but it's just, you know, a very low resolution data point. The last click would get credit and so that store provider would be like, well Google's clearly working for me, we should give them more money even though they're really just capturing existing demand. And so I love what you're doing and other sort of unbiased third party companies are doing outside of the big analytics tools because they're always going to tell this very biased specific story also because they only have a really limited view of what happened prior to that purchase and I think it's going to get even worse. I guess another question for you is the type of analysis that you're doing, is it really only at this point still for large businesses because you need so much data and your spends are in the millions. So of course it makes sense to do everything you can to optimize and you have enough, you know, you have enough statistics to be able to say, well this is working versus a small business where it might just have to be vibes and you know, you know, other than like the things they see are working and you know, can have some sense because they only have so many marketing dollars at work. Is this still a tool for like large businesses or like what are your thoughts here?

Yeah, it's a, this is a really good question and I'll, I'll address one misconception first, which I think, I think is important. So an interesting thing about media mix modeling and this type of econometric modeling is that you actually don't need that much data. Right? This is an interesting problem because it's a small data problem. Multi touch attribution, digital tracking. That's where you need lots of data because you have every touch point for every user that's visited your website. Like that's a lot of data. That's a big table. You're going to even up for a small E commerce business. You're going to have millions of rows in that table. The data that you need for an MMM, a media mix model is often just like 600 rows, right? Like how much did you spend on marketing over the last two years? So 700 rows around that, 5, 10, 20 different columns, right? It's a very small table. It can almost fit like you can almost look at the whole thing on your laptop screen. So it's a small data problem, not a big data problem. You need to have consistent and clean data over that time, which can be a challenge for some brands, but it's not like, oh man, we need all of this data. It's actually sort of the opposite. You need a small amount of high quality clean data in order to do a media mix model. So what size brands should do this? This is a good question. You can build a good media mix model for a very small brand. Totally possible to do. Absolutely will work. The problem is that often the juice is not worth the squeeze. As I alluded to earlier in the conversation, building a good media mix model is quite hard. It requires a lot of specialization, a lot of sort of tough. You have to tackle a bunch of tough statistical problems, very easy to get wrong. And therefore it's often just not a good use of time and resources for smaller brands. If you're only advertising on Facebook and Google, for example, the additional value of a media mix model probably isn't that high for you. You should be focusing your time and energy on other things, working on creative, working on the offer, working on the product. All of those different things are probably going to be higher leverage for you. And you're like last touch attribution. Absolutely not perfect, but it can get you a pretty long way. Like I tell a lot of people who are in marketing analytics, like if you're sort of still a subscale brand, the report that you should be looking at is like first touch attribution, last touch attribution, post checkout survey results. And that's really it. Like you can build a pretty big business just doing that, especially when you're only on a few digital channels. So like, great, focus on that or just do that and then focus on other things as you're scaling your business. Once you start to get to scale and you have more mid funnel and upper funnel media, so you're doing more video, you're doing more audio, you're outside of just like Meta and Google. That's when makes sense to start exploring something like a media mix model. But generally we're talking about, you know, millions of dollars a year of annual media budget before that really starts to make sense. And that's actually like a good use of time and energy for your organization.

Yep, that makes sense. I, I didn't mean in terms of data, I meant in terms of costs, which, which you said the cost is.

The cost is real. And the cost is real. Whether you work with a vendor. Right. Recast is fairly expensive. All of the vendors in the space are fairly expensive. Or if you do it internally like you can do and you can use an open source tool, but like that ends up being more expensive because you have to hire someone who knows how to use it or if they don't know how to use it and you get bad results, that's going to cost you dollars because you're going to invest poorly. So it's like there's costs everywhere. There's no real getting away from that. And you're totally right. If you're only spending a couple hundred thousand dollars a year on paid media, it's just like the juice is absolutely not going to be worth the speaker, please.

Yeah, so I want to get back into your tool in a bit, but I want to talk more about sort of your background and philosophy of measurement so we can understand that a bit deeper because I think your background's really interesting. You were at, you built the marketing science team at Harry's and I'm curious of like how your experience there and what you saw because they're really successful brand and you know, you saw a lot of tests and you saw a lot of different tactics. How did that influence how you built this company?

