Your Conversion Rate Doesn’t Matter...

You can incentivize customers into bad behavior. >> Remember Publishers Inquiry? Ed McMahon would show up at someone's door and give them a million dollars. You'd subscribe to different magazines and that was your entry. And then the magazines loved it cuz they got a ton of subscribers. But then they realized none of these subscribers are actually buying any of the advertiser stuff that advertises the magazines. The original goal was like, well, they're going to renew. You get this for a year for three bucks or whatever it is. They're going to renew for $29. And very few of them would ever do that. The only time they would renew is the next year when the flyer came in the mail to get it for $3. Again, >> the thing is is that if you don't have a really tight feedback loop on observing what those customers are doing, you can get upside down really quickly by getting a customer base that behaves poorly. >> You're watching Marketing Misfits with Norm Ferrar and Kevin K. So, Norm, you ever wondered about your conversion rate in your personal life? I know you've been married for a long time. >> I have I but but what attributed to that conversion when you got Connie? When you got married to Connie? When Well, you met her when she was like 16 or something, right? Or early. >> That's right. 16. >> So, so what actually led to you getting that conversion? You did the final work. It was you and your charm and your your wit and your good looks that that that got it. >> But what led you to get that conversion? Do you know? Was it your buddy John? Was it uh something that happened? What led Do you Do you know what that was? >> Yeah. Uh um yeah, I was asking out the wrong girl. >> I was split testing. I was split testing >> because your AB split. So you're doing it the old the current way that most people do, which is AB split testing. But you know what? Sometimes attribution is not so good and you don't actually know. And and so I think our guest today is going to be talking about a different way to approach conversion whether that's how you actually look at it and actually do you want that conversion because maybe it converts to the like me I got conversion on my ex-wife but it was the wrong conversion. I had to send her back. I had to return her. So I had to actually send her back and return her. the negative negative conversion. >> It was negative conversion. So actually, how do you prevent that? Cuz maybe you don't actually want that kind kind of conversion. >> It's called a bullet though. >> So yeah, it's called it's called a bullet. So I mean when it comes to conversion, I mean it's and now with AI and with some of the stuff that we can do and with the tracking, you know, back in the dinosaur age when you and Connie met, you know, there there was just someone had to go knock on the tree, you know, to send the signal. Now, now we know we know everything know everything about every leaf on the tree. So, it's going to be I think a fun talk today and a little bit different um and maybe change some people's perspective on stuff. >> Yeah. And I can't wait to talk to Matthew. So, let's bring him in. Matthew Barnes. >> Matthew Barnes. >> I'm good. >> Good to see you, man. >> Yeah. Thanks for >> I'm just going to wait. I'm just waiting for you to leap over leap over a wall or something. You're like Spider-Man. Yeah. >> What's What's the I never heard of it. I I was looking you up and I was like, "Okay, what is this? Parkour parkour? Is that how you say?" Like, "Are you into that?" >> And I was like, "Okay." >> Oh, cool. >> Yeah. >> I watched a little quick little video on it. I was like, "Okay." First thing that came to mind was like American Ninja Warriors, but I was cuz they jump on all this stuff, but then but they have props, they have ropes, they have stuff. But you then I saw a little TED talk video. Uh, and it's the girl's like, "No, we leap on walls and we jump over fences and we do all this kind of stuff. It looks pretty cool." I was like, "I wish I could do that, >> but I'd like to see you do it." >> You norally see me face plant. >> Yeah. So, I started I started training parkour uh back in college. So, I went to Purdue Purdue University and uh back then it was a it was an online sport. So, it was like people all around the world who had gotten connected through YouTube. This was like back in like MSN Messenger days uh of of people connecting like in the UK and a couple of people in the US and some people in like Eastern Europe. Uh and we were just sharing these like >> 480 resolution like videos uh training and like coming up with new moves and things to do. It was like very much like early days of skateboarding like on YouTube basically. And it's really a mental sport. Parkour is really about training right on that edge of things that make you feel uncomfortable and scared, but are like physically really quite possible. Now, as you train and expand that circle, it can get pretty like impressive uh visually, like running up walls and doing back flips and all of that fun stuff. Uh but it's a really a fundamental discipline. My my son was uh he didn't get right into it, but for about a year or so, he was getting into it. And I was amazed because I got to see when I was living in Hawaii, it was unbelievable. The guys that were training in parkour, you know, I couldn't even imagine even starting, but it's just a slow progression, I guess. >> That's exactly it. Yeah. Now, I actually I taught parkour uh for a couple of years and the it was amazing because you'd have these people that would come in that would be like really physically fit, but they weren't like mentally fit be having a like a comfort with like how to manage like fear and emotions and that sort of thing. And then you would have these folks that would come in that be like a physics major who is like never done a sport in his life and like a year later he sticks with it and he's like become very athletic because it's really uh it kind of weeds you out mentally because there's no amount of uh ego that can jump a gap. Uh which is which is really the thing I love about it the most. >> Have you ever used it in a real life situation? not where you're like, you know, competing, but we're like, shoot, I need to I I lock I left my wallet in the in the third floor. Uh, and the door is front door is locked. Ah, no problem. I'll jump the wall and and and get it or or I don't know something. Is it >> I lock I locked myself out of my apartment and uh scaled up the uh the building up to the the second floor roof and jumped from the building across onto my other roof and then went in through my guest uh window. uh one time after. Yeah, >> that's great. That's awesome. That's so that's so cool. So So I I see though how it it is, but doesn't it isn't there a physical limitation? You said it's mind over matter where you think you can't do it, but you can. >> Yeah. Yeah. So the the it's more of like you can always find something that is mentally challenging that isn't physically challenging. You don't have to like have this like really high peak level of like fitness to be able to find something that challenges you mentally but can stay like quite safe reasonably