GUEST INTRO
Swami (00:02.2)
Hello everyone. Welcome to another edition of the SuperCMO Contramans podcast. In this episode, I’m talking again with Scott Desgracelius, CEO and founder of Vickitables, a leading attribution platform based out of the US. The last time I spoke to Scott, we really had a lot of fun talking about attribution, the challenges marketers face with big tech platforms in terms of
attribution and how to optimize their marketing spends. Since the last time we spoke, AI has disrupted attribution algorithms as dark funnel AI and visible AI referral signals are increasingly difficult to track and measure. Scott spoke to me about the new platform that they are releasing and publishing as a part of the AI led transformation that Bickett reports.
is making and he really then broke down his experience of how attribution is changing with AI. He specifically spoke to me about why websites must be AI ready and how they must be tagged for research-driven conversations. He also mentioned the fact that increasingly with AI, brands will become more and more important as your
three or four steps away from a click in the AI world. Therefore, he said marketers must be conversant with using attribution as not just reports that they look at, but having a conversation with AI platforms. Wicket Reports is becoming an AI led
platform where you could ask questions and answer will be there given the vast amount of knowledge they have built over a decade. So as a marketer, he mentioned that it is important to ask the right questions because the answers will be embedded in attribution platforms. So if you’re a marketer asking the right questions to
Swami (02:20.994)
manage your spend and understanding of how to build your attribution will become increasingly important. So you need to know the context before you ask your question. And that is going to be your biggest competitive differentiator and the expertise that you need to build. Let’s go in and dive into my conversation with Scott Desgrosseilliers.
EPISODE TRANSCRIPT
Q1
Scott, I wanted to you know kick off this conversation with the impact of AI, right? and that has been absolutely you know earth-shattering in the way attribution is going to change, right? Because the AI dark funnel as we call it, okay, is really, really you know, getting up to speed and literally the AI referral signals are increasingly difficult to capture, right? Because the conversation happens pretty much inside the you know the AI chats, and therefore typically your ability to do attribution is becoming far more difficult. So, can you talk to me about what’s changing since the advent of AI in the attribution you know, market.
Scott Desgrosseilliers (01:56.897)
Sure. There’s a couple phases to that. There’s tracking, did AI refer? And then there’s tracking, did those referrals, you know, first, are you getting traffic from AI? Then there’s is that traffic amounting to anything? And now there’s AI advertising. I’ve been experimenting with OpenAI Chat GPT ads. I can talk about those. And then the the biggest one that we have been grappling with is that everyone’s bolting AI on top of their data, either chatting, uploading a CSV, using an MCP server, or chatting within. But no one’s really asking the question, at least I’ve seen, is that is the data you’re pointing your AI at true?
Because AI will take the data and then use what’s called generative AI, which is amazing, it’s awesome, but it’s probabilistic, which without getting too nerdy, it means that it’ll approximate or guess things or fill the gaps in. And when it’s writing something for you, then it does a fantastic job. You’re like, that’s great. But with data analysis, it’s quite risky because you need something that’s what’s called deterministic, which has rules and has if-then-else and has don’t do this if that and you need those rules which is based on a lot of you know expertise and you know measurements of unique fields just like anyone any unique niche and without that expertise you get steered the wrong way or you miss a lot of opportunity.
So I’ve been grappling with all of those. So w which one do you want to dive into first? I want to I’d like to talk about all of them with you though.
Swami (03:35.927)
Yeah.
Q2
So first talk to me about how is AI changing this whole you know method and the algorithm of attribution. Maybe you can talk about that first because many of our marketers who are listening to this conversation, I’m sure are grappling with you know, how do you really build attribution given how AI is changing search, right? So therefore, that’s something that I would want you to talk, and then we will deep dive into many of the points that you just mentioned.
Scott Desgrosseilliers (04:11.4)
Okay. So yeah, what you’re gonna find even w with any attribution tool, you should find is that your search traffic is going down. because AI’s take you know, Google has been foundation of a lot of people’s businesses and now most of the real estate is a it starts as a chat and now it’s changing to where you can chat back and forth with it. So those if you’re not on the first page, it’s gonna be a very desperate searcher, which you know, if you do get those clicks, they’re probably quite valuable because someone’s willing to sift through, but everyone’s getting more impatient and used to a chat result that they’re not really gonna do that research unless it’s a real pain point.
