Transcript
Michael True: There's definitely a race to make things more challenging within these like walled gardens of what's happening across all the different advertising channels. I think the race is just starting and eventually I think there could be an anonymous Internet.
Daniel Darling: Welcome to the 5 Year Frontier podcast, a preview of the future through the eyes of the innovators shaping our world. Through short inside pack discussions, I seek to bring you a glimpse what a key industry could look like five years out. I'm your host Daniel Darling, a venture capitalist at Focal where I spend my days with founders at the very start of their journey to transform an industry. The best have a distinct vision of what's to come, a guiding North Star they're building towards and that's what I'm here to share with you.
Today's episode is about the future of online advertising. In it we cover copilot for marketers, brand ecosystems, disrupting ad agencies and a privacy first Internet. Our guide will be Mike True, CEO of Prescient AI, a marketing technology company advancing the use of predictive AI models to maximize paid advertising efficiency and return on ad spend. Their technology works counter to how 700 billion doars in online advertising is done today by not relying on tracking users via cookies and instead building a platform to help marketers succeed. When a privacy first Internet arises and cookies go away with the most precise solution in the market, Preston can forecast future campaign performance across multiple channels. Three months out at a remarkable 90% level of accuracy. Calling over 100 iconic E-Commerce brands as their customers, Prescient has raised 20 million dollars from VC firms such as Headline, Blumberg, as well as the first check from US here at Focal.
Mike True has been helping brands take advantage of AI and analytics solutions throughout his career. Prior to starting Prescient, working at IBM, Oracle and Apani, Mike has become a great friend and a founder held in high regard as reshaping the online advertising industry.
Hey Mike, nice to see how you doing.
Michael True: Doing good man, good to see you. Thanks for having me on.
Daniel Darling: Let's dive in. We're spending over 700 billion do on online advertising to sell products, services, brands, etc. Can you frame up how the industry is structured today, how this money is being spent and how you think it's going to change over the future?
Michael True: Well I think traditionally when you think of larger media budgets at an enterprise level, even today there's still majority of the budgets are going into bottom of funnel Google search. Where we're starting to see the transition is their desire to go find new audiences right Scaling top of funnel their untraditional channels. TikTok, TikTok Shops, YouTube, Podcasts - where these diverse audiences are thinking about putting on a Joe Rogan podcast, right? Or doing these large media buys across connected TV devices. The diversification of media budgets is now starting to transition into more awareness. While obviously still knowing you do need to show up from a brand perspective at the bottom of the funnel.
Daniel Darling: And it sounds like with all of that you need some new competencies. Is the industry broken in a way or not set up to be able to fulfill that as it stands?
Michael True: I think about it from the notion of doing out of home billboards and direct mailers and linear tv, you know, has always been a really difficult thing to measure, right? Because you know, you don't have somebody that has a direct mailer that sees a coupon and they go into a retail store and they purchase something hard to really tie that back into it or somebody sees a campaign on a linear TV and then goes and purchases something online. There's always been kind of that disconnect. Traditionally when digital advertising started to become prominent with social media platforms, when they started to grow, data privacy was still kind in its infancy stages. And so the ability for cookies and fingerprinting and all these ways to track a unique individual customer journey was easier historically. But as data privacy started to become more relevant in front and center, even measuring these digital channels became more challenging which impacted the way you think of like using a multi touch attribution which was kind of looking for IP addresses. It's being able to track somebody with a cookie. And so I would say the ability to identify which even digital channels is starting to get in that same realm as out of home.
Daniel Darling: We all love cookies, right? Those fun little things that we have to click accept on every single website. It feels so archaic that we're still doing it and it must be inevitable that this is going away. So when will these cookies start to be phased out? What is the industry start to readjust itself as you say, to become more privacy centric.
Michael True: I reserve the right to change my mind on the dates. Chrome has been this moving target of every year. It's like hey, July 2023 cookies are gone and then 2024 and it's like we're not going to get rid of the cookies anymore. I don't know what's going toa happen with Chrome. I've always looked at this data privacy as a war for market share between Apple and Google, where Apple comes out with iOS 14.5 and their advertising revenue goes from hundreds of millions to billions of dollars because they're restricting other people to accurately measure what's going on across their devices.