speaking. >> Well, that applies to business too then, doesn't it? >> Yeah, absolutely. Yeah. No, >> when you get in that mental state in your business, it's kind of crazy what you might be able to accomplish. >> Yeah, it's it's and the the thing that I love the most about it is it's it's it's gaining a relationship with risk, Rick. Uh so uh one of the things is I I have two daughters. They're uh 10 and seven and I started training parkour with them. Uh like I I took them out into you know out in the yard, teach them how to fall, take them out like training, jumping on things, you know that sort of thing. And it's it's I love it because, you know, it's we get to share in this thing together where I'm helping them learn to manage their own fear and like assess risk, right? And I think that's like a really really good skill for anyone, but it does just massively apply to business. I use it all the time, right? When something something bad happens, right? You're assessing your emotions. You're trying to figure out what like the real risk profile is here, right? You're like figuring out how to calm down your central nervous system and make a decision like uh with with clarity, etc. It's it's highly applicable. >> No, I I think that's so amazing. And like you said, it it applies exactly to business, family, like it's all aspects of life. >> And especially when you're dealing and as entrepreneurs, you you sometimes you're dealing all the time, not sometimes, with crisis management. >> And if you have that mindset, crisis management goes away. >> Yeah. Yeah. >> And then it just becomes an obstacle you have to overcome >> figuratively or sometimes literally. Yeah, >> it's risk management too. I mean, what about injuries? I mean, do you get hurt doing this stuff? >> Yeah. I mean, the thing is is that done done well, right? It it carries like a similar risk to other sports. You know, in the UK it's like really really common for people to break their leg playing soccer, playing football there, right? Because they get slide tackled from the side, etc. And so, uh, one of the things we always talk about was you train to fall. So like uh one of the first things that I would teach a student is is how to fall properly uh so they can absorb impact safely without uh injuring themselves or how to take a drop from height uh or how to short a jump uh and we drill that stuff over and over and over and over again. So it becomes kind of second nature and I think any any good sport really probably has that component to it. >> It's a martial art isn't it? >> Essentially. Yeah. That's it. is like you're you're you're you are learning to sort of like uh have your mind and body like commune with the your environment. Yeah. >> How can you even apply this like I know you can because of the mindset but how can you apply this in business? Well, so I guess the the thing is, right, is setting goals, right? And like and uh and like understanding what it takes to achieve those things and like breaking them down into their components, I think is like really important. Like I had this uh so there's a there's a move that you can do. Uh this would Okay, so back in the day there was parkour and there was free running. Parkour was just the efficient movements and free running was more of like the stylistic like flipping things. And then over time those things kind of merge and people just call it parkour now. But there was uh a thing that I really wanted to do on campus. Uh I wanted to do what's called a Superman front flip. So uh what that is is on the second store like on the second floor of the campus university building there's like a railing on this balcony. Uh and what I wanted to do was do a dive over the railing and then at like like Superman and then once you get over the railing you pull into a front flip and then you land on the ground like below. And there's like a lot of pieces of that that come together to like build that move. You have to learn how to take the height properly and be able to go into like a roll. You have to learn to clear the obstacle. You have to be able to learn the aerial awareness of being able to open up in a flip like that early. There's lot lots of different pieces to it. And so again, just like in business, right? You may have like some goal you want to achieve. You want to do some like high tier thing like that, right? And you have to break down like okay functionally what would I have to have in place to make this thing like happen right and so actually spent like six months training for that move. Uh we set up like really similar scenarios in like a gym like a gymnastics gym and like practice like learning the timing of like popping out and then doing the roll and then took it to other places on campus and like drilled it there so that when I did it, it wasn't just like yoloing like this like really hard move like for the first time. I had done all of the component pieces, right? and I had like worked toward uh building up to this like big thing. And I think you know in in my business we approach things in like a similar way which is like you know we're we're trying to figure out like instead of just like hey would it be cool if we just like built this thing or we did this thing like etc. It's like what are the functional pieces of this right that necessitate precipitating the thing because I think you know in in business but in in this as well like when we when I think about a goal of something we want to achieve or something we want to build like we have this really fantastic software uh that we've we've built in our business we didn't start by just deciding what software we wanted to build right we found the things that were like useful that would precipitate the thing at the end so okay if we master how we're going to store this data here if we master how we do analysis. If we master how we deploy the tests like etc. Right? We can precipitate this amazing software as opposed to trying to blindly without any context like roadmap it out like ahead of time and that sort of thing. And so uh it's really fast and functional and has worked really really well at scale because I think we've taken this like non-traditional approach to building it. >> That's like a first principle kind of thing, right? >> That's right. My background originally, right? engineering to figure out a better way. I mean, that's how SpaceX came about, you know, building the rockets is they applied first principles >> engineering build stripping it down to its parts and like how can we put this back together in a more efficient, better way >> rather than how looking at the whole and like, well, how can we make this rocket faster over here? It's like, no, let's take it all apart and put it back together. And that that sounds like it's similar to that. >> I'm so glad you didn't say this to me, Kev, because I would have said Mr. Campbell. He was my first principal. >> He was your first track. >> Shout out to Mr. Stump over here. Yeah. No, >> but those those things those things work >> and I don't know if this is what your tool does. We'll talk about that, but this works in AI and it just changes everything. Um, and so it's fascinating that from an exercise or parkour to