I found it with a number of things recently where I had very niche things I needed to, like I had a weight dumbbell set of my sons that wasn’t working right. And the AI kept giving me the wrong answer for the wrong product. And finally I had to scroll through. But it was only because I had exhausted the easy button. And so that’s happening everywhere. And so your attribution tool, anything that’s accurate, should show less traffic coming from search. That’s a pretty much gonna happen across the board, unfortunately.
But then how to change that is you know, there’s a number of there’s a couple different ways of talking about it, but it’s that you need to get more content that AI likes. So that starts with marking up your website AI friendly. And then you know, you can talk to AI to get all the specifics. So I won’t go into all of them, but you can there’s specific ways that AI likes to read your site. And then they it needs unique information from you that it’s not generic resourced from elsewhere. So we’re gonna start publishing unique research that I talk about. We’re gonna start posting our unique research in an AI marked-up format consistently. Then there’s also you can, well, this isn’t for everyone, but if you have an MCP server, which is a way of connecting your system to AI, you can give it instructions that help the AI learn more about your product or offering. And this isn’t just for data. I saw someone that did a really good job of it for hiking trails. And they created an info source where it just pulled in some of its hiking trail data. So if you said, hey, I want to go on a hike and I have a nine-year-old child, it would then tell you which hikes in your area were best.
Swami (06:28.366)
Okay.
Scott Desgrosseilliers (06:42.55)
Which was a really unique way to do it, but also then AI starts referencing your product more, then that spreads through its brain, basically, and then it becomes more well known. So there’s some unique things that are very early adopter type things that can get you a big leg up. And I think that’s one not a lot of people have done. Because you think MCP server and it sounds like, I have a big complicated software. I mean, we have one coming. But it could be for any business niche.
Where you could post and just post some basic stuff up there that’s gonna help people chat with your expertise better. And then you can so that’s one that I would highly recommend people start looking into because it doesn’t the the technical hurdle hurdles aren’t great. I’ve been deep into AI and it’s not insurmountable to create these things. You just need to have unique knowledge translated in a way AI likes. And so that’s where the work is, but it’s a it’s a it’s a one two-month project, not a a year-long one.
Q3
Swami (07:44.079)
So how do I convert my website into AI friendly research content?
Because sometimes what I really read from whatever I’ve kind of learned is you know, it’s almost like you will start giving it a direct attribution when actually you know AI has actually influenced okay your visit to the website and things like that. So therefore, your ability to kind of add build attribution which is influenced versus direct itself is actually changing given that you are doing most of the stuff through AI search, right?
Scott Desgrosseilliers (08:59.434)
Yep.
The first thing you should do, and anyone on this can do this very fast with AI, AI will help you help yourself. So the tagging is not elaborate. It it’s basically so tagging for those that I don’t know what no, normally a CMO level will understand this, but HTML is a way of tagging so that your browser understands what to display. And then behind the scenes, there’s tags that don’t display that used to be called meta tags, which would tell search engines about it.
Very similar, there’s certain tagging so that AI will understand the intent of the page, what it’s about. To take it to the next level though, AI likes it it’s a computer-type brain, so it likes like FAQs and how-to guides. So you’ll want those on the page maybe at the bottom or if it doesn’t fit the page you can put them in just… it’s not going to show on your maybe on your page because you like the layout, but behind the scenes you have a how-to or explanation for AI in the page.
And you can do this by uploading your page to AI and saying, hey, I need to be better more discoverable on AI. Here’s my website. Here’s my page. The intent of the page is this for this. So you want to be specific, not like it’s about marketing is a great way to have no one reference it. Like we’s ours would be like for for a particular page, this is about marketing attribution for people using Shopify and spending at least $20,000 a month and struggling because Google Analytics 4 doesn’t give them what they need. So the the longer the prompt that’s more informative, the better, and that’s generally true with AI in general, the more detailed it will give you exactly what to copy and paste into your website editor.
And it’ll be just this tagging where you’ll go to like the code version of your website and you’ll paste it in at the top or at the bottom. Or if you already have a script area that might like with like HubSpot’s for our website, there’s a little area where we put the meta tags for a search. Well, we’ll add them for the AI as well. And then you just you pay you paste it in whatever the tags are that it gives, or you have to might have to create some content and put in there.
Swami (11:13.047)
Okay.