Yeah, so Harry's was a really amazing experience, really amazing team that they have over there and very like forward thinking. So you know, Harry's was sort of part of that like last generation of VC backed D2C companies and one of the sort of few that, that sort of reached escape velocity. And it was really interesting to work there because I got to work with a team of marketers who are very highly, highly analytical. These, you know, a lot of like Harry's did this thing where they would like hire people from Bain and BCG and then come and make them marketers. And so you have these like deeply technical, analytical people approaching the marketing problem from that background and that whole business again. At the time when I joined, it was really like a DTC only company. It's now changed a lot since then. So this is all like very outdated stuff. But at the time it was D2C only. Cell raisers online. Facebook was still sort of a newish advertising channel. Podcasts were just getting started. And so there's a lot of opportunity to exploit these different channels and really efficiently build a business really quickly. But the whole team took a very aggressive approach to analytics. Thinking about, okay, what's the cost per Acquisition on these different channels, how is that changing over time? Where's their opportunity? How can we move budget from channel A to channel B to exploit that opportunity? What is our payback window? What's the payback window? Target? All of that sort of stuff was really in the DNA of the company. And the CEO was looking at these reports alongside the marketing team on a daily or weekly basis. And so it was like the whole company was very focused on this, very analytical and this idea of having a marketing team that's extremely responsive to the dynamics of the market. CPMs are going up, we need to adjust. We're going to go double down in this other channel because the CPMs went up on this channel. And being able to do that really dynamically in order to again identify inefficiencies in the, in the market for impressions and then exploit those really quickly felt like, okay, this is the future of where marketing is headed. And this is, you know, again, today this feels very familiar, but like six or seven years ago this was actually like very new, right? If you think about like 10 plus years ago, the way the vast majority of marketing dollars were spent was on linear television, right? And you go back 10, 15, 20 plus years, really the vast majority of marketing dollars were allocated one time a year at the upfront when you buy all of your media for television for the next 12 months. And so this idea of being like, we're going to dynamically update the budget on a daily or weekly basis to try to exploit inefficiencies in the market for advertising, we're was really sort of crazy. But today it's very common, right? Everyone listening to this is probably like, oh yeah, that's obvious. Like we do this across all of our different platforms. But again, back then it wasn't. But then you start to think about what are the tools that those marketers are going to need going forward, right? It's a very different set of tools than what you need when you're buying all of your media upfronts. You need tools that help you understand incrementality, but in an ongoing way on a daily or weekly basis so that you can take advantage of those inefficiencies, move around, shift your marketing budget, evaluate whether or not that actually worked as you expected, do this sort of ongoing experimentation and hypothesis testing. And it's that environment that first of all got me really excited, right? Because it's like, okay, not only do we need to measure, but then we're going to take action as well, but also got me thinking about okay, There seems like there's a real opportunity to build tools that could support that type of team in this new world of always on dynamic digital marketing. And now, right, even, you know, the video platforms, whether it's online video or streaming TV or whatever, YouTube, right, which is now sort of actually like has, I think now the majority of eyeballs in people's homes. Like, people just watch YouTube constantly. The way that that is bought is not bought once a year at upfronts. It's bought dynamically every day. Impression, you know, the cost per impression can move around. You can dynamically change your targeting setting. You can very dynamically change. Change what creative you're running. And so the teams now that are investing these really large marketing budgets need the tools that are much more dynamic. And that's the environment that my company recast was born out of.