business to first principles to rockets to whatever, this is where you get an edge ahead that most people don't understand. >> Yeah, that you're completely right, Kevin. So, as you as you mentioned, my my background originally was structural engineering uh and I was getting a master's degree in uh uh mechanical engineering for seismic design. So, I was doing like earthquake engineering research and I started working in e-commerce for the small apparel company and as I started going through everything uh you know I was like okay cool so like we care about customer lifetime profit like against like the cost to acquire their attention and someone's like what and I was like you know like you want to sum up all of the gross profit that the customer like produces and like that's the number you use right and they're like no we just take like all the revenue from like the first sale and divide it by all the marketing cost and then if it's greater than two then we ship it. And I was like that sounds dumb. Uh so uh as a as a mechanical engineer I was like that's that's so gross. Like you have all the data like right here like you could just be calculating like profitability like do you count returns in that? No no. Uh okay like discounts sale like margin any of it? No. Uh so I was like well this is dumb. I'm just going to calculate it myself. So, uh, as I, uh, started working in e-commerce and then eventually started doing conversion rate optimization, we were just like handcalking the stuff all the time. I was like, I've taken like four statistics courses like this is not hard, right? But we would take it and look at like, okay, so like the best customers, right? The ones that correlate with like high lifetime profit. They came in, they you bought from this product category or they did or did not use this discount or like all these like kind of leading indicator things. And we started reporting on like testing from the perspective of qualifying like what these like good or bad customers were, right? Because I'd come in to like work with these businesses and they'd be like, "Oh, it's amazing. Our lifetime value is, you know, $100." Uh, and you know, our cost to acquire the customer is only like 50 bucks. And so we're kind of clapping along. And I would go into it, I start like teasing out these like customer segments. And I'd be like, "No, it's not." I was like, you're when you factor in cost to acquire, you have a cohort of customers that you lose $10 on every single time and a cohort of customers that you're making $250 on. Stop acquiring the ones for negative $10 and the business will make more money. Like you could I could murder your conversion rate and just ship profit uh and then put a backfed signal back into the meta algorithm to for when it's looking at what a customer is. I'm just going to artificially cancel all the ones that lose $10. Uh, and that's like the approach that we really started taking with like I hate the idea like the the the nomclature of like conversion rate optimization because I think the way that most people do it is like objectively bad. Uh, but uh so it's like but you can't say what do you do? I do customer lifetime profit optimization. Uh people like I have no idea what that is. So uh all the existing software tools out there like they suck. uh because they're they're just they're taking like a tertiary slice of like that perspective where they're either looking at just conversion rate or they're making like here's a wizzywig editor, right, for like which is cool if your business does like no revenue uh and you just need to like ship it yourself, but it's like especially when you like coming from like apparel or like footwear and like return rates are like 20% sometimes. You really really care if you're acquiring customers that just come in, take free shipping, like return the the the product, etc. and there aren't really great tools for that. So, we had to build one. >> Hey, Norm, I've got a quick question for you. I'm trying to manage all my affiliate and creator programs from Amazon, from Shopify, from Walmart, but it's just a freaking mess. I mean, I've got reporting coming from here and there. There's all these different Slack messages. Do you know if there's like a unified dashboard where I can do this all in one place? >> Yeah, absolutely. And you're right, it is a mess. A lot of brands are complaining about that, but there is a place that has a solution. It's called Lavant, and they let brands recruit partners, track performance, manage payouts, send product samples, and even run creator programs across every major marketplace all in one place. And guess what? Brands can spend less time on tools and more time making profit. Is that the one that you sent me a link for? Like a 10% off coupon, the gold or enterprise plan a few few days ago. >> You got it. >> Oh, cool, man. I think I've got that link here. Was it lavanta.io/misfits? >> Yep, you got it. >> L Va.iomisfits. >> Awesome. I'm going to go uh go hit him up right now and get that 10% off. >> Perfect. Me, too. Well, they say they say that like a lot I remember publisher remember publishers inquiry where Ed McMahon would show up at someone's door and give them a million dollars uh during the Super Bowl and everybody those there's magazine subscription so they'd send out something there direct mail back in the day and you'd subscribe to different magazines and that was your entry but and then the magazines loved it because they got a ton of subscribers but then they realized none of these subscribers are actually buying any of the advertiser stuff that advertises the magazines and none of them would actually The original goal was like, well, they're going to renew, you know, when they're they get this for a year for three bucks or whatever it is. They're they're gonna renew for $29. And very few of them would ever do that. The only time they would renew is when the the next year when the flyer came in the mail to get it for $3 again. >> And and the same thing happens with like uh what's what's it? Uh Groupon. You know, when Groupon first came out, everybody's raving about Groupon and I'm going to have them I'm going to give away discounts and give get people in to get cheap haircuts or whatever it is. And most of those people never came back >> and they had massive sales and like yeah we were flooded out the door and it was awesome but it turns into nothing. So I how and now with the internet you can track this stuff like with precision. Um but you still get confusion >> on attribution. Uh, and so like it correct. So I mean there's tools like triple whale and some of these that can kind of follow it across Facebook but it might and I've seen people that you know we had someone on the podcast John Moran one of the top face Google guys and he he was like yeah you know sometimes I'll see people turn off their Google ads and because they're like there's no I'm looking here at what Google's telling me. He's like, "Yeah, Google is showing you what they want to show you to keep you advertising, but you're saying you're not getting sales, so you cut it off cuz Facebook's got all the sales, but then all of a sudden, Facebook dies." >> Yes. Yes. >> How do you