Scott Desgrosseilliers (11:13.088)
You paste it in, you hit save, and then you recheck your website, and it shouldn’t show anything new unless you edited the WYSIWYG, the what you see, what you get area. But behind the scenes, AI will now know this stuff. You’ll then go back to your chat and say, hey, I just updated. Can you confirm what would AI now interpret the page to be about? And so you do that.
So that’s the mechanics per page, but at a higher level strategy, real quickly, you would want to say what say I need to get more visible to AI, and then you would say who who you are and what you what target market you have.
And so for us, you know, marketing attribution for e-commerce brands, five to fifty million in revenue, they’re spending at least twenty grand a month. I want to be more visible for them. And then it will spit out all the questions that they’re asking AI, which is valuable because AI will tell you, here’s what people are asking, which then is the type of content you should be creating.
So that’s a big t that helps you anyway with your strategy. But then based on those questions, it looks at your website and will surface probably some areas you didn’t think that you needed, or other pages how you might need to repurpose, or some that you just want to text. Be some laundry list of action items and you just go off and there’s just work. You got the you get the, you know, make someone. But it’s not technically difficult work is what I want to explain. It’s just…it’s more you’re gonna have the expertise in your niche, but you should have you don’t have a business, hopefully.
Swami (12:42.317)
Okay.
Perfect. So therefore, the first thing in the attribution in the post-AI era is about making your website AI ready, right? That’s the first step that you’re talking about. How is this connected with your MCP server? And can you just explain with an example on if I were a brand, you know, say e-commerce brand, looking at say beauty and cosmetics…how should I be looking at you know using what you just said, which is an MCP server? how do I connect my content? And just if you can explain it with some analogy and you know example, that would actually really excite the listeners.
Scott Desgrosseilliers (13:19.702)
Sure.
Scott Desgrosseilliers (13:30.505)
Yeah, actually that beauty and health, it’s a great example.
So everyone’s chatting with AI now and getting excited. Everyone likes new shiny stuff. So an MCP server, it it technically sounds daunting, but it’s not. What it means is I’m going to install something in my AI for me that’s going to connect with my user credentials to some system and then pull out either the data that I have in that system or the this is where it gets interesting, or the unique skills or knowledge of that brand. And that’s where I want to take it for the beauty and the and the makeup.
So beauty and makeup, let’s contrast it with what we’re doing, because ours is more a logical step. So we have proprietary data and we have proprietary ways of analyzing things that we want, we want our point of view in the AI with our data that’s different than other data sources they’re gonna reference. So there’s a way we need to instruct the AI.
Hey, this is why we’re different. Here’s how to use our data, how to compare. We have to give it instructions because then the AI takes the data in, uses our instructions to help get the user to the outcome they’re trying to get to, whatever question they’re asking. Now, with a beauty per let’s just say you’re trying to make your I don’t know, your eyelashes look thicker. I have a couple daughters. So I’m trying to get my eyelashes look thicker for the prom or whatever.
Swami (14:54.882)
Ha ha.
Scott Desgrosseilliers (15:00.476)
So what a beauty brand can do is have an MCP server. It has to be very easy to install. But basically they have a a unique way they’ll be able to capture leads this way. They can say, hey, use our, but you market it, use our AI tool to help look better for the prom or whatever.
And so people, you’re also gonna go, great, I want to do that. So they click a few buttons and then they’re in with their chat GPT on their phone or Claude and probably chat GPT, although Claude’s more for the or engineer types. And then when they’ve connected it, it means that whatever you’ve put into that MCP server, which we’ll get into in a second, is now the AI is going to use whenever there are beauty questions asked.
So when they say, and then you’re going to and then so behind the scenes, the beauty brand is gonna be it’s basically it’s text files and you’re typing up your expertise. Like what makes lashes more thick is, you know, something to do with the the brush that they use. I don’t even know the terms. The brush that they use on their eyes and the you know, all the gobbledygoop marketing that goes into your lashes are thicker. Well, they put all that knowledge in there called a skill.
And so then they tell the AI, hey, when people are looking to make their eyebrows thicker or eyelashes, use our eyelash thickener skill. And it’s a list of their expertise. And then separately they’ll have one that says, here are the products we recommend if they’re trying to do that, which is of course what they really want. But you gotta lead with advice first. So then you’re in there chatting away, you get a helpful tip, and then naturally you want products. You know, we constantly have Sephora coming to the house. It’s like mind-blowing how much money we spend on this.