And then when it comes to pixels, Apple announced this thing called, and I think everybody should keep an eye on this called Apple Private Relay and it's part of your iCloud subscription. Word is they might roll this up by default. But if you go into your iPhone, you turn an Apple Private Relay it literally blocks the IP address on your phone. Google comes out with a Google one VPN to battle them which is essentially blocking that pixel. So there's definitely a race to make things more challenging within these walled gardens of what's happening across all the different advertising channels. I think the race is just starting and eventually I think there could be an anonymous Internet essentially.
Then the question is, well, what are you stuck with to look at what inputs remain for you to start feeling confident in your measurement. And this is where more MMM statistical models, incrementality studies become more relevant because you just have spend and revenue and maybe some impression level data, collect data, cpm, so on.
Daniel Darling: Absolutely. And so you know, as this battle wages on and let's say the more privacy centric side becomes more prevalent like you say you need another model to not track necessary but target and identify how to use your ad spend across different channels successfully. Core to Prescient’s approach is around a probabilistic approach which was invented back in the 60s, but you've really innovated it, called Marketing Mix Modeling. Maybe you can highlight how that is starting to help advertisers in this new world?
Michael True: I think going back into 1962 CPG goods were being sold inside of box stores and these companies had advertising budgets but they were for billboards, radio, tv, catalog, newspapers, so on so forth. They still wanted to understand which channels were most effective so they could help plan out their media budgets for each year. And so they built these statistical models that tried to collect as much data as you could associate it to a billboard or households that are being sent to and whatever matching they could do, they'll see some of these larger MMs, they'll run a model once a year at the channel level.
And now you're starting to see companies like ours that are taking a much more dynamic approach where you can use a model that we are able to run every single day. So think about transitioning from an MMM that ran once a year at the channel level. We have a model that can run every single day at the campaign level and it's able to do cross-channel measurement. So in our example call it a retail industry, we can do your e-commerce store, we can do your marketplaces for example like an Amazon and then also over to your retail model. So think about it as more of like a dynamic or always on MMM where you're getting this cross channel media measurement but you're allowed to do these budget optimization and forecasting in a much more dynamic way. So you can be more fluid with the information that comes in to adjust these media budgets to making sure that you're always on track for your optimal profitability.
Daniel Darling: Fantastic. And what kind of data do you need to rely on to be able to accurately use these models to predict out what is happening across these channels?
Michael True: So there's data that you're going to get directly from the ad platforms. So think about your TV data, you're going to get your spend, all of your ad platforms are going to get some sort of spend because you use that for the calculation. You're going to get some sort of impressions and this data will never go away regardless of data privacy. I think there'd be a civil uprising if Meta ever took away how many people saw your ads, people clicked your ads, what was the CPMs and people seeing those ads. But it's predominantly spend impressions looking at some session level data and then what was the associated top line revenue and then what was the revenue that's been reported from their native ad platforms. That's the data you have access to.
But with MMMs, right, your spend doesn't drive 100% of your revenue. There's seasonality of the business, there's people talking about hey, I've been loving this new Kombucha. You talk to your friends, they might go to the online store and research this brand and purchase it. So there's a lot of statistical assumptions that you have to make alongside you're spend to being able to explain 100% of your revenue with those assumptions.
Daniel Darling: One of the things I saw that you're able to predict out future campaign performance by channel by up to 90% accuracy three months out. That seems like an incredible achievement. How are you able to gain trust in that assessment amongst your customers and sort of make those kind of claims?
Michael True: MMM, they typically do a lot of what we call back testing against historical data. So with the MMM, we ingest all of that historical data and we call it learning the brand's ecosystem. And how you learn that ecosystem is identifying their seasonality, which holidays impact them, trying is your best to quantify what is that impact of word of mouth. Once you have that kind of feeling good about what that truly organic revenue figure is or impact is, then you can start doing a lot of back testing against the actuals from the ad platform and the revenue of the brand. The more data you see, the more you learn, the more change that you see and spend, the more confident that we get in that relationship. And so there's a variety of different forms of back testing.