So I love that. I wish all marketers thought like the world you described, where they take a more quantitative approach to where they actually buy media or even in some cases, just pay attention to the world. Because I've seen so many marketing teams which are still run by leadership that thinks all of their users watch TV every night and cable is still their golden goose and they're still 60 billion a year going to linear TV. I mean, some stupid number, I don't know if it's that high still. It's trending down. But our sector is probably split on the capitalists that are operating in, you know, in a world that you describe and are biasing to where users actually are and perhaps so sophisticated that they're exploiting, you know, they're finding arbitrage and, you know, cheaper buy somewhere and it makes sense versus the teams that are still doing the, you know, you're out planning and are still spending absurd amounts for, you know, dying TV networks or maybe even spending obscene amounts sponsoring, you know, influencers that, you know, like some baby boomer CMO wants to golf with. And they're just going to throw them, you know, $20 million. Right? It happens all the time. It happens all of the time. I see it all the time. I talk to other marketers who are frustrated when, you know, their leadership is kind of operating in an old world. So I wish we lived in the world you described. I think it's going to go closer that way. There will come a time where CFOs ask their marketing teams to be much more fiduciary and not just vibes based. Where I think the opportunity to be vibes based is in creativity. And when you understand, you Know the pulse of your market. Yeah, you should try vibes based stuff with, you know, your video ad should have the right vibes for users, it should increase sentiment, it should help increase sales. But you know, your media buy shouldn't be spaced. Right.

I, look, I totally agree with that and I was going to hit on that point, which is that sometimes people talk to me and they're like, oh, you know, this guy wants to take all of the creativity out of marketing. And that's just absolutely not true. Like, I have so much respect for creatives and the work that they do. I'm quite bad at that. Like, I can not generate good creative at all. And creative matters a lot. It matters so much and it's so important and it's a thing that you really need to get right. But you need to have a measurement system in place to avoid the things that you were just talking about where the dollars get spent frivolously because of ulterior motives that people have or even good motives that people have. But if they're just mistaken, you need to have that solid measurement system in place. And what that can do is it can actually really unleash the creatives to really start to say like, okay, we want to help the creatives find the message and find the creative that really resonates with the audience and actually drives the business forward. And that's the best thing for the business overall. That's what's going to allow it to grow and allow the marketers to get more budget to continue growing. And so overall it's actually a really good thing. Although again, some marketers sort of find it a little bit uncomfortable to have to think about measurement and have to live in this split brain world. This is what makes marketing really hard. We have to have both the creative side of, hey, how do we get a message out there that really resonates with people and really gets their attention? But also the numeric and analytical side of hey, how, how is this impacting the business? How are we driving it forward? Marketing is hard because you have to have both.

Yeah. And to be fair, on the creative side, we've seen any number of teams at large tech recently who walked back ads that were just strange because they were created in a room without, not even without data, just without understanding their market really well. Or maybe Bud Light paying certain influencers that, you know, we've had someone else, other people on the podcast talk about this where it wasn't, it wasn't even necessarily the approach, it was just how they did it. Which was the market found odd. So I don't know if I would look for salvation in the creative teams or some of the analytics teams from now. It's an interesting time for marketing. I think there's a lot of opportunity for really good analysts and for really good creatives. And I think these are the people that are going to help their companies get the most ROI and have the creative that people actually love and talk about. It's kind of similar happening, similar thing happening in the media world right now where you see big media companies just making more sequels that everyone hates because they create more dollars in the near term. But is this something that's actually like inspiring their market and having people excited to go to movies sometimes? I don't know. It's an interesting time and I think the crux of it is a lot of data being used poorly or data being used to optimize for the short term in both the marketing, the media side. And if you take like a better view of it, I think it all starts to make more sense.

So I think that. No, go ahead, let me. I'll add one point here just because it's sort of funny. Like, I am a data person. I've been a data person my whole life. I've run data teams, always been statistician, econometrician. That's the work that I do. I often find myself arguing against using data way more than people might expect. Often again with people who are, you know, CEOs or leaders. And I'm like, look, this is a decision that you just cannot and should not use data for. You do not have the right data to use it. If you try to use data, you'll be using the wrong data to make this decision and it's going to lead you astray. I think a lot of what you're describing is where people really want to be data driven, which I generally applaud. But you have to be smart about it. You have to think about where do we have good data for this, where can we use data and where can't we? And when you're doing things like strategic decisions, right, a lot of times you just don't have the data for that. And you need to, when it comes to doing, thinking about how are you investing in your brand, for example, that's the thing that often needs to happen at a strategic level. There just is no good data that can really guide you. And so you need to be making bets at a strategic level and then trying to use data as a crutch to shift that ends up leading you astray. And this is true again with the idea of like the movies that people are producing. Right. Like you can use data drive the short term, but that's actually maybe not the right long term thing for your business. I don't know enough about the movie industry. I actually have no idea. But you could totally see how that could happen. And so I think it's really important for marketers and leaders generally to think about what should we be using data for, for which decisions and which ones do we just need to trust our gut or trust our philosophy or our company vision and push ahead without data. And I think that, you know, the best companies are really good about differentiating between those two situations.

