Attribution Time - Ep.34 - Michael Kaminsky

All right, welcome to attribution time, Michael, thanks for being on.

Thanks for having me. I'm excited to be here.

I am pumped about it. I was thinking about it earlier. I get excited about all the people that are on. I don't want to. Don't want anyone to hear and feel otherwise. But this is a really fun topic. I've learned a ton from you and what you're doing at Recast with media mix modeling. So it's a part of it is just the world I don't live in, which is what makes it so fun. I live in the multi touch attribution space. And I don't know how somehow you got into my feed the LinkedIn algorithm working properly. You're talking about media mix modeling and it resonated really well with where I stand as far as the value of attribution. And so, yeah, tell me a little bit. Give me the story of how you got into what you're doing at Recast with media mix modeling and how you came about solving and using media mix modeling.

Yeah, absolutely. So thanks again for having me. The story of Recast starts with work that I was doing when I was running the data science and analytics team at Harry's. Some of our listeners, if you're male, you've probably heard of this brand. You've surely been advertised to. They used to sell razors on the Internet. Now they sell lots of different brands basically everywhere. Back when I started, they sold razors on the Internet and my team there, we did all of the stuff. We did multi touch attributions, I'm very familiar with that. We built our internal solution, we did experimentation, we did ltv forecasting, all of that stuff. But as the business got more complex and they started selling at Target and Walmart and Amazon and started advertising more on out of home billboards and television and radio, we sort of realized that the multi touch attribution tool that my team had built and so I loved dearly, but it just wasn't telling us the whole story about what marketing was most effective for the holistic business. And so that led me down the path of learning about media mix modeling, which is what CPG companies have been doing for like 60 plus years in order to measure the effectiveness of their marketing spend. If you think back to like, you know, how Pepsi would measure their marketing spend 30 years ago, like, there's no clicks, there's no E commerce at all. And so what they did was they used statistics and econometrics to do it. And so I started doing A deep dive on media mix modeling and how it worked. I thought that with the development of a bunch of new statistical algorithms and massive computing power and like the cloud, you would be able to build a new version of media mix modeling that was a lot more effective than what had been possible in the past. And so that's what we're doing at Recast. We're trying to build the next generation of marketing science tools with this sophisticated modeling software that will help the next generation of omnichannel marketers to effectively and efficiently measure and then optimize their marketing budgets.

Yeah, yeah, that's such a good point. Like, media mix modeling has been around forever, right? Not forever, but it's been around for a while. But it seems to have picked up some momentum as of late. And you and I were talking, you know, you like to think that you added to that with what you're posting on LinkedIn. I can't disagree with you because they got into my feed. But yeah, I think it's a really interesting concept of the balance of multi touch attribution and media mix modeling. And as you mentioned, it's using the statistical analysis to be able to project where to invest your dollars, where to go, which is different than the multi touch attribution space, which is, okay, we invested XYZ dollars into channels or campaigns. What's the outcome of that? And then that way you can know, okay, coming up next year, we can invest more dollars here and less dollars there. And they kind of work in different spaces in that sense, I guess. What's your approach or what's your experience with the, with the multi touch space as you worked at Harry's or other places?

Yeah. So I think multi touch attribution and other types of like digital tracking methodologies will always have played an important part in the digital marketers toolkit. Right. It makes a lot of sense. You want to know who clicked on which ads prior to purchase. With the app tracking transparency, which has given reduced some of the ability to track people across sessions and devices and then the removal of cookies from the Chrome browser, it's getting harder to do all of some of the fancy stuff that some of the multi touch attribution tools were able to do in the past past. But I still do think that it, it will continue to play a role, especially first touch, last touch attribution are they continue to be very valuable reports to look at if you do any sort of E commerce at all. And I think they will continue to be very valuable reports to look at far into the future. So I think they're, they're important tool sets. With any type of tool that you're using, especially in like the advanced analytics space, you just really need to be aware of what the limitations are. And I think this is the place where marketers actually got into trouble over the last 10 or 15 years is they got these really nice reports that were very clear about, you know, who clicked on an ad prior to purchase. But they, some people failed to think critically about exactly what that meant and where it could go wrong and what the core limitations of that report are. And that's a place where I think, you know, the next, this sort of, over the next couple of years, marketers, I mean marketers already have started to like, start to ask themselves the hard questions, which is one of the reasons why MMM has grown in popularity as a topic. But it's going to continue to happen where people need to start like looking at these reports and not just thinking about, oh great, you know, we have a CPA of $40 on a last click basis and start to think about, well, how could that be wrong, right? And what are the other ways that we can corroborate that or how should we think about that in the context of these other measurement methods? And that is what I think the next generation of marketers are going to have to do over the next couple of years.