how do you follow that? Is that what your tool in part? I'm assuming that's in part what it does. >> Yeah. So, it's a great question. So I think of it right in terms of like a loop because if you're just doing analytics, you're a historian. Uh that doesn't tell you, it tells you what happened. It doesn't tell you what will happen when you change something. So like to to your point about the group on thing, right? It's like or or any any person that is in e-commerce or whatever, right? Like oh when we acquire a customer, their lifetime profit is this. Okay. Well, cool. Try acquiring more. Do 10 times as many. Tell me if that number stays the same. Right? We know it doesn't. Uh and so the question is right is is teasing out correlation and causation. And so again like analytics without testing uh just gives you clues about how to form hypothesis. Uh it also doesn't tell you actually like how to do any of it, right? So like in terms of like what action should we be taking in terms of like development or whatever. What we built is software that actually completes like the whole circle. So it plugs into your eventbased data from like a Google Analytics or something like that, right? Which is like helpful on turning out what people are doing on the website. Then it also pulls in your Shopify data, right? But it's pulling in all the different attributes of the products so that you can go and like segment them by different product types or categories. You can pull all of the shipping profiles like etc. Um the uh subscription uh cadences, all of that fun stuff. Uh and it pulls in all the customer data, right? so that we can see this person purchased within 30 days they rep purchase. One of my favorite metrics that we in uh in invented we just make up right because a metric is is just a number and it usually has a numerator and a denominator. So, I hate the idea of like conversion rates because what's a purchase and what's a what's a customer? What's a session, right? Like we have to define these things. Like there's no there's no agreed upon definition of that. Is bot traffic count in your in your denominator? Probably not. How are you going to subtract that out in your numerator? If the customer never comes back and they purchase like an intro product that has a subscription, but they're negative lifetime margin on the first order. Should we even count the first orders? Maybe your conversion rate should be all the customers who placed at least two orders divided by all of the sessions that weren't bot traffic. How how are you going to calculate that? Uh so we built software that literally just does it because in order to run a healthy business, you actually have to understand how the business generates profit and then actually be optimizing for those numbers. Is that conversion rate? I don't really care what you call, right? It's just a label. Language is necessarily figurative. Uh so we're just making up definitions for all the all the words actually. So uh then we pull in the ad level data right so from meta from Google etc. So we can contextualize all the different touch points that we have. We can do our own sort of like loose version of attribution. We can look at okay if we gave first touch credit, last touch credit, all any touch credit etc. Where do these things fall? And that tells us more about like is to your point is Google actually like very good top of funnel and most of our meta people actually in in you know interact with Google at some point etc. We can look for some of that overlap, right? Uh then we also pull in all of their Claio data. So we can look at when somebody signs up for email, etc. Uh people have these giant Clavio lists. They're not all all those customers are not equal, right? You have some people that you have their email address cuz they bought something. You have some people that that uh you got their email address and they purchased the same day. So whatever your your promo is, like when they came into the website, that was the thing that like converted them right there. And then you have a third category, which is people that you nurtured, right? And then you have to segment those things out when you're doing things like running a test for what should our intro offer be or how should we do a pop-up or whatever, right? Just looking at conversion rate is stupid. Even just looking at average order value is kind of stupid because you're giving them a discount, right? You want to look at profit. You want to look at repeat purchase rate. Are you flooding your email account with a bunch of garbage email addresses that never nurture or are they absolute gold? Like we've had some uh brands that have used our software and have proven that like their emails are worth like 30 bucks like an email that they get in. That massively changes your strategy, right? So our software does the analysis portion. It also does all the testing portion so that you can deploy tests basically right out of the software and have it come back and and use the same metrics that you were measuring things like customer lifetime profit as your metrics that you're actually optimizing for. Uh, and then it has another component which is it does actually it actually does do some of the ID resolution stuff because I I don't like the way a lot of these other softwares do it. They either do it really poorly uh or just to favor whatever the thing they're selling is, whether it's ads or emails or whatever the thing is, uh, or they just do gray hat stuff that's like wrong half the time and probably going to be illegal in like 6 months. So, uh, the way that we approach it is we have like a firstparty app that sits in your store. It observes all of those things that are happening first party. Oh, they opened an email, whatever, right? The IDs are available there. They came in through Google Analytics, they clicked an ad, etc. We assign a global ID like many do, etc. And then we we also fire those events for our tracking purposes for testing, right? Because again, not just historians, uh, into like a Clavio. oftentimes we can match customer profiles for things like add to cart events like 30% better than like a triple whale or retention.com or whatever the thing is. It's really it's really interesting but that's a hard problem to solve right again not taking like the tertiary approach of like I'm doing event-based tracking ah we're deploying like wizzywig tests ah we're doing like just like marketing attribution etc. I think you we're reaching a point uh at in in the world of e-commerce with regard to like regulations of like harvesting like thirdparty data where like you really do have to have like a first party view of like how the business generates profit and the and the ways that you capture like that attention of the customers if you're going to win the game. Wasn't it Tesla that had when they first started selling cars online, it wasn't working because there was like I may have the numbers wrong here, but something like 74 different way 74 different options to ways to do your Tesla and they just weren't converting. And so they they took all the data kind of like what you said probably. I don't know if they were using your tool or not, but they they