Daniel Darling: Where can this prediction start to evolve? How accurate will this get?
Michael True: I think it's going to translate into confidence. If you have the confidence of recommending media mixes (which we do today). We'll tell them exactly how much to increase spend per campaign, decrease spend per campaign, how confident we are in that predicted outcome, predicted growth over their existing spend. Why would you have somebody click a button to go do that? I think about it like a Bloomberg terminal. If you had a model that could automate media buying based off of a confident score and predicted LTV to CAC and you're spending $100 million on a media agency and they're taking 10% of that spend, essentially you're spending $10 million for somebody to press a button.
Daniel Darling: Each of these brands would have their automated media buys dictated by these models from there and then the agencies and the brand contribution will be on the creative side is kind of where you see it?
Michael True: I think that and like, I mean think about like you have the planning side of marketing and finance executives and then you have more of the activation side. Right? How do you figure out models that can help with planning and help with activation in a very automated confident back tested upper bounds of what could happen, lower bounds of what could happen to give a pulse on that? I firmly believe that's going to be the future.
Daniel Darling: Makes a lot of sense from that perspective. But then you also hear about how much innovation is happening on the creative side as well with sort of the actual individual ad, creative being customized and developed on the fly using generative AI. Do you not think that that is also in the realm of being automated?
Michael True: I think that's the other piece of the puzzle. We share a board with a company called Motion (shout out to Reza and the team over there). They’re doing some unbelievable things when it comes to creative analytics. MMMs should be what's used to power forecasting, planning and optimizations. That's what they've always been used for. Typically at the channel level, we go down to the campaign. Marketers still want to optimize at the creative level. Right. And so that's kind of that other piece to the pie. I see the future, and we're seeing this every single day, is I make this very, very clear to all of our clients. The MMM is not your source of truth. The MTA multi touch attribution is not a source of truth. Your incrementality test is not a source of truth. Platform, GA…the source of truth is the marketer.
What we talk a lot about is triangulation where incrementality tests are trying to validate the incremental performance of a channel at a certain point of time. Multi-touch attributions are trying to find a deterministic customer journey through IP addresses. MMM are trying to take a holistic view of all the things that can drive revenue. And how do you triangulate all those data points to feel as confident as you can to the decisions on how you spend your budgets are going to drive profitability. We see a lot of our clients, Cody from Jones Road Beauty (shout out to him). One of the most savant marketers of E-commerce - he uses a MTA, he uses an MMM, and he uses incrementality. What he'll do is he'll go turn on the channel, he'll find the incremental performance of that channel and then you start using some of our optimization models and come in, select which campaigns, press a button, we're going to tell you the mix and then spend it and then use the incrementality tool to validate did our recommendations work and then give confidence into using our MMM for further optimizations. And he's like, I did a 3-cell holdout on YouTube and it validated the results. And so now when he looks to our recommendations for YouTube, very hard channel to measure, he feels confident in his ability to go and scale that channel.
Daniel Darling: If you start to think about it that way, does the role of the brand marketer and almost the agency need to have a bigger injection of understanding the data science of how to use all these different types of tools and understand what the information is that they're seeing and how to act on it?
Michael True: Yeah, education like incrementality and MTA are pretty cut and dry of what happened. MMM is where the education needs to come in of asking the right questions to the vendor about how are you doing back testing and what models you're using, how often do they run. Because these are purely statistical models. And so I think the education is (1) feeling confident in using an MMM, especially for larger organizations. It’s so funny, we'll get on the phone with large corporations and they'll kind of be like, “how the hell are you running an MMM on a daily level at the campaign level, when they've just been getting a PDF report that runs once a year”?
And so how do you take this new dynamic MMM and then the education that comes is how do you use this along other forms of measurement? Yeah. I mean, what if you just had a creative studio that was automated based off of the automation I described? Right. So now it's maybe smaller agencies, more tactical engagements. Maybe it's, you know, the easy button for somebody to say, okay, this looks all right. Click the button right. Or this doesn't look right. Tell it why it doesn't look right. And it goes back and comes back with something say, does this look right? And then eventually it's gonna stop asking does this look right? It's going to know. And so I think of an automated agency that's purely focused on creative and identifying new channels and new partnerships.