Yeah. And a lot of the bets with truly asymmetric upside were, are not predicted by historic data. And what would happen? Like, I think I was reading something, something like there was McKinsey predicted like the total mobile market for like cell phones would be like some stupidly small number in the 90s, I think they said like no more than 7 million people will own mobile devices. And then you know, that obviously turned out to be wrong. So they were looking, you know, at a historic trend and they didn't foresee, you know, a visionary like Steve Jobs taking something that actually already existed. A lot of people had BlackBerries and you know, I had a Windows CE device. I'm a nerd. And he just made it like, yeah, he made it more palatable for consumers. And you know, it's just there's like the mid twit meme that people like to use about, you know, using data versus, you know, sometimes some of the things are obvious. Like to me, even cell phone adoption was obvious. But it's not something you would see in the dashboard. It was just something that was going to happen. And I worry about companies that become so, you know, too data driven to the point they, they miss the opportunities right in front of them. They might, you know, it could be something small like, you know, optimizing a product and removing a feature that just so happens like you're 10% of your, your most loyal users loved it happens all the time in tech. And then suddenly everyone starts to hate you. Or it could be something like, you know, your, your big corporation lays off a huge customer service team. Cause it was a big cost center. And it turns out the customer service team was what was accountable for bringing a lot of repeat business. So there's a lot of ways to use data really poorly that make users hate you. And you know, you Gotta avoid that in marketing or, you know, operations, whatever you're doing.

100%. I could not agree more.

So I wanna talk about, you know, your background in econometrics. You have a background as an economist, which is super cool and I would say pretty rare for someone building a marketing company. Usually us marketers have like, you know, our marketing analytics background, which is we run a B tests and you know, we understand statistical significance of, you know, really basic things. We're not, you know, doing cohort analysis. We're not, you know, like, like a biochem student who would be following people and watching viral knockdown over time, things like that. So how do you apply the principles that you've learned to marketing analytics at Recast?

Yeah, it's a really good question. So the sort of through line of my career is that I've always cared a lot about doing causal inference, which is to say, how do we learn things about how the world works from the data that we observe? And you can, you know, I've applied this in the field of environmental economics where did some research looking at, okay, when the pricing schedule for water changes in Tucson, how does that impact people's consumption habits? Right. We wanted to understand what's the price elasticity for water in a desert city. And so you have to collect the data and then do a regression analysis and look at the coefficients and understand them. And it's that sort of work. And that was sort of one of the very earliest research projects that I did in my career in academia. I did similar work in the healthcare field. I've always just been really passionate about use data and then use that to really understand something about how the world works that isn't necessarily directly observable. And it turns out that, you know, I love doing that. And it turns out that marketing has a lot of these really interesting problems, right? It's the same sort of problem, which is to say, okay, when you give, you know, someone treatment A versus treatment B, what's the outcome in the healthcare field is very similar to like if you show someone ad A versus ad B, do they purchase or not or how does that change their likelihood of purchasing in the marketing field? So you have a lot of these very similar problems, but it's a lot of these similar problems in a field where there's first of all a lot of just like economic activity, there's a lot of dollars that are invested in marketing. It impacts the future of lots and lots of different businesses. So there's like a really big impact there. And you get to operate in this world where, like, experimentation really is possible. Experimentation is quite difficult to do in the healthcare field. Like, running a clinical trial is a huge pain. It takes years. You spend. You can spend a decade, like, doing research, setting up your clinical trial. You run one test and it fails, and it's just like, it's all over. Marketing is so much more dynamic and so much more exciting that it's really fun place to be doing this, you know, this work that I really love and care about in causal inference and statistics. But getting to apply it to a world where people actually take action, right? People use recast to drive experiments, to drive their business forward, to generate huge returns for their business. And that's just really fun to see. It's a really fun and engaging atmosphere where you get the intellectual side of how do we do this really, really hard and complex statistical problem, and then you get the benefit of getting to see it actually put into action, like four days after the results are in the water. It's not like you have to wait five years to see what happens with the clinical trial. You get to see people using it and actually using it to build. To build businesses and drive them forward. And that's just incredibly rewarding for me.

