Michael is a trained econometrician with experience in causal inference in fields ranging from healthcare outcomes to environmental economics. He previously built the marketing science team at men’s grooming brand Harry’s before co-founding Recast. He spends his days helping brands eliminate their wasted marketing spend…

Questions and topics we covered include:

• The different methods of measuring marketing’s performance (attribution is only one of many methods)…
• Do attribution tools lie to you?
• The difference between “attribution” and “incrementality”
• The relationship between measurement and forecasting when it comes to marketing
• How to use Marketing Mix Modeling (MMM) alongside Multi-touch attribution (MTA)
• How to get your work measured if you’re not a performance marketer
• Will 100% certainty in measurement ever be possible?

And more!

Connect with Michael on LinkedIn – https://www.linkedin.com/in/michael-the-data-guy-kaminsky/ 

Connect with me on LinkedIn – https://www.linkedin.com/in/kennysoto 

Full Episode Transcript:

Michael Kaminsky  0:00  

There should be no marketing dollar that is not held accountable to some goal, right? When we’re doing attribution in digital tracking, right, we’re doing that because we think that it has some relation to increment out. If it didn’t, we wouldn’t do it at all. When we run experiments, same thing, we do historical trend analysis, same thing. But none of these measurement methods perfectly capture incrementality.

 

Kenny Soto  0:24  

You are now listening to the people of digital marketing, the number one resource for marketers who want to both impress their boss and eventually become their boss. And in today’s episode, we are talking to Michael Kaminsky. Michael is a trained econometrician with experience in casual inference in fields ranging from health care outcomes to environmental economics. He previously built the marketing science team and men’s grooming brand Harry’s before co founding recast. He spends his days helping brands eliminate their wasted marketing spend the best way to describe what he does and what he knows. And in today’s episode, we will be talking about attribution. And prior to our conversation to Michael, I always used to think that attribution was just one thing. But apparently, there are multiple methods of attributing marketing dollars to revenue. At the same time, he covers this tricky word called incrementality. And you’ll notice in this episode, unfortunately, sometimes when he says the word, the audio fades out here and there. It doesn’t happen all the time. But about like four to five times in this episode. Unfortunately, the audio fades out, just wanted to give that quick disclaimer before we start the show. And if you’re a marketer, who is trying to learn more about marketing measurement, especially as we enter a world where there is more and more difficulty with tracking, marketing success, this episode is for you. So without further ado, let’s listen to our conversation with Michael Kaminsky.

 

Kenny Soto 2:11

Hi, Michael, how are you?

 

Michael Kaminsky 2:14

Hey, Kenny. I’m doing great. Thanks for having me.

 

Kenny Soto 2:17

We were just talking, pre recording about Dallas traffic and all the fun that that that entails. And today I want to talk to you about a subject that I’m I’m trying to gain more appreciation for even though it’s the bane of my existence. Before we jump into that subject, I want the listeners to get a better sense of who you are as professional. Could you tell us the story of how you got into digital marketing?

 

Michael Kaminsky 2:47

Yeah, absolutely. So my background is actually in econometrics and statistics. That’s always what I’ve been really interested in is how do we learn about how the world really works? I started my career doing environmental economics and then moved into healthcare research. And then from there moved into the startup world where I spent a lot of time working with a broad variety of teams on data analysis on a metrics causal inference, and just sort of randomly ended up spending a lot of time with marketers broadly and digital marketers, specifically working with them to help them understand, okay, what are the actual levers that we can pull here to drive the business forward. And so, you know, through that partnership, and then through a bunch of the research, that I’ve done my own reading the best of the marketing science literature, I’ve just gotten to really know this space really well, and then spent the last couple of years building this company recast, where we’re in the marketing mix modeling space, which is effectively doing causal inference on marketing. So been spending really like the bulk of my career now focused on how do we learn things about what types of marketing, it’s effective.

 

Kenny Soto 3:56

So for the listener, who hasn’t caught on yet today’s topic of conversation is attribution. In my eight years of being in marketing, attribution has always been something that I found very difficult to grasp.One from a setup perspective, which is a whole different topic in and of itself, but also from a way of using it effectively over time, both for single channel attribution as well as omni channel. So taking a step back, I want to start off by just getting a lay of the land, if you will, what are the different ways of measuring marketing performance?