Yeah, absolutely. And to take a step back, one of the key distinguishers with MTA and MMM is MMM is that high level overview of hey, we spent this money and how. And then it helps you understand where to continue to invest your dollars and be able to forecast based upon a variety of different things depending on whatever you put in your system. But you know, you know the most common would be like a pandemic ahead. So like you can, you can actually track and add these things into it. So if it gets a part of that computation. Multi touch attribution is individualized. So Even in a B2B perspective, it's still one to one as far as those engagements. So as what Michael's saying here with, and I think this is, again I agree with the reason why multi or media mix modeling is increasing is as the deprecation of cookies happens and people have more power around, yes I want to be tracked or no I don't want to be tracked. Media mixed modeling still enables you to be able to report on where to invest your dollars. And as far as an ad level and conversion rates where the multi touch attribution may lose some of that insight. And though you know, people are engaging on specific web pages because they might fill out a form or whatnot. You may lose some of the additional insight of ad spend and things of that sort. So again that's where you start seeing these combinations take place. But to your point Michael, like 100% people are becoming much more intuitive, a lot more asking a lot more challenging questions. Where back in the day it was just you had Google, Google told you that their ads accumulated to revenue and that was, you had no reason to ask otherwise because you could easily see one to one and you could just pump money towards it. And we're getting to a point where people are having to be a lot more strict with how they're spending their money. You know, there's a lot of other factors involved in that of why that's the case, but money's not as easy to come by and some of these departments and so it's really helpful to have a media mix tool. I also think so.

I think that's 100% right. And I think the other factor is, well there's sort of like a, an ecosystem maturity and then as well as like an individual company maturity. And I like my belief is that, you know, if you're a startup or a small company and you're only advertising on one or two digital channels, like you do not really need to be thinking about a media mix model. It's not going to add very much to you like first touch, last touch report. If you're only advertising on Facebook and Google, that's really all you need. That's going to get you really far. That's awesome. The place where it starts to get more complicated is when you have these like much more complex full funnel marketing campaigns where you're doing billboards, you're doing TV, you're doing radio, streaming TV, YouTube and then all of the social channels. That's the place where there's a lot of reason to believe that you're not going to see all the clicks. You don't get a click from a billboard, you don't get a click from a linear TV channel, even a lot of streaming TV and YouTube. You're not going to observe clicks at like there's a good chance that a lot of those ads are having a lot more of an impact than what you're going to observe via clicks or even impression tracking. And so that's the place where you know, as brands are starting to get more, get grow in maturity and get more Omni channel and sort of the ability to just like we're only going to advertise on Facebook and Google for years. Like that's less, it's less possible now, at least as far as, as far as growing brands go. And so that has led people to start to ask this question of like, okay, if I want to get advertising on TV or digital video, what is the right method of measuring that? And it's not like mta, I think, has more issues with those types of channels just because of the nature of inherently being able to, to observe clicks or necessarily observe impressions and link them back to different individuals.

Yeah, absolutely. Yeah. I'm curious about the TV world. I mean, the QR code, speaking of things that have been around forever. The QR code has been around forever. And when, when the pandemic hit, it became like, now everyone can use a QR code, but it's been, you know, it's one of those things and you're starting to see that even on TV ads where people are able to scan that. And now that information is tracked. Right? You're like, you, because you're watching an ad on tv, you scan that. Now you can have that inside the system. Like it's.

Yeah, but how often, how many times have you scanned one of those codes?

That's fair, that's fair. I know, that's why I'm like, does this scale right as things move out elsewhere? But I'm right there with you. Way to call me out on that, Micha. I can't, I can't. And I can't take it out of the podcast because that's too real. Yeah, I think that's an interesting point though, with the maturity of data. So how would you, what's your approach with that maturity of data as far as here's all these different sources as you mentioned. Hey, if you're doing one or two place like sources of data, medium mixed modeling is not your case. You could easily do basic forms of multi touch attribution to get what you need. But as you get to those more complex paths of all these different data sources and I mean I've seen it where people, not only is there company silos, marketing and sales, but that then leads to data silos of those different places of those different units within the work. Like and then within marketing you have those different data silos because Legion's responsible for all their Google and paid ads. And then you know, the demand gen teams work in these other streams and so they're, they're monitoring those. So how do you, how do you manage that?