Yeah, I think it's cool what you're doing and what everyone in ad tech is cool because basically, in a lot of cases, especially for analytics, you're doing a lot of the hard statistical thinking, coding, all of this stuff for the marketer. So now the marketer has a set of tools where they basically have more of an easy button or not having to have a whole data science team to help them understand what to do next. And I think it, you know, we talked about creativity versus, you know, analytics. I think the good tools basically do all of the hard analysis work and coding for the marketer. So then the marketer's whole time is spent on the creative and fun aspects. And I think, like, really good teams have processes there where it's pretty clear. So, you know, the marketer isn't asked to create some crazy report with an insane deadline for their budget, you know, that they have to come up with next week. That's kind of the old way versus, you know, actually having a set of tools and a process to say, you know, here's our path of experimentation, here's our series of tactics for the next quarter. You know, it's kind of already laid out. And it's not just we're guessing as to what we're doing next and always.

Spinning our wheels yeah, 100%. And I think it's, it's, it's really fun to get to build these tools that let marketers do more of what they want to be doing, what they're really good at, and less of that, like spreadsheet munging, trying to put the budget together for the finance team, like that's such an unpleasant process. And so I, it, it really is fun to get to like do the stuff that I love doing and then allow marketers to do the stuff that they really love doing.

Yeah. And look, you're gonna. If the more you force creative people to work with numbers and data in a way that's annoying, I'm not saying that can't be creative work. You're just taking them away from like the high value thing that they could be doing for your company. So yeah, let's talk about people a bit more. You have a unique approach to hiring talent and so I'm curious how you find people for your team, whether coders or analysts or even marketers. Like how do you source talent and what have you learned that maybe you could share for? Probably other startups I think could learn from you and what you've done because I know you guys are growing quite quickly and everything seems to be going really well. So like what, what could other startups learn? Maybe not competitive startups, but just even other tech startups.

Yes, I think in general, you know, we. Well, so first of all, I'll say the same thing that everyone says, which is that at a startup the talent matters more than anything else. Right. And so if you are running a small company, the people that you hire are just so incredibly important to your success that you need to put a lot of time and effort into to getting really, really good people on the team. At Recast. Very early on, my co founder and I made the decision that we don't want to do a lot of active management. And so basically we don't want to spend a lot of our time doing a lot of training, a lot of oversight. We want to hire people who come in, set their own direction and self manage from day one. And what that implies is that you need to hire people who are pretty senior and pretty good. And so we basically made that a policy of we're mostly going to hire people who are very senior, very experienced, don't need a lot of oversight, training and management. They've been in their careers for 10, 15, 20 years already. They know what they're good at, they know what they're bad at, they're able to operate really effectively independently. We can set the goals at the high level and they're going to be able to go and execute on that. And that has worked really well for us. Again, it takes, you have to have the right attitude and approach to management. Those people, you know, they want to be independent and so if you're sort of by nature a micromanager, that's not going to work as well. They're also expensive. And so you need to be very thoughtful about how you're going to finance the company if you're going to do it this way because you're going to be paying a lot more per head. Hopefully it works out that they're more effective and you're saving on management costs. And so there's, there's savings that you get. But you know, there can be some sticker shock when you talk to some of these people who are 15, 20 years in the career and really, really good. Like they know what they're worth and you're not going to be able to like it's not, you know, like you're hiring some 24 year old, two years out of school and you can sort of slip one by them and pay them a below market salary. Like that's not going to happen with these people that we hire. So overall that's been our philosophy. Again it's worked really well for us. But that is related to how my co founder and I like to manage and the sort of company that we've been trying to build.