And so then it’s going to naturally start recommending you and then it will learn locally on that person, but it can spread then if it’s successful. And so it’s a real cutting-edge thing where you can get your expertise and point of view in there. And if you do it right and it’s not all sales pitchy, then it can spread.
And furthermore, then you know how I mean how my daughters and all her friends talk all the time. Obviously they’re like one hive connected mind themselves. That spreads, they’re gonna say, hey, look, I’ve got this cool AI tool that helped me with my makeup or whatever, and then other people will go download it and you get a lead magnet that’s very unique, and then you always have your you’re getting your best chance in front of your customers of being recommended.
Swami (17:24.76)
So so so I’m trying to you know get this right. So this AI tool sits in your website or does it sit you know independently in a you know chat GPT or in a Claude chat? Where does this sit? And therefore if I were say The Body Shop, okay, or if I wear Sephora, you know. Are you talking about the AI tool sitting in my website, or are you really talking about this AI tool being there in my you know, because I have Claude, I have Chat GPT, and I’m chatting with it. So where is my brand really coming in into my con in the conversation and how is it connecting with the expertise skill-based MCP server?
Scott Desgrosseilliers (18:17.75)
So you ha you or you hire someone that creates the MCP server, which is it just lines of code. It doesn’t actually have to sit anywhere as a server, it’s just code in the form of different text files that know how to talk to each other.
And then it it’s in a format that people can install it into their AI. No different than when you install a browser add-on. Shoppers install like Chrome extensions for coupons. Very common thing to do. Commonplace now, hundreds of millions of people use them. When I first started, it was kind of like, huh, what? I got to install some of the browser. It’s going to be very commonplace soon. So this is like cutting edge to do this. But it’s installing something in the AI so that your brand’s instructions are in the AI.
And so it’s a very new space. So people can’t you you can’t just go in and say, hey Claude, always make people buy my eyelash stuff. Like it won’t do that. This supposedly safeguards, but you can put your expertise of makeup or beauty in there and then recommend your stuff as a adjunct thing. what I like about it is that your point of view can get in there. Your actual expertise, it’s not just you know, generic stuff and then because if people are responding well to it, then Claude will continue to use it more. So it it it’s yeah.
Q4
How are you transitioning Wicked Reports as a platform, right? if you can break it down into three or four steps, how you’re changing it, that’ll be very valuable in terms of how you are transforming it and therefore what outcome and impact will it have on brands.
Scott Desgrosseilliers (25:53.758)
Massive. Massively.
Scott Desgrosseilliers (26:10.602)
Well, that sentence is is where we’re focused. Like now we have all this data, we’re not building more reports. It’s all about what’s the profitable outcome from all this analysis. Like I have playbooks upon playbooks of doing this for eleven, twelve years, step-by-step checklists… in-app walkthroughs, all these things. And it still involved people had to think. Attribution has a lot of complexity. It’d be like flying a plane. You have all these different gauges and certain gauges you only need because it’s nosediving to the ground, or certain ones you need ’cause you’re crossing the water, you’re in a fog. Other times like you can see the airport and you just gotta gonna dr guide it in and you don’t need a lot of guidance.
Attribution in the same way. There’s a lot of complexity in it. And what I’m ex most excited about is AI is allowing us to realize all the potential in the data we’ve had. So we now have what we have coming, it’s about three to five weeks away, depending on how the testing goes, is that the outcomes are on a literally served on a platter like a buffet at a restaurant. You pick, like, I want meat today. And then we have all the meat entrees. This could be the customer LTV, or you’re focused on new customer acquisition. And we have a menu list of all the outcomes we can give you. And you click on it, and then it pops up with the answer. And then you can chat with the answer to get further insight if you want. Or you’ve already got the answer. You don’t even need to go, you can go look at reports. I love to look at them, but you don’t actually have to ever look at a report again. Only because you care to validate what we’re telling you.
And what came this came about, this was always my goal with Wicked Reports, is to realize the potential that we have in the data so that you always know what your next best customer is trying to tell you. Because they’re out there, they’re looking, hopefully have the answers for them. How do you get them? How do you get to them? And so all these different playbooks are revealed, and it’s because we we didn’t know this at the time, but we built a very deterministic AI, which was very rule-based. Remember my product manager was very stressed. He’s like, There’s not enough AI. It’s still like what we know. And I was like, Yeah, but no one’s not enough people are doing what we know already. Let’s just get that in their hands. And now what we have is a hybrid where we start deterministic, which means we start from our expertise.