 

Michael Kaminsky  4:33  

Yeah, I’m glad you phrase phrase the question this way, because, you know, the word attribution is a little bit overloaded, right? A lot of people use the word attribution to really refer to marketing measurement. But I think that’s a little bit misleading. And so I like to talk about marketing analytics and marketing effectiveness measurement as a whole of which one type of doing that is attribution. So attribution, right if you think about what the word really means, in general, it’s this idea of we We have a sale, how do we attribute that sale to some marketing channel or set of marketing channels. Most people today, especially most digital marketers, are familiar with this type of digital tracking, attribution. So what we’re going to do is we’re going to try to track people across the internet, their activity on different devices, whatever they’re doing in their normal daily lives, and then assign credit to some conversion event back to one of those channels or multiple students. And that’s this idea of attribution. But I want to, I want to really emphasize that that’s one way of doing marketing measurement. Right? There are other ways of thinking about doing marketing impact measurement from surveys and polls, to running experiments to doing statistical modeling. And all of those are different types are different ways of getting at what’s truly effective. One of those ways is attribution or digital

 

Kenny Soto  5:58  

tracking. And would it be helpful? Or would it be accurate to assume that all of these different ways should be leveraged in different points of accompany sunyer?

 

Michael Kaminsky  6:11  

So I think that that’s definitely right, I think mature companies do all of them, right, because for mature companies that are spending lots of money on marketing, it’s worthwhile to try to use different methodologies in order to understand what’s truly happening. All of these different methodologies for marketing measurement, have different strengths and weaknesses. And so mature organizations want to use all of them understanding the strengths and weaknesses of every different type of, of measurement, and then use that to build a picture of what’s truly working and what’s not.

 

Kenny Soto  6:48  

I always struggle with measurement, specifically, because of two reasons. One, I find that even if the data is showing you a trend over time, sometimes there’s an anomalies week, over week or month over month, because you’re looking at historical data. And that historical data can take into account changes in culture, changes in market dynamics, changes in business dynamics, whether it’s headcount reductions, or new headcount or lack of resources, or more resources, etc. So there’s that the fact that when you’re looking at statistics, it’s always historical, you can’t really, you can use it to help you infer what might happen in the future. But at the end of the day, it’s still a guess. And then to on that same vein, is taking an intuitive approach, which I see a lot of creatives and brand marketers to, where instead of just solely focusing on the data, sometimes to the detriment of the performance marketers that work with them. Brand, marketers tend to go with their gut, and their feeling, saying, Here’s what we understand about our current customers. Therefore, we can use these insights to then create the right creative for specific channels are. So I say all this to say that, from your history, and from your perspective, how much should we be relying on marketing measurement?

 

Michael Kaminsky  8:17  

So it’s a really good question. So I think, you know, the short answer is a long, okay. All right, I think we should definitely be using a lot of marketing measurement and thinking really hard about it. But you’re also right, that there are fundamental limitations. And so I think really good marketers are really good marketing. And analysts are very thoughtful about exactly what those limitations are, and how they might be being led astray. One thing that I like to to try to get marketers to focus on is the idea of incrementality. The word incrementality. Also overused also can be confused a little bit. The idea behind incrementality is really just means causality. Right? When we do some more of some marketing activity, how many additional conversions or how much additional revenue does that draw? That’s really the fundamental question. And that’s the fundamental question of all of business really, right. Like, if we do this activity, how much does it truly get us? That’s the thing that we really care about. And as marketers, it should be the thing that we’re most focused on. Right? If I didn’t spend that $10,000 On that marketing campaign, what wouldn’t have happened otherwise, right? And if the answer is nothing, like the business wouldn’t be any different, well, then that means that that $10,000 wasn’t very effective. If we say if I hadn’t spent that $10,000, we would have gotten, you know, $100,000, less revenue, well, then that $10,000 was really effect, right? It was a 10x incrementality to it. And so incrementality is the thing that we really care about causality is the thing that we really care about. All of these other marketing, measurement methods are just ways of trying to get at incrementality. Right, when we’re doing attribution in digital tracking, right, we’re done doing that because we think that it has some relation to increment out. If it didn’t, we wouldn’t do it at all. When we run experiments, same thing, we do historical trend analysis, same thing. But none of these measurement methods perfectly capture incrementality. And so as we’re going through these measuring methods, what we need to think about is, how close is this to incrementality? And how might it be misleading, right, there might be other things going on at the same time, other business trends happening, that are impacting our ability to look at the retrospective data and get a read on incrementality. But incrementality should be the thing at the center that we’re always focused on trying to measure, I

 

Kenny Soto  10:37  

might be connecting two different concepts here. So let me know if there’s a distinct difference. But when you mentioned incrementality, what comes to my mind is confidence levels and statistical significance. Are they similar? Are they are they related?