I love that. I think that if I ever build a company I'll do something similar. And I think for more senior talent, especially in tech at this point they're probably, if not totally financially independent. If you're not doing well after the greatest bull market of all time, maybe you just had some awful bets and you tried to become an angel investor and realized that that was too hard. So know that probably happened to those people. But I think what's more interesting, at least for me, like why I joined, you know, my team is, is probably equity and having skin in the game and like I'm okay with probably a below salary of what my peers are getting but I have like asymmetric upside and inequity. So I think that for me that was more interesting than going back to a big company where I'm buried in an org chart. Now I'm part of a pirate ship and I'm not micromanaged and I also have skin in the game and so I feel like an owner and so like I Think like that's how we're doing it at AD Quick. But I think that other startups like that ethos just works so well. It's like why so many startups had great people. I probably stayed in the corporate side of my career too long and it's just, it's like comfortable. Right. Like you get used to, you know, the fang equity which is probably going away. Honestly, I think like they fix the glitch on it. So yeah, I think that's how you get great folk and certainly don't try and nickel and dime them. But yeah, give them skin in the game. It's like the way to do it.

Absolutely.

You mentioned also on your site, I think I was going through and doing some research on you guys and you talk about building a sustainable business from day one. What does that mean for, for recast? How does that, how does that manifest?

Yeah, so for us it's just always been important that we have an economically viable business. We wanted to first of all from day one sell something. So like a lot of venture backed startups will do this thing where they'll give away the product for free for a while and then hope to be able to convert that into something that pays and is economical. And we never did that. From the very first day we were looking to have customers pay us to validate that, hey, this is something that is actually valuable and creating value for their business. Beyond that, we also wanted to make sure that the business was sort of like unit profitable. Right. You can scale to a large top line number if you sell dollars for 90 cents. Right. Because people will come in the door and they will take that all day long. It's a great deal for them. But that's obviously not a profitable business at its core. And so what has always been important to us is to say like look, you know, we can take outside capital, that's great, we can use that to fund development. But the core product that we're offering has to be profitable always. Otherwise we're not building in the right direction. And so that is when we think about what makes it a sustainable business. It's really those two things. It's that people are willing to pay us. It means that it is creating some value for their business. They are receiving more value than what they're paying us. And then two, that it's on a unit basis. At the very least economically profitable. Like this business can continue to grow. When we are adding new customers, it is adding additional free cash flow to flow to the business. And we always want to make sure that we keep that front and center. We could grow top line by like cutting our prices a lot but then it wouldn't necessarily be profitable and that wouldn't be sustainable and we'd have to like fingers crossed we can somehow flip the switch to profitability down the line and we didn't want to put ourselves in the that position.

Awesome. This is, you are another company that goes into my folder of Zerp is over. So you know, no more. Basically just trying to run things at a massive loss. I, I, I think some things like AI tools that have, you know, massive numbers can probably do it for a little while but you know, money's no longer free so we all free.

But I mean, you know, there is like there's a reason that venture capital exists like SaaS tools. You have to spend a lot of money on expensive developers up front, build a tool and then hopefully get paid back in the long run. That's not a crazy way to operate the business, but it does have to be unit profitable. You do have to make sure that you are getting, you are not selling dollars for 90 cents. You have to make sure that you're getting some amount of profit on the unit basis. Otherwise you end up in the really bad situation.

I do miss the like $10 Uber rides when I was, I also missed.

That there was, it was a really.

Good time for a while and there was also a time period I think I ate for free from Blue Apron for like half of a year by signing up, getting a few, you know, getting like six free deliveries and then unsigning up and then signing up again. And there was just a glitch of infinite free food from Blue Blue Apron.

So it's really interesting that tale of two businesses, right. Blue Apron I think went out of business. Yeah, I think it sold or something.

Like a penny stock and got acquired.