Things like the E-Commerce industry, if I was to recommend the larger agencies, like look at people like John Snow from the Snow Agency and Jordan West. Like these people are all over TikTok shops and their agencies are growing really fast because that's where the next generation of people are. So it's getting more strategic into new offerings of ad platforms. A great example about this is AppLovin. If you guys check out what's going on in the projection of their stock price AppLovin was a 40 billion market cap that I think was doing like 4 million for mobile app downloads. And they come out and they create an ad network for D2C brands and agency brands to leverage all of their inventory. And that push their ads are within mobile games, that pushes them out to their Amazon store or their Shopify store. Brands are pumping on this channel now. And for us, we stepped up and agencies are stepping up. Like now we have the ability to measure Applovin in a very trusted way. Agencies are coming in to now offer this new channel. And if they didn't have to think about this, right? If they didn't have to think about the automation and the media buying, it's much more innovative on partnering with the ad networks and being more creative. This next generation is just wildly different than previous generations. And how do you talk to them which mediums do you talk to them? And that's why I feel like creative, creative is king right now. Having the right measurement stack alongside that is going to allow brands to scale to, you know, the goals that they want to hit.
Daniel Darling: Are there other parts of the ad tech industry maybe outside of MMMs that you know, you're paying attention to as potential big disruptions ahead?
Michael True: I'm curious to see what happens with data privacy and like the larger DSP first part. I'm curious to see what they're going to do with the reduction of first-party data based off of data privacy. That's why you've seen a lot of these publishers put paywalls because the amount of data that they're able to collect on audiences has drastically gone down. So you start to see them moving towards a subscription model. So can they push back the need for that subscription model through advancements in collecting data or acquiring other companies that have the ability to have access to other data sets.
And then there's the other side of the advertising industry where people are now trying to start to measure. Like people couldn't measure the impact of awareness through impressions. That's always been the thing, like we got this much reach but there's more things to look at related to number of retweets on a brand's Twitter, the type of views that they're getting on their Instagram, their stories. There's this whole other pure brand awareness that really isn't tied to much at the same time, that isn't tied too much data. You have people going viral, brands going viral. What are the impacts of these events that are taking place? There's some really sharp people that are going out that are just trying to solve and measure this brand problem alongside obviously the performance side. So that's something I'm really interested in watching as well.
Daniel Darling: And how about with the arrival of more search and online engagement happening with LLMs and AI agents, rather than going to a websites or going to Google, how is the industry trying to position itself to capture some of the user intent or the customer intent in the agent world, and start to market against that?
Michael True: If the data was made accessible to a unique ID of that person's email and then their search categories serve them up an ad within this agent, they click that ad, they go to the website, you're going to get that last click from the agent into the website. So perhaps there's some attribution that you could measure tied to people interacting with an agent. It's so crazy to think about actually, it's a fantastic question, by the way.
Daniel Darling: All right, Mike, well, hey, thanks so much for coming on and chatting to me. It's been a lot of fun and congrats on everything you're doing at Prescient and looking forward to watching it grow.
Michael True: Appreciate you having me on, thanks for the conversation.
Daniel Darling: What a timely conversation with Mike that really highlights the dynamic evolution of the advertising landscape as privacy regulations, emerging platforms and advanced technologies reshape the industry. Mike offers a compelling vision where automation, data science and probabilistic models like dynamic MMMs revolutionize how brands allocate media budgets and measure effectiveness. From TikTok shops to connected TV, the diversification of channels and creative innovations emerge as critical for engaging the next generation of consumers. Meanwhile, the shift towards more privacy centric models and the potential of AI agents raises intriguing questions about attribution and marketing an increasingly automated ecosystem. To follow Mike, head over to his account on X @Michael_True_ . I hope you enjoyed today's episode. Please subscribe to the podcast to listen to more coming down the pipe. Until next time. Thanks for listening. Have a great rest of your day.