whatever they did, they and they figured out that there's really only three things that people want on a Tesla. And some of these we can combine in the same choice or whatever. So they combined it all down to >> I may have the numbers wrong but three choices >> is very minimal. Three choices and relaunched the website and sales went through the roof. >> So >> yeah. No, I mean this it's so fascinating, right? Like uh there's a there's a brand that's using our tool right now and uh they're going through a website redesign and they have one of these like navigations where uh on the website where like when nobody could make a decision they just put everything in. So, it's just like this like awful like list of 60 things. They used our tool basically to look at where do customers go on the site that they don't land uh that correlates strongly with repeat purchase behavior and high lifetime profit. That that was the question, right? So, uh it turns out writing SQL is really boring and most people don't know how to do it. Uh, our tool goes and like pulls all of that data that I mentioned before into like a data warehouse and then like preps all of it and like sanitizes it and sets it up so it can be easily joined together so that you can write SQL against it like trivially essentially and we we materialize that data once an hour. Uh, and so you can use a prompt right where it turns out having LLMs actually calculate things is the most dangerous thing you could possibly imagine doing. having it write SQL to calculate things in an existing tool or write JSON to fill in forms like for building out like an analysis etc. trivial uh they're translation machines, right? They literally just take language and turn it into another language. So they take English and they turn it into JSON or they take English and turn it into SQL. So that's all we let the LLM do in our tool is it does you could do anything that the LLM does in our tool manually. It just sucks. So uh no one wants to write the SQL or fill in the fil forms. So through MCP. >> They did an analysis >> like uh it's just we just have the conversation. >> Yeah. >> In your tool. >> We we so we have the conversation prompt directly in our tool and it has access to every uh every input and output field that a normal user has and it'll call those uh those skills or actions in directly and then it lets you go through and actually like review those metrics or segments that it sets up for the analyses uh etc. So what they did they built a report basically looking at okay customers who came into the website what collections etc did they go to that they didn't land on uh that correlated highly with like repeat purchase behavior etc. Now, here's the thing. Job of a historian does not tell us what what we should do uh if we change things, right? So, what that did is it led to a hypothesis of we should promote these product categories, we should demote these ones or remove them entirely off the site because they look like they're dead ends or people just come and buy discounted products like etc. And they administered a test uh in their navigation that uh paired that list of 60 down to 20. and it wasn't the 20 that they thought. Uh, and there was like a reasonable amount of complaining uh, when they went through it of like, nah, like I don't know if we really ought to put this on there. So, they ran the test. Uh, increased conversion rate by like 3%. Fine. Uh, increased average order value by like 10%. Uh, increased lifetime like the essentially what is like the lifetime profit by like 15% because it was driving people away from the sale products. And so this stuff gets really really exciting because you can you can basically set your website up using a tool that can complete the circle like that to be literally optimized for customer lifetime profit to choose the customers who will come back and purchase again who elicit a good behavior like in their first order and all of that fun stuff. Uh and you can do it on purpose. So when a business is working with you, so I went to your website and you're dealing with all these huge companies, but what about the small medium-sized company and how often does the owner or founder have to get involved? Because you're talking a language where they would have no clue for the most part. >> Yeah. So the thing is right is it's it's when you you know when you are at a smaller scale right the the your goals are different than when you're at the larger scale. When you're at the larger scale you have like all of this like uh fine grain data where you're trying to like uh you know get more juice for the squeeze. The thing that I think is really interesting with the with the smaller brands is a lot of times the mistakes that we see people make is they will get something that gets traction on paid media and it's usually some sort of like intro offer into like their their product that ends up like poisoning the well. Uh, and so the thing is is that if you don't have a really tight feedback loop on observing what those customers are doing that are specifically coming through those ads that you're like trying out and trying to figure out which ones you can scale, you can get upside down really quickly by getting a customer base that behaves poorly. So, uh, for example, there's a small brand that we work with, uh, that was using our software tool, uh, because they were running meta ads to a toothpaste brand, running it to anti-avity, uh, and they were like, "Oh man, this is sick." Like, we have this, uh, great anti-avity ad or whatever, like it converts really, really well. We're like going to increase spend on it, etc. Now when you do uh a analysis on customer lifetime profit and plot it like a histogram. So for everybody playing at home that didn't take force statistics courses as a histogram is where you have buckets of like they lost you $10, they were neutral, they made you $10, they made you $20, etc. And you see how many customers fall like essentially into each one of those buckets. Usually what people will do is they average it together and they pretend like it's a bell curve on with one with one hump right in the middle and it's nice and concentrated. The thing that we tend to find sometimes with things that like get good traction like that, right, is it's actually looks like camel humps where you've got a whole bunch of customers that suck and then a whole bunch of customers that are amazing. And like uh consumer package goods are like this a lot where it's like you're not going to stop brushing your teeth probably so you're just going to keep buying toothpaste. So, you're basically either going to be like a oneanddone person and you're going to lose me money, right? Because toothpaste isn't uh particularly expensive or you're going to keep brushing your teeth with that toothpaste for like the next 5 years uh and you're going to be the best customer that's ever happened, right? So, the thing that we found is there was like a uh they using the software they were able to like see right even at the small scale that the customers just fall in these two danged buckets. So they're able to like basically form some hypotheses around like okay like what are the patterns of like a