So let’s give an example. I want to improve my new customer acquisition. Where should I spend my budget? That means a lot of things in attribution. You gotta pick the right model, you gotta pick the right time frame, you gotta look at the right segment of traffic, and then you gotta compare it against how you normally do versus what you actually did versus what you’re forecasted to do. That’s all these different steps.
That we all have and we get all excited, we demo them, the prospect, my god, this is gonna be amazing. Then they get in there and they’re busy. I don’t have time to do it, maybe. And now all of that power is harnessed, and we just start with giving them the answer go put your spend here. Why? And then we can say, well, and then we can unleash all the reasons why the data supports it, but it’s data because it’s deterministic.
It’s actual facts of attribution backed with us doing this for twelve years. Back with what’s most important, which is the foundation of the data has to be in a way that accurately attributes new customers correctly. So if you have the you need the right foundation, otherwise the the first you need the right data foundation, which we have, then you need to get to the answer, and then you need to back it up.
And so we have all that now on a platter where each each click is a prompt to our MCP server, a prompt that’s like 600 lines long, but it’s hidden. With all this expert all these agents below, these little mini agents that are experts in their little fields of expertise, and they all inject in if they need to. We have this orchestration agent that looks and says, Okay, you’re trying to answer this. I need these particular global skills, and I need to call these two workers, and they’re gonna work together with these skills to give you the answer, which is all based on our eleven years of work, twelve years actually.
And then if you want to chat back and forth, it becomes generative, where it now we have a correct foundation of the data and the expertise it built into that chat window. So that now we know if you freestyle and start asking other things, we’re comfortable that the generative AI, which is amazing, is gonna have the right foundation. So it it has the guardrails. It’s not gonna go off and start talking about taking taking a cruise on a boat when you’re looking to stay in a hotel room, for example.
So that’s like I’m really like proud, but also like it’s it’s our end game because now we’re giving and now it’s just a matter of are the out do you understand the outcomes? Do you trust the data? And if not, we can dig into that with the people and customize it, but we’re pretty sure we’re accurate. That’s what we built the business on.
Swami (31:21.848)
So let me dive a little deeper into what you are saying. all this data that you have is the learnings that you’ve had on multiple categories, multiple spend levels, multiple geographies. you know, I’m just giving an example, you know, multiple customer segments, multiple price points of products, okay, and you had some amount of learnings out of that which is anonymized, and that is something that you carry in, you know, if I were to call it you know the wicked report.ai is really something that you have, and you then start asking questions to say, Hey, we I am Wicked Reports AI attribution advisor, now tell me where should I you know spend my money if I’m a beauty and a health brand. Okay. And you basically give out answers which will then allow you to really optimize your current marketing spend, then throw that data back into your Wicked Reports platform, and then further optimize it based on the results. And therefore, it’s a kind of a feedback loop that you’re building as a part of your AI platform. Is that the way to kind of understand it?
Scott Desgrosseilliers (32:47.146)
That’s a great summary of it. You know, I said I had new talking points, so I’m a little boisterous with some of them. Not I’m not as concise. That’s a better way to say it. I appreciate that. Because we have this thing, this this very taxing thing for people to mentally get. And when they do, they become very they really enjoy Wicked Reports. And until they get it, they don’t. And this and it’s leads take clicks take time to buy. You don’t you do get last click sales, but they’ve generally had some up funnel influence.
And now we have that built into all the analysis. So you don’t have to go off and cross-reference cohort values and past convert. We call it the lag curve, conversion lag curve. So we factor that in. So if you’re looking at your new you’re trying to buy new customers, and look we have all your historical conversion data, and we know it takes two weeks for them to buy. And it’s been four days and the performance doesn’t look good.
Well, we have your historical data that we’ve learned on, and yes, other brands in your niche and your AOV and your LTV. And so we can forecast out now, which is new, we for have the forecasted amount you should be at, but where we’re forecasting it’s going to go based on your other new customer acquisition activities. So we can be much more confident and say, hey, we got 85% likely you don’t have to panic, even though the ROAS looks bad.
And so that’s the type of stuff that we didn’t have access to where it was just too much cognitive load for you to do all that. Or people would export it out and have their own complex spreadsheets color-coded with all these other indicators to try to do that. And now we’re just gonna be able to do it for them. So it’s it’s exciting for us.
