How AI Turned a 3x ROAS Industry Into a 12.47x Growth Machine

Welcome to technology. Half the money I spend on advertising is wasted. The trouble is I don't know which half. That quote from John Wanamaker has defined marketing for over a century.

But what if that uncertainty is finally disappearing? I'm your host, Ken Yamasky, and today we're diving into how data, AI, and advanced modeling are transforming marketing from guesswork into precision and from unpredictable outcomes into scalable, repeatable performance. My guest today is Sheldon Poon, co-founder and CEO of Drive Marketing, a company that's redefining how organizations use data to drive measurable growth. Sheldon has spent over two decades at the forefront of emerging technology.

From early work in AI using frameworks like TensorFlow to building high-performance marketing engines, he's helped transform companies from millions to tens of millions in revenue, while consistently outperforming industry benchmarks. At Drive Marketing, Sheldon and his team are focusing on one thing, turning complexity into clarity and data into results. Sheldon, welcome to Technology. I'm glad you joined us today, and I'm looking forward to hearing how you've made Drive so successful.

Yeah, thanks for having me on the show. Appreciate it. I'm really happy you're here. So let's start with your journey.

What led you to build Drive, and how did your early work in AI shape your approach to modern marketing? um so uh the company came about pretty organically actually um my my co-founder and i had worked together as freelancers and um as we continued to take on more clients we're like hey let's let's build a name for ourselves and let's put it out there and make it clear that you know we're partnering together on these things um they saw my my uh my former business partner was a very talented or still is a very talented uh graphic designer i brought in a lot of the technical knowledge and in the early days we were doing mostly web development web design um over time as the company grew and you know we became we moved from being kind of like on the ground doing the fun work to moving into like more of a management role and running an actual company um they saw decide to kind of move on and kind of left me at the reins um so when he left um i didn't feel that I was on the creative and design stuff, I wasn't strong enough.

So we kind of pivoted away from more of the content side and just really doubled down on the technology side, which is where most of my strengths are. So fast forward today, and we've evolved from going from building websites to getting into more of the marketing side, and then ultimately ending up where we are now, which is very heavily focused on ads and ads performance. And I think the thing that really makes us stand out is that in 2026, with all the tech that's layered on, like it or not, advertising has become, I would say, it's skewed more towards a technical and engineering exercise today that's based on data, that's based on algorithms, more than it is based on just good creatives, and it's less of a design exercise.

So I think we're well positioned in that sense, because we can go a lot deeper into the algorithms and understand a lot of the tech on a level that a lot of our competition just doesn't. And that understanding on the engineering side gives us a bit of an edge. Yeah, it's really not the same thing as in Mad Men, is it? No, not like you and I talked about that back in your days of the agency life and back when my father was in advertising, things have changed quite a lot.

Back then, it was, you know, the big splashy ad that grabbed a lot of attention. And that was the thing that really moved the needle. And that's still the case, I think, for a lot of the big, big, you know, international brands, national brands. But for more of the day-to-day, like, on-the-ground guys, like, the reality is that, like, in today's market with all of the noise that's out there, you have to be advertising.

and the day-to-day guy, the thing that's given to give them the edge is making sure that they're not wasting money on inefficiencies. They're not just throwing money into the meta and Google kind of black hole and not knowing which half is making money and which half isn't. That just doesn't fly anymore. The competitive edge is making sure that your tech stack is kind of taken care of.

So yeah, quite a lot has changed since the Mad Men days, unfortunately, for better or for worse, I guess. I miss those days. Well, my take on it is marketing is all about performance now, and data is what's driving that performance. I know you work with the well-known clothing brand Bench.

How do you ensure that the data you're relying on, especially in scaling brands like Bench has been totally rock solid because I think if you don't have the right data, it's garbage in, garbage out. So talk a little bit about that. Exactly So one of the things that I think everybody getting distracted by today is all of this talk and hype about AI But the thing about these models is that they are only as good as the data that they trained on The models themselves are not the problems I see people you know when things don go well oh they blaming this model that model But it not about that It the quality of the data So whenever we're onboarding a new client, right, we're looking at not just the accuracy of the data, which of course is important, but the completeness of the data.

A lot of the times when we're first onboarding clients, one of the first things that we see is that they've kind of installed the generic default, yeah, okay, we're capturing data, and then they left to that. So they're missing a lot of micro actions that are very, very important when you're trying to model an entire funnel and you're trying to understand top of funnel all the way to the bottom of funnel where are people dropping off what are the micro actions that they're taking and that's not something that is generic that's just out of a box you have to do a little bit of bespoke adjustment you have to think about what the client messaging is you have to understand for that brand what is the client journey and what's relevant and what's not and then adjust accordingly to make sure that the data is complete.

So that first and foremost, that's, that's the thing that I think a lot of people miss is that they think that the, the default, you know, GA4 tag, the default Facebook pixel tag is going to capture everything. It won't, you have to do a little bit more work. So the completeness is the first piece. And then in terms of accuracy, we, when we take on a client, we really do become partner with them.