oneanddone person or whatever right because again they want to scale this but if you scale that stuff too early meta is just going to like just home in on these easy to convert really lowquality customers and scale it through the roof and you've already burnt six months of budget before you figure out that none of the people are coming back and buying anything right and you're sunk. So I think the the game when you're at the smaller scale is to figure out how to evaluate product market fit in like the macro scale like to scale like responsibly if that makes it I maybe I answer your question about the the the owner founder >> smallmediumsiz business and they go to your site and how do they even get started? Is it just a plugandplay or you need information? >> Yeah. So the thing is is that uh we've you know the way that we built this software successfully is partnering with businesses like as they were like going through that scale and basically like shephering them through that process uh of uh taking the things that are important to them and then figuring out how to like triage those into like an actionable testable environment so that they can like essentially like again like learn those like first principles of how to responsibly gener generate profit at scale and then kind of scale it up. So, uh I think it it helps to have it helps to have uh like an a a good uh operator that understands these things can like utilize that. We've seen the whole spectrum, right? We have some folks that are just like absolutely clueless but really happy to be here. uh and they've gotten good they've gotten good product market fit uh in spite of themselves and uh and need need all the help kind of coaching and adopting that kind of mindset and then other people that are like oh my gosh where were you 5 years ago like uh okay thanks we'll take it from here right and we go in and and uh chat with some of those folks and it's like everybody from like the person that used to do data entry that doesn't do it anymore all the way up to like the private equity that's even involved in the business is all using the tool basically running all these queries like etc. So definitely varying levels of sophistication but you're you're you alluded a little bit to like is it is it helpful to have some help here. Now we have we have done a thing which I think is like uh quite helpful which is we have a thing we call playbooks uh which is there are these analyses that are universally helpful especially to a business that's just trying to figure this out for the first time and it's things that we've like battle tested against like a billion dollars worth of like collective revenue where it'll be like hey run this playbook it'll build all these analyses comb through your data sanitize it ask you a couple of questions and then print out okay this is this is your buckets of email qualification this is how much an email is probably worth uh that kind of stuff. I think that's tremendously helpful. Um same thing with like setting uh tiers for free shipping, right? Where it's like a couple of little inputs of like what do you actually pay for shipping? It can already see what you charge each one of the customers and all the different levels and then you can do an analysis again looking at those orders and then set a like okay cool. It's here today. if you move it by this many dollars, it'll move it'll capture 15% of people this way or that way. And it kind of like tease up the test for you uh in terms of like doing those analyses because it turns out there's only, you know, there's a finite number of interesting shapes uh of of data in e-commerce. Now, where the rubber meets the road of actually applying it uh is easier said than done, but getting people started to like kind of build those thoughts um we found the playbooks thing is really really helpful. Hey, Norm, do you know any sellers out there that are just burned out during this uh ecom game? >> You know, I I know a lot of people that have talked to us, you know, when we go to events, and it's not only that, they don't know where to start. >> And who would you recommend they talk to? >> The first one that comes to mind is is Quiet Light Brokerage. And here's why. They're going to build you up. They're going to understand your company. And at the end of the day, you're going to know how to maximize your valuation. So, the very first thing you need to do is go and get your free confidential uh valuation at quietite.com. They're going to ask a couple questions. Uh you're going to meet up. It's one-on-one with uh somebody over there and then, you know, let the games begin. >> Awesome. What What was that website again? >> It's quiet.com. >> Awesome. I'm going to head over there. >> So these smaller guys, I mean a lot of them they're they're they're thought of testing is AB testing. Let's test A against >> Yeah, they have no B1. So let's go with B. And maybe they make B the more sophisticated ones make B the control and then keep testing against that and that's that's all they can do. But now with with AI and especially I mean you have it built in it sounds like to your to your your system but with AI and MCP Norman and I have a a rule in Dragonfish that we don't touch software that doesn't have an MCP because that way I if it doesn't have an MCP I'm not interested. I'm not going to go by your dashboard. I want your data and what you've been assembling, but I want to ask it my questions my way and be able to tie them to different data sources because if you know how to do that analytics and that tying together, >> it it's like freaking a magic potion. Uh, and a lot of people, a lot of business that's beyond the pay grade of most business people and most people using AI, which baffles my mind. Um but >> well and and you and and when you get into testing you get into really weird stuff too because uh LLMs don't really have a lot of discipline on like committing certain uh statistical fallacies. And so uh the thing that we see especially at like the smaller scale is people abusing statistics not on purpose. So you run into things like the peaking problems like Simpsons paradox multiple comparisons problem etc. just like ways that you can use things like an LLM which can like pull massive amounts of data really really quickly and then expose you to tremendous amounts of noise uh is is like usually kind of like the flip side of that. So we've built a lot of guard rails into our tool uh one two uh you'll ask it to do something that is like not a good idea uh like hey I want to look at 30 different metrics for this test. You may not look at 30 different metrics for this test. Uh that would be committing to the multiple comparisons problem. Your sensitivity to noise now, right? There's a have you you guys you know XKC? Do you know this web comic? >> No. >> Oh, you're going to love this. Okay, so it's a it's just a it's a nerdy web comic uh that just has a bunch of like physics and statistics jokes and stuff in it. And there's this one where it's uh it's a bunch of panels of these people like study like uh researchers doing a study of every different color of jelly bean. Uh, and it's like, uh, oh, there's a, you know, P95 red jelly beans