 

Michael Kaminsky  10:55  

I think those are I think those are two very distinct ideas God incrementality really is about it’s just this idea of causality. How much additional business are we driving with each additional marketing activity. And that core idea is really important. And you you can assess incrementality with statistics, you can assess incrementality other ways as well. But the core idea is the thing that’s really

 

Kenny Soto  11:21  

got it got it. And when it comes to just taking a step back when it comes to attribution, I heard on a conference the other day, who was involved 2023, shout out to them. Rand Fishkin was one of the speakers and he said that attribution systems lied to you. Google is lying to you. Would you agree with that statement?

 

Michael Kaminsky  11:44  

I don’t know about lying. Pretty morally heavy word. But I would say that it often that they aren’t measuring incrementality. Right, not directly. And that is the problem. And so I think when people like Rand Fishkin say that your attribution tools are lying to you, what he really means is that they aren’t necessarily measuring incrementality. And this is the thing that’s really difficult for a lot of digital marketers is, there are some cases where it actually does measure incrementality, or where it’s gonna be very close to an incremental measure. And then there are gonna be other cases where it’s not. And I mean, I think this is a this is a phenomenon that lots of digital marketers are familiar with when we talk about branded search. Right? There are lots of reasons to believe that in many cases, branded search spend is not increment. Why might that be? Well, by definition, branded search means people are searching for your brand, which means that they have already heard of your brand. And so are using that channel to get access to your website, right? People are very used to going to Google and typing in some brand name, and then clicking on that top link, because they know that that’s going to take them where they want to go without needing to know the actual URL. And so if you’re paying Google for all of those clicks, from people who are planning to purchase from you anyway, then there’s a case to be made that that spin is not truly incremental. There are other cases where that’s been might be. And so, but I think the thing that Rand Fishkin wants you to think about, and that I want marketers to think about, is, when might it be incremental? And when not? And how can we get more clarity about whether it’s incremental, or it’s not for our particular business at the particular time that,

 

Kenny Soto  13:30  

hey there, if you’re enjoying this episode, and you’re a first time listener, when I hit the Follow button, my goal with each of these episodes is to introduce a new marketing concept, or dive deeper into one, so that you can become a better digital marketer. Hopefully, through these episodes, you join me on this journey, the path to CMO. So I’d love it if you subscribed. Thanks for listening so far. That’s the magic question right there. I find that this this is an onion, if you will, I like using this analogy where there’s so many layers, that you can just keep peeling away. And I want to peel another section of this onion, if you will, where it comes to the relationship between measurement and forecasting, when it comes to marketing. Can you describe that, that relationship between those two concepts? Yeah,

 

Michael Kaminsky  14:27  

really great question. And actually like probably goes much deeper and more vital than you even expect in this answer. So, measurement and forecasting. Ideally, if our measurements are giving us true incrementality true causal reads on how our marketing spend or activity is performing, then we should be able to use those reads for forecasting into the future. Right if we look at our Google Analytics report on attribution and we took All of those numbers at face value, we should be able to say, Okay, we’re getting an ROI of 3.4x on Facebook and an ROI of 2.5x on search and an ROI of 10x on branded search. And then we should be able to make a budget for next month, multiply those numbers together the budget for next month times those return on investment numbers from the Google Analytics report or wherever. And they get a forecast for how much revenue we’re going to do next. If those numbers are truly causal, that forecast should be perfect, effective, ideally, ideally, ideal. Now, you know, marketers who have gone through this exercise recognize that that’s probably not true, they have gone they’ve done this in Excel, and the forecast has not come through or given wildly implausible results, which just means that those numbers aren’t picking up the true causal, incremental reads of how effective those channels are. Because if we knew how truly effective every channel was, we should be able to predict the future.