We do ask them to open up their books so that we can compare the revenue that the algorithms think they're capturing versus the revenue that the accounting department is actually looking at. And that critical step makes it so that right from the get-go, we can figure out, is the tracking accurate to what the accounting is seeing? And if it's not, why not? And that's another critical step that we've seen so many companies skip, and then they don't raise the red flag until accounting goes like, hey, the numbers you're reporting or not good.

And by then it's too late. You've already lost the client. So the completeness and the accuracy of the data is the first thing that we look at. And just by my description, you can tell that a lot of that is not automatic.

We have to sit down and have real conversations with the stakeholders. We have to talk to the client. We have to talk to the people who are on the ground. They know, our customers know what their customer journey looks like.

They've been doing it for 20 years. They are the best source of knowledge for us to make sure that our modeling is correct. They know what their internal numbers are. There's no way for us to magically know that.

So we really have to work closely on that human level with our partners, make sure that the data is accurate. And then the easy part is just feeding that into the algorithms and letting them run and do their thing. So let me ask you a question. AI and predictive modeling are changing the games across business, not just in advertising.

But what role are they playing in boosting the performance and driving consistent returns that you're looking for, your clients are looking for, and are really changing the game, as you said, in the new world of marketing? so i think that's another thing that a lot of people uh again this is kind of anecdotal but when we're onboarding new clients and we're talking to um the previous agency or another agency one of the things that people seem to miss the mark on is how to use the ai and how to trust the ai so when uh early early on when uh google and meta were rolling out their predictive ai models they were not very good.

They made a lot of mistakes. Our take on it is that the initial adoption was not very high. So people would try it out and then say, this isn't working, and then go back to the old school way of like, I'm going to set my demographic targeting. I'm going to try to narrow in on locations and I'm going to create the customer profile.

And people worked against the AI. Fast forward to 2026 and targeting people by demographic does not work as effectively as carefully letting the AI do its thing. So the way in which you properly leverage these predictive models now is that, so going back to like old school marketing, right? Attention, interest, desire, action.

Old school Ida. What we do is we make sure that we build a library of content that represents each of those layers properly. And again, that messaging comes from working with our client, leveraging their decades of knowledge in their space, knowing their audience, knowing their brand, knowing their messaging. We take all that and we build out that content so that each layer is well represented.

We take all of that content and we're able to throw it into a library on Google and on Meta. And what those platforms do is that they will use the data that they have on each person. And so Google has, I forget like how many hundreds of data points on each of us. And what they will do is internal testing to see which content resonates with what customer pattern And they just let the math do its thing And even though we're not telling it what piece of content represents what layer, what will emerge from that predictive algorithm is that Google will be able to create an audience that is at each level of the funnel.

And when we compare the content and what each level is connected to, lo and behold it matches up with what we thought was going to work with messaging so that's how we leverage the predictive algorithm without getting it its way what we're doing is we're giving it all its tools we're giving it the data we're giving it the content we're giving it access to to everything that it needs and then we're letting the predictive algorithm actually do its thing and then figure out okay this is the audience and based on what i've learned now i can predict that that I can go after another 100,000 that match this pattern.

That's how you leverage it properly. And we're not seeing a lot of agencies do that. We're seeing a lot of people working against this. Yeah, that's counterintuitive to, you know, the advertising industry as a whole, you know, even 10 years ago.

It's really interesting. So let me ask you another question about, you know, I'm sure you do a lot of planning and have a lot of strategies. And, you know, you're very, very deliberate in what you do to get to market. but uh walk us through a moment where campaign or technology failed you um and either mid-execution mid-campaign whatever uh it failed and then how did you deal with that how did you pivot how did you uh how did you deal with uh you know not not having the whole thing fail on you but that one element.

Yeah, absolutely. So going back to our client, right? They're very, very strong in the Canadian market. They're found in Costco, winners, like all these big retail brands.

They want to make a move into the US market. And we were totally gung-ho on that. We said like, yes, we can totally do this in the US. We have US clients.

We've done it before. um the thing that worked against us was our recommendation was to keep the data clean um make sure that you have a separate website for the u.s market because that's what we've done with all of our brands that we that we've uh kind of gone into the u.s with canadian site u.

s site completely separate nice and clean um they were running on there or they still are they're running on Shopify and Shopify at the time had just released a new feature which allowed you to separate out markets. There were some, this is also, this is one of those situations where like it's above our pay grade, so to speak. The higher ups did not want to spend the money on building up a separate website for the US market because of the cost incurred, because of the way their ERP was set up, because of a lot of the way that they're on the back end of the business, right?

There's logistics. It's understandable. There's a cost incurred. It's not just simple, hey, I'm going to put up another website.