I mean, look, it's hard. This is a place where I and this is like a, you know, almost a do as I say, not as I do situation. Like Recast is a startup and we do like we do advertising and we do marketing and our data warehouse is not set up. It is not good. So like we're still very early in the journey but for, you know, we work with some like very big B2B companies that you know, are very mature and I think the best in class companies invest in data quality early. They see it as being a core part of how they do business and how they are able to drive efficiency in their go to market teams. And so they've invested in data warehousing and you know, they have a central location that has all of their relevant marketing and go to market activity. And that could be sales contacts, it could be, you know, the marketing spend on the different marketing channels that they're active in, which conferences they went to. Right. All of that information is stored in one location. And the great thing about that is like one, it makes like answering lots of basic questions about what's happening a lot easier because you can, you know, they have a BI tool, you can build these nice reports and dashboards. And then beyond that it also makes doing more sophisticated analysis a lot more possible because all of a sudden, okay, we already have all of the data ready to go if we want to analyze, you know, what was the lift in sales qualified leads associated with going to this conference? Because they know the date of the conference, they have the sales qualified leads tracked, they also have all of their other marketing activities. So you can, you know, build a statistical model that controls for it. Right. There's a lot of stuff that that opens up, including fairly straightforward analyses, but going all the way to doing a marketing mix model. Like it's very easy for Recast to work with them because we can just plug into the data that they already have set up. So I think if you're one of these companies, right, you want to figure out like, you know, what is the right time to do that investment. But I think if you are on the path to becoming, you know, a really large, you know, multiple hundred person company, like it's, you probably want to think about, hey, it is going to be a good investment to get all of this marketing data into one central location so that we can start to do stuff with it.

That's right. Yeah, exactly. I mean there's a rise in my mind with the snowflakes of the world and just whoever else is managing those. I mean They've been around forever. But with these larger organizations, and I say this often, but especially with the rise of machine learning and AI, like, if those data points are disparate, then it's hard for anything to really learn off of it. So same as media mixed modeling. To be able to take this and run against that, it's algorithm. It's the same for AI, it's the same for machine learning. Like, you just can't get the insights you need. As much as the teams don't want to work together, that's just essential.

I think that's 100% right. But also, like, if you're a leader at one of these companies, you need to get these teams to want to work together. Like, that's the other choice is like, you know, hopefully if these teams are like antagonistic, like, you have maybe a different, bigger problem you need to go solve first. But hopefully you have a world where, you know, sales sees marketing and knows that, like, you know, they need marketing's help to get the leads and marketing sees sales and it's like, okay, we need you to help close the deals and everyone's incentivized to work together.

Yeah, I mean, that's. That's one of the things I've really enjoyed about attribution, especially as I started in the space, is I was talking to marketing teams who purchased it, and then I was saying, hey, we have to pull someone in from sales to have XYZ conversations. Right. To understand how are they logging data in their salesforce and being able to start and being a kind of the mediator between that and not breaking the.

Fundamental assumptions of your modeling methodology by including things that are endogenous to the model or that are actually dependent on marketing activity that's happening further up the funnel. Hopefully that made sense. We got a little bit into like the technical statistics, but hopefully that was clear.

I like it. I think I got maybe 50% of that. And that's okay for those listeners. Hopefully they're able to get a little bit deeper into it. There's a lot more intelligent people out there than myself, you being one of them. So I think. I love when people really get to start nerding out about what they're passionate about because it's fun to listen to and obviously I get to learn from that. Let me ask you, let's take it the next stage. So you have those different pieces and I think this applies a little bit. But what if. Or can you. And do you see this where people use mmm for obviously the initial Sell of the deal. But what about like different sales cycles? Like so now we have renewals, things of that sort. Are you running those models in those circumstances or do you see it primarily in that early new. New customer engagement life cycles?

Yeah, so we primarily see it in new customer oriented KPIs. So we, so we think about like the model will have some outcome variable which is the KPI in general. The reason why that is because marketing teams tend to be the ones who are most focused on using an mmm. And marketing's job in general in most businesses is to drive new business. And you know you might have like a few campaigns that are renewal oriented but that's going to be some very small fraction of your total budget. And I think a lot of what drives renewal is not the marketing. Right. It's product things, it's the customer success, it's all of those other factors. And so like you could maybe run a model but I just don't think it's going to give you that many useful insights. So the vast majority of people are using it for hey, what's driving sales qualified leads for us and mostly being focused on the paid media that is driving that activity.