don't cause acne, yellow jelly beans, and it's like 20 different, uh, like colors, right? So, the joke is, right, is when you, uh, when there's a 5% chance of, uh, error and you test 20 different uh, colors of jelly beans and you look at every single one of them, your expected value is that one of them is going to be noise. And so they ship this like paper and then it's a headline of a newspaper at the bottom and it says this just in green jelly beans found to cause acne. Uh and people >> Yes. And and it's and this is one thing I see with AI AI and like LLMs is uh they they can't they're translation machines, right? They cannot simulate. They cannot mentalize which is fine but you have to know what you're dealing with because they mimic human beings and language right so they they you people very commonly perceive them as being something that has a worldview or has a goal which they do not because LLMs can't do continuous learning in the nonlinear case right and so what ends up happening is you can ask an LLM and it'll be more than happy to calculate anything you want uh and just let you like moonwalk past abusing it unless it has like good guard rails on it. And so I think that's one of like the flip sides of this, which is like have access to literally all of the data, calculate all of the things, but when you're making statistical choices, having like some underlying like rigor there that doesn't have you just like chasing infinite noise because you can run really fast, I think is like part of the game. Uh it is really really funny though, right? Because it's just a it's we thought we we we we wanted to build uh androids, right? We wanted to build Androids with AI. What we really built was mech suits. So it's like whatever you normally do, you now can do it 10 times as much. So if you are just like an absolute loose cannon and have like no discipline in like the the general like work that you do now you can do it 10 times better. It's like I meet these people that are like love cloud code. I use cloud code all the time. Like please stop, right? Because it's just they're just doing terrible work 10 times faster. Uh and just generating like fake problems and like complexity that doesn't need to exist. If you're a five on the knowledge scale, AI will make you a 10, but if you're a 10, it will make you a five or or something. It'll it'll it'll take you back. Um, so so it's I mean that's one of the things like you what you're talking about there, Norman and I, we have a company called Dragonfish that does email marketing for ecom brands and we one of them we do warm email and we do cold email. >> And what we've been building um is it with the five beta clients is a a brain system. So we have 11 different brains and they're the guardrails. So instead of telling Chad Claude saying, "Here, go write a go write a cold email sequence uh for company XY and Z. Make sure you do this, this, and this." No, it's like reference these 11 documents that were created independently and then cross-cheed against each other to actually write this. And it's super powerful. >> Uh and you don't get the same old slop and you don't get the same old stuff that everybody does. And >> we have it on competitors. We have it on their keywords. We have it on their brand statements and what they stand for. And we have them do a first principles document themselves that we then incorporate into this. And it it's it's it's a pretty cool system. I think um that really differentiates what comes out of it. >> And we I think we'll be talking to Matthew after this call, by the way. >> Yeah. And we I think so. And when we've we're in beta now, so we've been building it's almost done. and and being deployed. But we can with a touch of a button, we can create any kind of landing page. We can create any kind of image. We can create any kind of quiz funnel. We can create any kind of optimiz. It has all the optimization stuff. I mean, it it's a pretty pretty comprehensive and it's not perfect. It can be still made I know a lot better. Um but but that's what it's that's what baffles me is people I don't know. They they don't think it through. uh they just want the easiest way and the most general way. I I gave a talk on this recently and the the so I was uh you know we've we've done this lots of times historically right like if we we say like hey what's like the the closest like parallel to what we've created with LLMs it's essentially the assembly line right because what it allows us to do is take something that is essentially like unsk like unskilled labor and produce skilled results right and it's it's only as good as the procedure Right? And so the thing is is that procedure is cool because it moves everybody to an average. Now when you have people who are like uh exceptional uh with regard to like their ability to problem solve like etc. they all hate process because it brings them down to like the average. But when you're when you are setting up like an assembly line right you used to have to say I want to make shoes. I need to have be go apprentice like a a shoe maker for for two years to learn how to make boots really good, right? But instead, we built the assembly line and it was like, "Hey buddy, your job is to put the the the rivets in the sole." Uh, and then you punch for the laces and then you attach the, you know, and so it's each of the individual parts. And so the person only has to have knowledge about how each one of their their little pieces works and they don't have to have a concept about how to do all the skilled labor. And I think when you have something like that sounds like what you've built, right, you talk about how to use like an LLM like successfully, you have to break the context window down to punch a hole in the leather right there and it'll do it an infinite number of times really really well, right? And when you can chain that stuff together, you can get really, really good results. But if you walk up to the assembly line worker and you say, "Hey, we're thinking about what we're going to do for marketing for new shoes for 2028. Uh, do you have any ideas?" That would be a stupid question to ask somebody who punches the rivets into the soles of the shoe, right? But uh going to an LLM and saying, "Here's my marketing data. What should I do uh for our marketing calendar right now?" is equally as stupid, right? Because an LLM cannot cannot mentalize. It doesn't have a goal. It doesn't have a world view. It can't hold it can't anticipate the thoughts of something else. It's guessing what the next damn word is, right? And that is a phenomenally powerful thing that will create a trillion dollars worth of value. Please don't ask it things that are just like intentbased, right? There is there is a whole world of action out there that can be put onto the LLM assembly line and you're asking it to make like critical business decisions. So like if you go into our tool right now and you say like, "Hey, based on the data, what should I do?" It will say, "Hi, I'm an LLM. Uh, I am not capable of actually having any goals or mentalizing. I'm just going to give you an answer that looks like what a reasonable answer looks like. Here is all of the data. you should consult