You have to do a lot of work to make sure that all of the customer service shipping, all the other stuff that goes into it. So when they did the cost analysis, they're like, no, no, we're going to put everything on one site. At the time, Shopify had just released their multiple markets option in the back end, and it was still very experimental. And we said, don't do that, but they want to try it out anyways.

lo and behold when we tried to set up the shopify site to handle two different markets canadian us we we had to separate out the data because obviously the canadian data you don't want mixed in with the us data simply because um there are two different markets there's different shipping to be considered there's different warranty issues in some cases depending on availability, there's different products. So we insisted on separating out the data. When we tried to do that, it didn't work properly.

And what happened was we started seeing massive gaps on both sides and it risked basically taking down both sides of the business, the US and Canadian side. So it really kind of showed how important the cleanliness and the clarity of data was to the performance of the brand. And what ended up happening is we saw that as the US side kind of increased in sales, it was somehow cannibalizing the performance on the Canadian side. So they weren't making, they weren't doubling their market.

They were just kind of like, you know, taking away from one and putting in the other bucket. So it didn't make any sense. Unfortunately, ultimately, we had to stop the US spending, fix the Canadian side, which took a few weeks because the data was so corrupted at that point went to restart some of the algorithms we took a big hit there and we shelved the u.s marketing for the time being um i i believe now they're reinvestigating it to see like with all the recent updates with shopify this is something that we can try again now that the company's a little bit more stable now that the the the the website's a little bit more stable so we're gonna see if that's been fixed but definitely that was a huge kind of more technical hiccup, but it definitely had a huge impact in terms of breaking things.

Just not having that data structure set up properly was just a complete disaster Okay Let me turn the focus a little bit more towards perception In a lot of categories including fashion e-commerce, there's a perception that success is partly because of luck. When you look at work with brands like Bench, who you've been very successful with, how much of that was strategy versus timing so um i think when it comes to marketing it's kind of a race right and uh while yeah some of it can be attributed to luck uh i think because it's a race you don't have to be you only have to outrun the next fastest guy and what we're seeing is that a lot of the times the next fastest guy is is not putting in all the work that's involved they're cutting corners so that's where we're able to edge them out and and get ahead so where i think uh the more traditional mindset is that it's an element of luck because you know something goes viral or some piece of content does this i think what we're starting to see now is that while that's still true, you can have a piece of content that suddenly goes viral and everybody falls in love with your brand and blah, blah, blah, blah.

You don't need to depend on that. If you're putting in the work and making sure that you're working with the algorithms properly, you can still get mixed in there. And then to me, that's not luck. It's a repeatable process.

And actually we've seen it because we've done it for other brands. and the repeatable piece is falling back on the fundamentals of marketing, understanding the strategy and the framework of, you know, people first hear about your brand. What is it that brought them in? What is it that grabbed their attention?

Make sure that there's a connection between the attention and the interest piece. Make sure that you retarget them often enough. You have the multiple touch points so that you build up that desire. You know, old school marketing, make sure the branding is on point, making sure there's no red flags, making sure that people start imagining themselves using your product and service, and then ultimately make a frictionless action so that they can just easily, easily take that final step.

Those fundamentals, I think, have been lost a lot because people are now relying so heavily on tech and AI. And while our strategy is reliant very heavily on tech and AI, I think we've struck a balance where we recognize that while the tactical piece keeps changing and evolving and how we're doing these things is constantly moving, the understanding of the why, that strategy piece, that higher level thinking, that hasn't changed. And the successful brands, you still see that level of thinking.

You still see that human nature is still human nature. Tech evolves very fast, but humans evolve very slowly. And ultimately, by mixing those two together, we've been able to repeat the same success over and over and over again by understanding how the algorithms work with us. And we're not working against the technology by layering on that higher level thinking, talking to our clients, doing the hard work of making sure that the data is accurate.

We just spent my team just spent four days like my dev team just spent four days tracking down a credit card tracking issue on one of our smaller clients. that's the type of thing that larger agencies are going to go like, oh, there's a problem with the data and tell the client, like, go figure it out. No, we're like, no, this affects, this is affecting our cost per acquisition. Let's make sure that we dial this in and let the client know that we're fixing it, that working on it.

That's the type of hard work, I think, that makes it so that it's not luck, it's work. It's work that other people are skipping over. It's work that people are trying to not do because they're trying to leverage AI and I think they're doing it wrong. Great.

Well, Sheldon, unfortunately, we've run out of time. This seems to happen at every one of our technology sessions, but I want to thank you for the insightful discussion we've had. And what stands out is that success in modern marketing isn't about luck. It's very deliberate.

It's about building systems that learn, adapt, and improve over time. Drive marketing is a great example of a powerful example of how combining data discipline with advanced technology can turn marketing into a true growth engine that's predictable, scalable, and measurable. So thank you again for being on the podcast this morning. And just like Sheldon helping companies unlock performance through data and AI, Performix builds custom AI agents designed to reimagine how businesses operate, boosting innovation, efficiency, and profitability without massive upfront investment.

So Performix mission is really about AI for every business, and it's about making this technology practical, accessible, and results driven. So if you're interested in how AI could transform your organization, please visit performancebiz.com. Thanks for tuning in to Technology Today, and until next time, stay smart, stay data-driven, and keep innovating.