 

Kenny Soto  16:02  

Yeah. And it seems to make sense. Well, then let me ask you this question. To what degree are forecasts necessary?

 

Michael Kaminsky  16:11  

Good question. So for some businesses, they’re very important, right forecasts are very important for if you have a physical goods business, you need to forecast inventory, you need to forecast how much you’re going to be able to sell. And so forecasting is actually a core business function. Because you want to make sure that you’re not going to go out of stock, or that you’re advertising the right things, and that you’re going to be able to maintain your business going. And so forecasting is really important in a number of different contexts, depending on the realities of the business. In other contexts, it’s not that important. But it can be really useful to validate if we actually have causal or incremental understanding of our marketing channels. If we do this forecast error size, and we miss widely, it means that we actually don’t understanding. And therefore, we need to do more work on understanding that incrementality in order to be able to actually optimize our business. Because if we’re making decisions off of reads of marketing, effectiveness, cost per acquisition, or return on investment that aren’t incremental, it means that we’re not actually efficiently optimizing our marketing dollars across our channels. And so the forecasting exercise, while it might not be business critical, it’s actually very useful for understanding our marketing performance and making sure that we have that true incremental read on how every dollar is, for us. Every

 

Kenny Soto  17:44  

business has so much nuance that this answer, I asked it, I know it means like, it’s It depends, really, but I still want your opinion on this, when it comes to forecasting specifically at what stage your business as a marketing team needs to start doing this. Because I I’m only assumption that if you’re like a seed stage business, forecasts really aren’t your priority, you want to get product market fit, but post Product Market Fit once you start doing forecasting?

 

Michael Kaminsky  18:10  

Well, I guess like, again, to the extent that you’re using forecasting, to evaluate how effective your marketing measurement is, you should start doing it when you’re spending significant amounts of money on marketing, and you’re spending money. And so it’s really about, again, to the extent that this is an exercise to understand is our marketing measurement, actually getting incremental reads, you should start doing it as soon as you’re starts starting to spend meaningful amounts of money on marketing, because that’s the point at which you might be meaningfully wasting marketing dollars. And so I would encourage any business that is spending meaningful amounts of money on marketing to start thinking about this exercise of how do we validate increment? And how far away from the group mentality are our measurements today, and you might run this exercise, if you’re a small business and be like, Look, every dollar is super incremental, the Facebook reports are a little bit off but close enough that we’re not going to worry about it. And then you say, like, thumbs up? Great. We’ll keep doing this. And then we’ll check it again in six months. But in six months, you might be like, Oh, wait, something’s really off here. If we just multiply these numbers through, we get totally different performance than what would be implied by a true incremental read. And so therefore, we need to start thinking harder about how we’re going to measure the incrementality of these different channels.

 

Kenny Soto  19:21  

Let’s talk about two other concepts. One I believe listeners are familiar with while as the other one, I’m still trying to learn myself. There’s multi touch attribution. And then there’s marketing mix modeling. Could you start off by explaining what each one is? And then talk about the differences between the two?

 

