00:37:44 · Henry Innis
[00:37:44] S0: Definitely like psychological theories around that that that advertising get more effective over a long period of time if repeated. You're a 100% correct. But to say there is no short term payback is also incorrect. If it can't sell today, chances are it can't sell tomorrow. Right? Like, and there's a lot of research around this. So like, Byron Sharp, who wrote a book called How Brands Grow, Ehrenberg-Bass Institute, one of the best institutes in the world in like marketing research for long term brand... Branding. It said an ad that pays off in is... In a short term is more likely to pay off in the long term versus an ad that doesn't. Right? So they actually did all the research across like 50 brands and they found this. Jenny Romaniuk, who's one of the all time market... Like researchers on brands, thinks the concept of like pure play brand campaigns that don't mention products is one of the most dangerous things that have ever been created. And so I think that's a huge problem in the sense that we built this expectation up that brand marketing somehow should not deliver a sales result and performance marketing is the only thing that can deliver a sales result. Both are not true. One is really about mental availability. So, you know, building kind of, you know, presence in someone's mind. And what bottom of funnel is about is physical availability. Like how do we actually distribute the products effectively? And so because of that, you measure... When you're measuring and modeling by being trying to attribute to one channel, it's actually a complete fallacy because actually it's a combination of channels working to influence the overall sales result.
00:39:11 · Sachin, Henry Innis, Adam
[00:39:11] S1: Maybe an example of that is like if someone listens to one that adds on our podcast, but then goes and Googles the company afterwards, that's attributed maybe to the SEO that their company [00:39:20] S0: is That's right. So like, you know, think of it this way. Say someone's buying like a, you know, a new tech license and they're kind of analyzing that from their perspective of, okay, we're gonna we're gonna like walk down straight to see a billboard, see a podcast, they see like all of these different things. Yeah. And then eventually they come and like, like search. Like search is capturing demand. Right? And that's what a lot of performance marketing is. It's demand capture. So I think that's a huge issue. Like definitionally, we put ourselves into two camps, which is creating a really toxic environment versus going, hey, we use these combination of tactics. And if we use these combination of tactics, this is how we get the optimum result. [00:40:01] S2: Right? But how... Like, can you actually attribute if something is coming from multiple factors? [00:40:06] S0: 100%. [00:40:06] S2: If you've that podcast, SEO,
00:40:06 · Adam, Henry Innis
[00:40:06] S2: If you've that podcast, SEO, [00:40:08] S0: all that. Because you don't track by the person, track by the aggregate. Right? So if you think about stats, I'll give you another example. Hedge funds. Right? Hedge funds look at a combination of economic and market factors to predict the future value of a stock. They get it right a lot of the time. Yeah? Same thing exists in marketing. All we need to do is look at the combination of factors including advertising spend, pricing, all of these different factors, model it through really complex algorithm. And what happens? Hey, presto. Much like Bloomberg, we can start to model the value of stocks. Right? We start to model the value of actually, we do these combination of things. This is the likely sales result that this business gets. [00:40:43] S2: So you're trying to isolate like every single variable and see its like influence on the output.
00:40:47 · Henry Innis
[00:40:47] S0: Yeah. You're trying to see if that variable goes up or down. What's its influence? Right? So if I look at it, if I look at everything, you know, just at an individual personal level, of course it's impossible to track. But if I look at, hey, this business does a combination of activities, when it's doing more or less of this, what was likely to happen in that following week? Suddenly it becomes really easy to assess your tactical mix. And that's how Mutinex fundamentally works because we've got these amazing AIs that are able to just do that at scale. Normally, would get like 50 analysts to do that. You're sitting there around a room and you'd be kind of running around and going, okay, like everyone's built this complex model. We've got all these correlations, stuff like that. But like, it's a perfect built problem for AI. It's perfect built. Like it's literally the best problem an AI could do because it takes a complex data problem, like, you know, responding to someone's questions, for example, chat GPT, you know, responding to a heap of time series data in the Mutinex world, and it simplifies it into something that people can use and consume really, really quickly. And, you know, if you think about AI, what's it really force multiplier in? If you're looking at why models work, it's because they take reams of complex data, reams and reams of it. Yeah? And they make it consumable to humans. We're all looking at AI as these big complex things. What they do is they execute complex tasks and things like that with simple instruction from humans, or they allow humans to consume information in far more simple ways.
00:42:07 · Sachin, Henry Innis, Adam
[00:42:07] S1: How much kind of marketing spend do you need as a baseline to these models to start to work? Like, how much data do you need to train? [00:42:13] S0: It's more of it like... Well, so I'll give that answer in two parts. One, for us, it's probably you need to pay... Have a marketing budget of more than 5,000,000,000 for, like, off these to work. Right? But I think in in general, it's actually variation of spend and variation of sales that counts a lot more. You can have a really small marketing budget of like $500,000 and you could still build them with pretty good market mix model. [00:42:33] S1: If you're turning things on and off. [00:42:35] S0: Correct. And most media plans, they have natural variation, so it kind of works. [00:42:39] S2: Yeah. Yeah. Interesting. So basically what you're doing is that you're integrating every part of like someone's marketing data, all the different platforms, and then you're finding all those variables and you're solving those calculations. [00:42:50] S0: Yeah. Totally. It's like just seem like a Bloomberg terminal. Right? Bloomberg terminals make it... Make capital allocation easy. I wanna build the Bloomberg terminal for marketing. [00:42:57] S2: Yeah. Why wasn't this done far before? Like, why wasn't this sort of data centric approach like the I... Well, it was first of all,
00:42:57 · Adam, Henry Innis, Sachin
[00:42:57] S2: Yeah. Why wasn't this done far before? Like, why wasn't this sort of data centric approach like the I... Well, it was first of all, [00:43:05] S0: so there's like heaps of market mix modelling vendors. The thing is they're consultancies. And when you have your entire P and L for 500 people was built around head hours, pretty hard to change a business. Right? Yeah. So I just think that that's a difficulty for a lot of businesses. I think the second kind of component is it just was expensive, like pre cloud, it was expensive to do this. Pre like proper large modeling techniques. It was hard to do this as well. Like a lot of the recent pushes in terms of, you know, how you can unlock large model libraries, how you can model quickly in the cloud, you know, how Vertex runs, for example, like that stuff all counts right now [00:43:40] S1: Yeah. [00:43:40] S0: For actually making it way easier to execute this sort of business. And and to go quite practical, what are [00:43:45] S2: some you think are underrated and overrated marketing channels right now? [00:43:49] S0: Oh, I'm sure you get myself in trouble. Yeah. Let's do it. [00:43:52] S2: Like, I'm sure you must have a view of all these different channels and nuances depending on what sort of business you are, but I'd love to hear some takes.