Yeah, no, that's good insight because something I've played around with my mind I was having this one, I think it was the second session I had it was with Drew Noel who's really intelligent that go to market space but he kind of like made me think differently about multi touch attribution. We don't want to always measure everything, right, That's a caution. But one of the things is there's a patternization across every motion of, of sales. Like whether it's the initial sale, there's a renewal. And to your point Michael, like you just said like it's not as much marketing but like even those data points like I was having conversation with who's in like partner marketing and partnerships and how, how can we actually understand and begin to monetize these different actions? So yeah, like here there's teams that are dedicated to these accounts. Can we measure these actions in some capacity to be able to run against this and say hey this is what we found to be, this is the secret sauce. And that you know that gets to the point where a lot of people say don't over attribute, don't do this and that but I mean we're at a place with the amount of data in which we can collect, the amount of data in which we can run these things, the speed in which AI runs at nowadays it just seems like why not? Why not try, try something like that?

I mean, this is an interesting question. This is, this is honestly something that I struggle with as a business leader who is trying to think about like, we're spending money, we have a bunch of different marketing activities that we do. We spend money on marketing. But I actually, like, I'm on the side of like, don't over attribute and don't try to overthink it because I think the problem is that there are going to be holes and there's going to be biases in these measurement methods. And so if you put too much focus on trying to track every click and make every connection, then you can often lose the forest for the trees. Right. And you're focused on the trees, which is like, what did people click on and what were we able to successfully track? And you lose the forest where you say things like, well, you know, we've been doing a bunch of this activity and it really feels like it's working. You know, we get on sales calls and people mention it. We don't have a lot of clicks to link for a variety of different reasons because you just can't. But it does feel like it's working. Like our volume is definitely up since we've put more effort into this and people are talking about it and we're getting more cold inbound than we ever have before. Like, I think as a business leader, I myself want to be attuned to that and sensitive to that. And I don't want to like get distracted by the channels that are easiest to track because they're good. Like LinkedIn is really good at tracking clicks. Like they're good at it because they've been doing it forever. But like there's other activities that we do that I think, you know, has more of an impact on the business than paid LinkedIn ads. And so I don't want to get distracted by what's the easiest things to attribute as opposed to what do I think strategically is really helping our business. And then even just like intuitively feels like it's moving us in the right direction.

Yeah, that's a good word. I mean that's, I think that's, I think that's wise. Right, because you can really do them. You got people solely dedicated on how to make sure all this data works because it takes time to manage data just in general. Like there's people out there, that's, that's what they do. They're managing a data warehouse. It's nothing else other than that. And so there's A lift in that. And you have to even consider those ROIs of. Is that worth our time to, like, pay for this person's time to just be doing that? Right. Like, there's the internal ROI involved in it as well. So I think. I think that's a good word in that. Um, so tell me a little bit more about what recast is doing. What. Where y' all are headed. What's. What's the 2024 looking like? It's a good.

It's a good question.

I mean, we're.

We're feeling great. Recast is growing really fast, which is fun. As you mentioned, a lot of people are talking about mmm, and so it's a good time to be in the measurement business.

We.

Our vision, right, is to build the marketing science and planning platform for the next generation of marketers. And so it definitely includes mmm, but we're starting to add on a bunch of these tools on, like, mmm. I think a lot of people think of that as like, okay, yeah, like, you'll get a report that says what channels are performing, which is interesting, but not that actionable. We sort of think a lot about, well, how do you actually move along the actionability spectrum to get to something that's, like, really useful for actual marketers? And that's where we think a lot about how do we do budget optimization. Right. And automatic automated recommendations for how you can reallocate your marketing doll, doing things like forecasting and scenario analysis of, like, what would happen if X or what would happen if Y. And how do you build a tool that allows marketing teams to explore those what if scenarios? And then I think, you know, there's a lot of room to continue expanding that platform to. To encompass more parts of marketing science and helping. Just really helping marketing teams make better decisions based on the data that they have. And so that's the vision for where we're headed.

Nice. I love it. Well, Michael, thanks for being on attribution time. This is super insightful. I know others that are listening to are going to find a lot of insight from it, so I appreciate your time today.

No, thanks for having me. This was a fun conversation. We could do three more hours of this.

Sure. Yeah, that's right.