Michael Kaminsky  19:45  

Yeah, absolutely. So multi touch attribution is in the family of digital tracking and attribution methodologies. It’s right there in the name. The idea is that we’re going to try to track people across the Internet and assign credit to one or multiple of the touchpoints they engage with prior to some conversion, tried prior to some conversion. So, you know, you could imagine customer journey where they see a display ad. And then they see a YouTube ad and they click on it. And then two weeks later, they see a Facebook ad and they click on. And so multi touch attribution could use just a latch, a last touch attribution, in which case, they would give all of the credit to that Facebook ad, which was the last thing that user engaged with. Or they could use a first touch attribution. And they give all credit to the display ad that they saw and didn’t engage with. Or they could give, you know, 30% 30% 30% to all of those three, display ads, YouTube, and then Facebook, or they could give 80%, to Facebook, and then 15%, to YouTube, and then 5%. And there’s all kinds of different ways of thinking about allocating credit across those different touches. And multi touch attribution is just this idea of, well, let’s give credit not to maybe the first one, or maybe the last one, but some combination there of the different touches. And there are a number of vendors out in the world who, you know, talk a lot about these algorithms that they have for assigning credit across those different I’m honestly like, pretty skeptical about the value add of that activity, right of like doing going from first touch, or last touch to some other combination, I don’t actually think it adds very much. I think that if you have a first touch report, and you have a last touch report, between those two things, you have 90% of the value that you’re going to get from digital tracking at all, and then adding on some other more sophisticated method for assigning credit among those different touches doesn’t really get you very much. And the reason why is because, well, all of these digital tracking methodologies, they sort of make the assumption implicitly, that you can observe every single marketing touch point, and you have the universe of those marketing touch points. So in the example that I gave about the display ad and the YouTube ad, and the Facebook ad, well, let’s imagine that they also have a connected TV, where they see ads, but not on the same device that the other journey. And so there actually is this other touch point or a billboard, or billboard outside their house, or there’s linear TV ads, or they see an influencer on Instagram, right, all of these other things that aren’t actually picked up by that multi touch, multi touch attribution methodology, they’re not being counted at all. And so this idea of like, let’s get really smart about whether it’s 3030 30, or 4060 25, or whatever, that is totally missing the point, because the problem is that we’re actually missing a bunch of these touch points that weren’t captured by our tracking methodology. And those are the real problems in this methodology. And so getting really fine grained about how do we allocate between the three different touchpoints, we observed feels like we’re just totally missing the forest for the trees, because there’s three or four other touchpoints that we didn’t observe that we’re just totally ignoring and assigning zero credit. So that’s a multi touch attribution and the flaws with it and like the real problem, and again, you know, I want digital marketers to sort of think about in this new world that we’re entering into, where there’s privacy regulations around what sort of tracking can be done the iPhone, which is getting sort of more and more severe. And so we are observing less and less of the true depth, there’s, Google Chrome is going to get rid of cookies next year, which means that we’re going to miss even more of those touch points, right? All of these problems are getting worse and worse and worse. And so this assumption that we’re making in the within the world of MTA multi touch attribution, they’re just getting worse and worse and worse, we really, I don’t think that anyone really believes that we’re getting anything close to 100% of the actual touch points. And that means that all of that methodology is really fundamentally flawed, except for the most simple businesses that are only operating on, you know, one marketing chip. So that’s MTA one, right? That’s that world. One concept marketing mix modeling is another concept that takes totally different approach to measuring incoming right, marketing mix modeling, is a technology that was developed 4050 years ago, right before the invention of E commerce before the invention of the air. And if you think back, if you think back to being a marketer at that time, right, let’s imagine that you’re a CMO of Pepsi in 1985. There’s no e commerce, there’s no digital tracking cookie. The word only refers to the things that you make in the kitchen. The internet exists, but there’s no ecommerce there’s no idea of selling Pepsi. And so if you’re the CMO of Pepsi at that, how are you going to measure marketing effective? Right, you have a couple of main channels that you’re operating in you spend on TV, you spend on On out of whom advertisements you send out in print, you spend on radio and then you spend on like in store advertising. And so, if you’re in that world, well, you don’t just throw up your hands and say, I have no idea I’m just gonna make up a number and spend that on on all these different channels, which you actually do is you hire an econometrician are a statistician, to look at your historical data, of when you spend more money on different marketing channels. How is that associated with the additional sales that are being driven? Right? So on a weeks where we spend more on TV, do we sell more Pepsi? And how much you run that analysis? You get some understanding of, okay, when we tend to spend more money on these marketing channels, how much additional revenue does that drive, and then you use that to calculate your return on investment numbers. And then you use that to make a decision about how you’re going to allocate your budget across those channels. So the idea is that it’s a top down statistical model, looking at historical data, to try to find those relationships, econometric ly, right. What’s the what it when if we were to spend more money, what do we expect would happen via a statistical model without tracking any individual human. And that’s the idea behind marketing mix model. And so this is a technology that, again, was developed back before the world of digital tracking existed at all. But now it’s sort of making a comeback, because marketers have started to recognize the limitations of multi touch attribution or digital tracking. And so are turning to this methodology in order to understand without needing to track any individuals at all, can we get a read on incrementality, just by looking at our store data. And as you noted, right, there’s lots of tricky things about marketing that a good statistical model needs to take into account, right? It’s not as simple as just looking at, you know, a correlation analysis in Excel, you really need to take into account all of the other complexities that are associated with marketing. But that’s the core idea is to use statistics to find those relationships, as opposed to tracking individual humans across the internet. 