Swami (34:16.973)
Hmm.
Swami (34:27.96)
Fantastic.
Q5
So given that you know as a platform now you have all this data and with AI coming in, how do you build the referral signals for this attribution? Because one is the way you’re now spending money is you know it’s moving from SEO to GEO, which is really your website getting AI ready, which means that those tags are giving some referral signals to your platform. That could be one way of building the you know referral signals for your attribution.
The second one is you’re already spending money on Google and Meta and things like that, and those signals are also coming in, right? And therefore, you know, some could be an awareness campaign, some could be an acquisition campaign. So you’re getting all that data back, and you also now are saying, you know what, you’re using your traditional expertise of what you really say the analytics and the insights that these platforms give you may not give you the true picture, right? So therefore that’s really where you differ. Can you explain that a bit to our listeners?
Scott Desgrosseilliers (35:44.319)
Yeah, so that difference, that’s w what the value is in our data and the way we measure, should be different than the platforms, or if it if if you sign up and it matches, then you you shouldn’t pay us because it’s the same. You’re paying for nothing. That difference happens for a couple of reasons. One being that all the ad platforms are grading their own homework. And they’re generally grading it biased they don’t want you to stop spending with them and spend on their competition, which is the other platforms. And they have a last click bias, which is whatever happened last. Particularly with Google brand search keywords being the most flagrant offender.
That brand keywords meaning if like for me, Wicked reports, if I start if I was running ads for include the word wicked reports, those are always the best performing keywords because people are already brand aware and typing in your brand to go there. And then you run ads supposedly to protect your brand, but they already know about your brand. And so it’s not generally where you should spend much of your money. But there’s also the the so first of all, we’re trying to be the independent, non-biased measurement of what’s happening.
So i if everyone’s claiming for a sale, which your clavio email will claim a sale for 30 days, your attentive SMS text will claim a sale for 30 days, your Google branded search and your Meta and your TikTok, all five likely with 30-day look back will all say, Look, we the made the sale. And what our job is to say they all touched it, potentially, potentially not, but where do we move the money?
Swami (37:26.616)
Mm.
Scott Desgrosseilliers (37:26.844)
So that you can be more profitable each time or at least find out you’re wasting money and stop spending it, which technically makes you more profitable. So we’re worried about most profitable decision, most profitable outcome for you as the person that owns the budget. That’s what we’re trying to do. So the first part is we’re traffic copying all five of them claiming credit. We’re trying to discern the truth. The second thing is because we measure for a longer period, we have infinite look back and look forward from when events happen.
We can go way back in time, if need be, to say, hey, what started this journey for this customer? Because the hardest click to earn is the first one. And so we’re always trying to be like, hey, what got them first interested in your brand? And we find enough of those reverse engineered journeys. That’s where you want to spend most of your budget. Because what you’ll find is once all the retargeting and all those other, that’s valuable, but you need to spend more than half of your budget on trying to find new people to first become aware. That first click or two. And usually before the click, there’s viewing and there’s AI chatter happening before. And so we’re trying to always figure so our strategic edge is one, you don’t overspend on bottom of the funnel somewhere, but two, where do you spend your top of the funnel dollars that lead to your overall business goal, which is usually new customers. You want new people. And so we measure new versus repeat, whereas the platforms blend it all in together because it looks way better when half of your half of the conversions they claim are probably repeat buyers that are already buying off email. Well then like Meta or Google says, hey look, I sold another thing for you, and they did. It was your email that did it.
Q6
Swami (39:08.75)
So given given that given that point that you’re making on the advertising side with Google and Meta, but on the AI side of the customer journey, it’s pretty much dark, right, Scott? Because you know, all my conversations are not tagged. All my conversations are you know, I would say not really visible. Okay. And given that, the first touch or the you know what I call as the first touch attribution is becoming harder with AI, right? So how is Wicked Reports now solving that problem in the new platform that you’re building?
Scott Desgrosseilliers (39:44.714)
Yes.
Scott Desgrosseilliers (39:50.879)
So the the undiscoverable is still dark to us at the moment. The best thing I’ve seen, but it’s not accurate enough to count on, is if you continually ask, you have to know the questions to ask, but you can ask the AI to score you on how how many, what percentage of certain queries they are referencing you, and they’ll give you that. It won’t be to a user deterministic level.