Yeah, bad, bad situation for Blue Apron. But like Uber doing well, they were able to flip that switch. Right. And now I was just looking at their, their most recently quarterly financial report. Their ad business alone just their ad business is now doing a billion dollars in top line revenue run rate. Like that, you know, might be the story where it actually worked. There's a lot of businesses that tried to follow that path where it didn't. But Uber is an interesting example where like sometimes this idea can work where look, you can get to scale if you have this sort of like monopoly type business with network effects. Like maybe it does make sense. It doesn't make sense everywhere obviously, but like it's a really interesting example of like, yeah, it maybe did end up working out for them.

Yeah, they also solved a really hard problem in a better way than anyone else. I mean, it's not like I have a problem getting food or even food delivery. Okay, great. You can order it online. Well, so what? How many other different places can I order food from? Yeah, the blue apron thing was, was interesting. Great. Well, I'd love to end on because, you know, you mentioned you like to hire really experienced senior people. I think that's smart. We, we do have a lot of young marketers and practitioners. Can you maybe give them, I guess, a few tips on maybe how to be better analysts or think like analysts to help them throughout their career? Because I think even, you know, even the creative types can, can learn a bit more here, especially from someone like yourself.

Oh man, this is a really good question. Let's see, what advice would I give? So I would say, you know, actually the. For, for people who are more junior career, don't be afraid to ask questions. That's actually the thing that I would, that I would. The best piece of advice that I could give is ask lots of questions. And especially in the world of marketing, like there's a lot of, you'll run into people who will, you know, have some analysis and they sort of don't want you to ask questions. They don't want you to go that level deeper. But you should feel very comfortable doing it. Ask the questions, you know, with good vibes. You don't need to attack people or anything along those lines. But ask questions, get to that level deeper. It will either help you understand better, which will be great for your long term understanding and career, or it will expose potential flaws in these analyses. One thing that I think is really important to keep in mind is that most analyses are flawed. Right? I've been doing analyses my whole life. It is so easy to get things wrong, to screw things up. The only way that you have a chance of really getting to the truth is by really like pounding the analysis from a bunch of different sides with questions. How could it be wrong? What are all of the different ways that this could be wrong? Are we comfortable with that or not? Do we need to do follow up investigations to understand if this is wrong or not? So I would say ask those questions. And as you ask those questions and hear the answers and think about the ways that the analyses can be wrong, that's really the best training that you can go, that you can go through. The best analysts in the world are the ones who are really good at doing that thought process themselves, right? They think, they, they proactively ask themselves, how could this be wrong? How could this be wrong? Can I, can I rule out this thing? Can I rule out that thing? Am I comfortable with this assumption? That's what the really best analysts in the world do. You don't need to be particularly quantitative to be able to do that. I'm actually like not especially quantitative. All of my, you know, the people that I work with are like really good at math, that I'm not particularly good at math. I'm very good at reasoning and thinking about how an analysis can be wrong. And, and that doesn't require particularly quantitative skills. And so you can do that. And that's probably the most important skill that you can develop, especially as you're more junior in your career.

I love that. And I think the other side of what happens there is you become not afraid to change your mind in public. And I think people who can do that, you know, in meetings be like, I was wrong. Those people are such a pleasure to work with because they don't just hold on to, to, you know, their hippo as we call them, highest paid person's opinion idea and, and they're willing to compromise and, or you know, take, take blame for things. You know, they're just such a pleasure to work with. So I think you develop that as well when you take on that mindset. Well, great. Michael, this has been really interesting. You've been a pleasure to interview. Can you tell us where to find you and where to connect with your company? We'll also put links in the show descriptions.

Yeah, so the place where I'm most active right now is on LinkedIn. So you can find me. I'm Michael Kaminsky on LinkedIn. You can find me there. Recast has a great YouTube channel that we're putting a lot of content out on. So if you like these sorts of interviews, we do discussions with really super technical people, we do code walkthroughs and then some just like fun conversations with, you know, people who are really interesting in the marketing world. So those are the two places that I check out. And you can check out recast website@getrecast.com and again there's a ton of great material. We have a blog, we have mmm Academy if you're just looking to learn. So a lot of really great resources for people who are, you know, again, not even looking to purchase, but really just looking to learn and start to understand how these different approaches to marketing measurement really work.

Great. Well, we appreciate your time today. And for listeners, we will see you next week with a new episode.