 

Kenny Soto  27:12  

When you bring this up, it brings up a question in my mind around associating credit to certain hods or individuals in a marketing team. My next question has to do with justifying both spend headcount and effort in the world of brand marketing, specifically, I find that this answer should be targeted towards those marketers, specifically, how should a brand marketer brand team approach this new world of attribution, and how it’s changing over time so that they can justify what it is that they’re doing? Essentially, building audience building top of funnel and making these bets for six months, a year from now, that actually driving revenue? 

 

Michael Kaminsky  28:02  

So this is a really good question. I, first of all, I’m skeptical of dividing a marketing team between brand and performance side. Let’s start there. Why? So I think that it leads to really unhealthy dynamics. I think the the way that this goes poorly, which I have seen as a is that the performance marketing team works really hard to drive a lot of revenue. And then the brand marketing team works on making like fun TV. And then like, that’s sort of it. And so the brand marketing team is making fun TV commercials, and then the performance marketing team is driving. And that’s like a terrible way to run a marketing team. Right? I think that brand is super important to a successful business. And great marketers should be thinking about brand all of the time, how can we continually reinforce our brand with our marketing messaging, and with our product, and with our customer experience, and marketing? On the marketing side? How can we use marketing to drive the business forward, there should be no marketing dollar that is not held accountable to some goal. And if the goal is conversions, then that’s great if the goal is driving awareness, and that’s great, too, but you should have a plan for how you’re going to measure that. I’m very skeptical of vanity metrics along the lines of like, just buying impressions. This idea of I’m a brand marketer, I’m going to make a really cool ad and then I’m going to, you know, spend as much money as I can to drive as many impressions. That seems like a recipe for disaster, going to be a waste of money doesn’t really make sense to me at any. I think that if you are a brand marketer and you say hey, look, you know, our unaided awareness is at x today, we’re going to run this campaign, we’re going to get it to x plus delta. And, you know, we’re gonna be able to show that with this. With this survey. And if we don’t achieve it, you know, our team is gonna get fired, I say great, like, that’s awesome, go out and achieve that it’s gonna be really hard. Good luck. I actually like would maybe recommend against that if you’re a marketer who wants to keep your job because it’s really, really hard to move in awareness number, but like, put your neck on the line, if you’re gonna go and do that, and CEOs and marketing leadership should be holding their teams accountable to these metrics, not things like, hey, our ad won an award in some industry magazine, such as the time you think, yeah, it’s such a wait time. So I think that really good teams, right? They’re unified marketing teams, they are all performance oriented. And then there are different people thinking about different stages of funnel, right at the top of the funnel, what are we doing? How are we measuring? Is that going to drive our business forward in some meaningful way?

 

Kenny Soto  30:52  

So Michael, you’re saying because I’ve actually never conceptualize this myself, instead of the classic way of doing an org chart where you have performance or growth and you have your product and you have your brand, then you have your PR and comms, which may sit somewhere else gave your content pod, you have your market research and insights, which is probably part of the data team and the straddle the two, etc, etc, etc. What you’re saying is that there’s another potential alternative, where the org chart is funnel based.

 

Michael Kaminsky  31:22  

That’s what I think should happen. And that’s like, necessarily, with the org chart being followed by, okay, the whole marketing team should be thinking about the whole funnel, and all the different, how the different pieces are playing together. Like I think I think about influencers as being a really interesting, where it shouldn’t be like, is it brand or is it performance, but really like where on the spectrum does that law and like influencers, I think can do top of funnel work in terms of building awareness and building credibility for your brand, pay this person who I follow who I have a lot of respect for really advocates for this brand, and seems to really trust it that helps to build up that part of the funnel, it can also try to drive performance, yes. And so maybe, right, I think really top to your marketing teams, they’re working with influencers trying to do but at the same time with the same dollar, they get paid back on the performance dollar, and it’s doing the other work of building brand for, there’s a lot of channels that can work this way, YouTube is another great channel that can help really help to build brand, at the same time that you’re driving performance. And so I think top two, your marketing teams are doing both, they aren’t just saying like, hey, there’s gonna be this brand spend that no one’s gonna hold it accountable for and then we’re going to have the, the performance marketing team like, you know, really driving hard with discounts and like, you know, not off brand activity, trying to just drive conversions wherever they can. Great marketing teams are trying to do both at the same time, and strategically thinking about, hey, how does our consumer actually make their buying decision? How can we help them make that decision, you know, in a smart, thoughtful way? And how can we move them along the funnel, using all of our marketing dollars all of the time. And I think really good brands are really good at doing that. As opposed to this idea of like, look, we’re going to split the teams. And you know what, the one hand isn’t going to know what the other hand is doing. That’s a recipe. Before