So we’re in the progress of figuring out how we’re gonna deal with that. And and I we are because I my mind got changed on one thing, which was I’ve always been against view-based attribution, but now I realized a way to allow it in our platform, which is to be transparent about it. Because all these people have models where you don’t know what the inputs are and you’re just supposed to trust the model. And I already don’t like things where you don’t know the logic, how you’re supposed to trust it.
So we created a user-controlled model where we set the settings based on your performance, but you can see our settings and change them, and it’ll immediately change the impact view has on your performance. So if we can do something like that for AI, we would add it in. It’s still, I don’t have the trustworthy percentage source versus the questions to map to the visits yet.
That makes sense. There’s still a missing it because it’s like view based. basically the the the AI saying, hey, we had a view impression here. Take our word for it. So we’re going to treat it similar. It’s just getting the reliable data feed we want to use. No one has one yet. I’m leaning towards HREFs version. They’re very solid with everything. So we may use their source if we can, but we haven’t. Still a it’s still a it’s still a blind spot temporarily, unfortunately.
Swami (41:48.309)
Okay, yeah, not so so true. So therefore, literally would you therefore say server-side tagging is going to become more important though it’s viewer based, you’re kind of looking at server-side tagging, which is really saying, Okay, you know what? So many signals came from people who searched for you know, dry skin, for example, and they came to me on my website. Maybe there’s some kind of a signal that you get. And from there on, I think you’ve got to find new ways for attribution given how AI is evolving, right? So I think it’s really not clear as yet, but I think the ones who are going to really crack this are the ones who are able to f think from first principles, right? It’s not going to be easy, and that’s really what you’re talking about, right?
Scott Desgrosseilliers (42:26.612)
Yes.
Scott Desgrosseilliers (42:43.56)
Yeah, well it you align it to business outcomes. So like what we did with views is that…you know, if your new customer acquisition cost blended across all your marketing is where you want it to be and you are growing healthily, then theoretically whatever your current KPI benchmarks are per platform are good enough.
And so that the the click-based accurate attribution can still be the command center directional guidance you need, even if what at a click bait basis it shows less profitable than you wish. Like on a like a Snapchat, which is much less clicking happens. The the return on ad spend can look low click-based. But if when you increase Snapchat spend, your new customer, your blended new customer acquisition cost stays the same.
And more new customers come out, then you got the outcome you need. So we can infer there. So that’s what we released with what’s what we have coming with that this summer release is that the ability to infer transparently against the business value. The mistake that happens right now is the ad platforms show view conversion from their limited siloed view, not taking your business into account. So it shows a sky-high ROAS that you’re not realizing.
But if we can tie it to the the bottom line, your dollars in and out, then it’s worth then it’s more trustworthy, in my opinion.
Q7
Swami (44:10.848)
And and therefore if I’m getting traffic from say five different you know AI platforms, right? I you know, I get from Claude, I get from Chat GPT, I get from Perplexity, then the real the real challenge that you’re talking about is on my website, I need those meta tags, okay, which will actually direct me to you know, saying that hey, you know, you get you’re getting these kinds of referrals, and these referrals are actually getting you into your website, or it’s getting you into you know some outcome that you want, and therefore you’re almost saying that this area is now opening up, given the fact that you know this is a new area, and therefore the whole idea is your past knowledge becomes your corpus for your asking these questions and probably you’ve got to do a lot of experiments and start learning with it. And that’s your competitive advantage as Wicked Reports, right?
Scott Desgrosseilliers (45:13.556)
Yes. Because you’re always just trying to beat your own benchmark. Really. I mean, we got benchmark data against other people, and it’s fun to see how you compare, but whatever your your current goal is, you’re trying to beat it. It’s the same thing like weight loss or running speed or you know, whatever your activity is, you’re trying to improve yourself. It’s no different with attribution. You get your own benchmarks based on accurate measurement… like we’re measuring things correctly. So if you go into an ad platform, it’s almost like you’re weighing something when you should be measuring the distance. It’s literally that different, the measurement philosophies in some cases. They’re just completely different.
So yeah, you you iteratively improve is how it works. And with the AI, you’ll keep wanting to if there’s a question you keep wanting to answer for and you’re not getting the visibility, you gotta keep modifying your tags and your content and then going and waiting and then refreshing and seeing if it if it worked or not. There’s no guaranteed blueprint, which is why it’s the Wild West right now, which is kind of fun. Just depends. It’s experimentation.
