Digital tracking is broken: How to use MMM to measure incremental marketing performance

You mature marketing organizations recognize that there's no sort of one single measurement method that's going to solve all of their problems. And so we work with a lot of these organizations and we help them think about, okay, which marketing measurement method should you use at which different time. The four pillars of marketing measurement that we talk about are one, digital tracking, two, experimentation, three, surveys and then four, statistical modeling. We think about them all as being complementary, but sort of having different phenomenon in the world that they are trying to measure. So the first we talk about is digital tracking. So that's this idea of tracking people across the Internet, seeing what they click on, what ads they engage with before they purchase. Obviously it can be hugely impactful, but has its limitations. Right? It has limitations that we talked about, which is that it's not measuring incrementality, it's not an experimental result. It can be very biased towards bottom of funnel channels. But that being said, it gives very quick feedback directionally. It can point you in the right direction for sure. If things are trending up or trending, that can be a very important signal for you. But it's not the only signal. And especially once companies start to move into a world where they're omnichannel in terms of their marketing, where they're advertising online and offline, then of course digital tracking methods are gonna have real limitations because you're not gonna be comparing your online digital channels to your offline channels. In the same way, if you have a last touch attribution methodology and you're also advertising on radio and TV like there's no click there to measure. So you're gonna have problems. You're gonna have to come up with some way of bridging that gap. And often that ends up being very like back of the envelope smudgy in Excel file sort of logic. And no one really believes that that's going to be fully optimal. So very useful tool, lots of limitations. Surveys also very, very useful tool, give you a lot of insight into how your customers think about your product, potentially where they might have heard about you, but doesn't measure incrementality at all. Experimentation super important. If you can do a true randomized control trial, right, that's the best way to measure incrementality. But doing a true randomized control trial is very hard. Right? That's why pharma companies pay millions and millions and millions of dollars to try to run these trials correctly. They're very, very difficult to run. The last pillar is modeling. That's the pillar that recast, fits into. And the idea here is that we can actually use statistics to try to bridge the gap or the gaps left by the other pillars, the other measurement methodologies. So what we want to do in the modeling world is we want to use statistics and machine learning econometrics to look at historical data that we have and try to find the patterns in the data. So we want to look back historically and say, okay, were there certain time periods where we spent relatively a little bit more on radio, or where we spent relatively a little bit more on influencers, or where we spent relatively a little bit more on TikTok? And then we're going to look and say, okay, in those time periods where we spent relatively a little bit more on those certain marketing channels, did we actually see relatively more revenue while controlling for everything else that's going on in the business? And that's the core idea behind this statistical modeling bucket. And that's what we do at Recast. And the idea is that by doing that, we can actually get a read on incrementality for all of the different marketing channels because it's top down, we're not tracking people across the Internet for all of the different distribution channels because we don't need to put a pixel on a website. So we can do this for Target in store sales just as well as we can do it for Amazon and just as well as we can do it for E commerce. We think it's great, we think it's a really powerful tool. But we love when our customers use modeling as an input into their decision making process alongside these other tools. There's all kinds of complex things that we can do here that actually give marketers the full picture from all of these different measurement methodologies.