 

Kenny Soto  33:14  

asking the last question, I quickly want to get a little bit more context and on recast, can you provide more information on what recast does? Absolutely.

 

Michael Kaminsky  33:25  

So at recast, who we are building a next generation, marketing mix modeling, the idea is to do marketing mix modeling, excuse me the way that I described earlier, and do it in a way that works for modern digital marketers that are operating across many different marketing channels, which are very dynamic, because of the nature of automated bidding, and all of the things that we’re familiar with. And so we have built a platform with a bunch of tooling. And we tend to work with enterprise marketing teams that are trying to understand the essence of their marketing standard without relying on.

 

Kenny Soto  34:03  

How do you know if recast, recommendations are correct?

 

Michael Kaminsky  34:08  

Great question. Very hard to know. Okay, there’s a couple of ways that we think about it. So we think about validating model results with experimentation, right? So we should be able to say, hey, look, this channel is really performing for you. If you go and double spin into your, into this channel next month, here’s what we believe happen. And the corollary of that, of course, is making forecasts and then observing that they come through. And so we put a lot of emphasis on this idea of using forecast out of sample forecast accuracy on data that we haven’t seen before, to validate our causal inferences that we are making the model, right. If we are correctly picking up the causal relationships in the data, we should be able to make forecasts into the future that come true. And so we rely on that as a way of validating our

 

Kenny Soto  34:56  

Michael, my last question for you is hypothetical, because time machines don’t exist. But if one did and you can go back into the past about 10 years knowing everything you know, right now, how would you specifically accelerate the speed of your career?

 

Michael Kaminsky  35:10  

Oh, gosh, that’s a really hard question. How would I accelerate the speed of my career? I don’t know, I’d probably play a little bit less StarCraft two and spend more time getting good at programming and software engineering. Actually, that’s it. That’s the easy daughter, right? Less StarCraft two more software engineering, that would have accelerated my career by a couple of years and probably a really powerful way. So that’s what I would go with. But overall, I’ve had a great career. So it’s hard for me to complain too much

 

Kenny Soto  35:37  

about nice. Michael, if anyone wants to say hello to you online, where can they go to say hi?

 

Michael Kaminsky  35:42  

Yeah, so I’m on Twitter, at Mike underscore Kaminski. I’m also on LinkedIn, you can reach out to me there or just follow me I’m constantly talking about marketing measurement stuff. And then you can check out our our website at get recast.com. And drop us a note there if you want to chat about marketing. Thanks,

 

Kenny Soto  35:59  

Michael. And I’ll put all these links in the show notes for the listener if they want to connect with you. And as always, I hope everyone has a great week. And if you haven’t done so, please do these three simple favors that I’m going to ask one, subscribe to rate and three. If you haven’t done so already, please share this podcast with a co worker because it’s the best way for us to all learn together. And that’s it. That’s the show. Thanks for listening.

 

Michael Kaminsky  36:25  

Thanks so much.

 

Kenny Soto  36:28  

Thanks again for listening to Episode 139 of people digital marketing podcasts. For episode 140. We have a huge guest on the show. We will be talking about positioning and who better than April Dunford to be talking to us about such elusive topics. She’s the master. She is the expert. She is the leading thought leader on positioning and I was lucky enough to get some time with her and sit down and learn everything that I could learn within a roughly 40 minute time span on positioning from April Dunford. So if you like this episode of Michael Kaminsky, you’ll definitely like next week’s episode of April. If you haven’t done so please subscribe. And thanks again for listening. Peace.