Q8
How do you see advertising evolving in the era of AI? Because you literally had search, search led to you know bidding of keywords, keywords led to you know clicks clicks led to you know a whole host of way you would attribute your spends and optimize your spends that’s really how the internet era built your advertising model right what’s your prediction of how will ai advertising model how do you think it’s going to pan out if you were to kind of look at it into the future?
Scott Desgrosseilliers (47:10.548)
Yeah, I think aver they will get better with their paid advertising options because there’s so much r I mean, you’re looking at how much I mean, Meta and Google are like, you know, hundred billion dollar profit businesses. I’m sure AI wants to get their hands on that spend. So it’s very likely it’s going to get better.
So I would anticipate more embedded advertising is going to happen because the money’s too great. So that’d be my first prediction.
Second one is the the importance of building a brand is going to become more important. And owning your audience because the audience is so hard to capture now and it’s gonna be even more obtuse or you’re gonna be more removed from them, less you can’t just pay, get a click. It’s gonna be more effort to get that paid click.
The importance of offering enough value to capture that click, capture that email address, that those email addresses are gonna be much more valuable. I mean, they’ve always been really valuable, but now they’re like, you know, gold. You can actually cut you know you can reach them. So without having to pay.
so I think those are the three things that I mean, br and building brands always been important. I think it’s just to be huge, hugely important to build.
Yeah, because that’s your asset, you know.
Q9
Swami (48:30.328)
So you so those are the two important things and do you s do you foresee you know first party data kind of becoming useful in the AI platforms because I’ve collected a lot of first party data and how do I really match my first party data with these AI platforms? Do you think you know is there’s a new way of thinking with my first party data in my AI platforms?
Scott Desgrosseilliers (49:01.226)
Yeah, ChatGPT already has a conversion API. So they’re already doing matching. Yep. And then also, I mean, the new the Google checkout AI, if it takes off, like to Shopify, where you can an agent can buy. I know that’s all hype-y. It’s not a lot of people aren’t doing it now. It remains to be seen. It might. It might take off. That’s, you know, a a transaction that then will need some. The first party data was triggered by the AI. So you’ll have seamless attribution there. Because then we can pull from Shopify or whatever your card is that the AI tagging in there, it happened because an AI shopper. So that’ll be cool. It’d be fun and e it’d be easy to attribute that ifit takes off. We haven’t worried about it yet, because not a lot of people doing it.
Swami (49:51.311)
Fantastic.
Q10
And how do you see the Wicked Reports 2.0 evolve in the AI era? So what do you think are the two or three priorities that you feel marketers will find valuable for the new transformation that you’re doing as a part of your platform?
Scott Desgrosseilliers (50:09.502)
Yeah, so I think the biggest will be decision certainty. You’ll have the decisions on a platter for you, and we’re lock now we’re logging them so you’ll be able to log and say, Yes, I’m gonna do this and no I’m not. Because we can we’re guaranteeing three times what you pay us now because we’ve certain that we can find three times what you’re paying us in decisions. So it’s outcome-based attribution pricing.
So that’s the biggest one. The second one is we’ll have more control over our customers’ experience because we as the data unlocks insight, we can deliver it to you on a platter. I mean no like hunting for the data or long emails imploring you to click here and do this and that.
We’ll just tell you, hey, go do this. So that’s fun. And then the third thing is we’ll get into you know, TV attribution and creative insights now in the fall because we’ll have the decision piece solved. So now we can move just where else can we help you make decisions is the third piece we’ll move into.
CLOSING
Swami (51:15.854)
Fantastic. Thanks, Scott. I think thanks for the preview to the product that you’ll be launching in the next three to four weeks. it’s interesting. I think it’s almost like the Wild West, as you said. A lot of questions were you know answered, a lot of questions people are experimenting and figuring out, but you’ve been a first mover and the fact that you’ve been able to experiment and build this.
is giving you a first mover advantage and I’m sure when we meet the next time there’ll be a whole host of learnings that you will be able to share with our audience and I am looking forward to the third conversation soon. Thank you very much.
Scott Desgrosseilliers (51:56.798)
Yeah, me too. Thanks, Swami. Always a pleasure. Great questions as usual. Take care.
Swami (52:01.858)
Thanks.
END