somebody who can actually interpret it. Uh, and I think that that's really really important. >> Yeah. Someone gave me analogy of photo Adobe Photoshop. They're like, "Hey, have you ever used Adobe Photoshop?" Oh, yeah. Of course. I've gone in and recropped a picture, resized something. Well, have you ever actually designed a piece of art in Adobe Photoshop? I'm like, uh, no. I I'm not a graphic designer. I'm not I don't think in that way or don't know how to use the tools in that way. It's like, well, that's the difference in AI. Most people go to AI and they use it like they use Photoshop. just crop an image or you know make it from a PNG to a JPEG or whatever but the the graphic designer is the one that knows how to use the machine and actually had has that background and that's the difference between an AI operators >> yes >> is something similar along those lines uh and that's what's differentiate or AI systems that are built by AI operators that incorporate that and that's that's the big difference and I hear all these people anti-AI or AI slop and like yeah there's a ton of it but the power of if you know how to use this tool is ridiculous. >> Yes. It's it's the game of separating intent and action, right? Intent is all of the the judgment type things that requires like the expertise, a worldview, an understanding of how other minds work etc. The action is just everything that flows after that, right? And so the that intentful portion in terms of like the entire time needed to accomplish any goal is very very small. And so in the same way that like if you are a good manager or you delegate well etc you you are kind of obsessed with separating that action and intent and really only having the intent and your job is to express the intent as efficiently as possible and have someone else carry out the action. You really have to treat LLMs like that to be successful as well. >> Sounds like this could be a good dating site. >> Yeah. >> Yep. >> Yeah. Uh yeah, it could it could be. >> We'll call Finder. >> Finder. >> You did something with Allirds. I mean, you've worked with a lot of big brands. Red Bull, all birds. I mean, your website has a laundry list of well-known brands. >> But you talking earlier, you touched on uh the shoes. You know, if you're selling shoes, you need to know the returns. You know, this and this and the other. So, you don't want these type of customers. Can you did an experiment with them where you like split hairs basically, I think, um on on something. Can you talk a little bit about that and what you learned off of that when it comes to conversion and like the value of the customers? >> Yeah. So, so like really really interesting stuff with you know with shoes uh for example, right? Like like that return behavior. Uh so uh you can incentivize customers into bad behavior. So, a lot of times what we'll what we'll see is uh folks will say like, "Oh, cool. Hey, we want to do this like intro offer for customers. It's this big of a discount on the first order or whatever, right? It'll have free shipping and returns, like etc. And these have knock-on effects for essentially like optimizing for tire kickers." Uh, and so you'll get folks that will be like very opinionated about like, "Oh, we have to collect all these emails, right? Like we do a lot of like email marketing, like etc." Or, uh, "Hey, this is a really good deal. like it converts well for us. Our customer lifetime value is X. And when you're applying things in broad strokes, all those things look like really really true. Uh but you can do things where you can say something like, hey, free shipping on two or more pairs, right? And that can all of a sudden like really kind of like change things where it's like the split on the hypothesis is like, okay, is that going to incentivize people to just buy a pair that they like and a pair that they're not sure and then just pick between the two? Are they going to be like on the fence on sizes, etc. or does it like weed out the people who are just tire kicking etc. We ran a test for a shoe brand that we work with. Uh and the really really interesting thing was we actually changed that instead of giving a a uh disc or basically doing free shipping on everything and giving like a discount on the first orders. Instead what we said is like hey free shipping and returns on two pairs of shoes and then we look to see based on how we were acquiring customers what the net result was. Now, the really interesting thing was there actually a non-trivial number of people who would come in, look for at a pair of shoes, see that it was free shipping and returns, place an order, and then just return the shoes. Like, I don't know this. I don't not for me, whatever the thing is, right? Uh the thing that we found though is that among the folks that were purchasing two pairs of shoes or more, they weren't doing it for uh the reasons that you might imagine, right, which is like, oh, am I 11 and a half or a 12? I guess I'll just buy both and like send it back. like once they really got like dialed in, they kind of knew what size shoe they were. They weren't buying different styles. They were legitimately keeping different shoes or buying one for their wife and one for themselves, like etc. And so the really really interesting thing on that was repeat purchase rate went up phenomenally. Return rate went down a ton. First order profitability was much much better, right? Because average order value is going up. You're not using a discount to uh convert those customers. And so you can like really change the unit economics of a business when returns are going down, repeat purchase rate is going up, first order profitability is going up, your uh you know your gross margin after like discounts and stuff is like going down on those first orders. And it can very rapidly change the ecosystem that you're you're purchasing customers in like as a result because sends a much stronger signal back to like a meta uh in terms of who you're allowed to acquire and if they're if you can with confidence acquire those customers even at a higher price point because you know you're just getting the good ones completely changes the business. I mean, an interesting just a quick little story on that is I have a newsletter in the Amazon space called Billion Dollar Sellers and I use Beehive to send this out and they just opened up an MCP. So, I went into and and I did some queries in the MCP to pull some data sets of different data and based on leads. I've been paying for leads for from this one one place and some of these are business business email leads and some of these are personal emails. So, I'm seeing like which is performing better. The business emails got me 15 much higher conversions, higher open rates, and decent click rates. The personal emails got me a much higher unsubscribe rate, a much lower open rate, but a way significantly higher engagement rate. >> So, it could be misleading if you don't look at the data all the way through. You would say, I want the business email addresses only because it's higher open rate and good conversion rate because these other guys are subscribing and much higher like three times as high. But the engagement rate is seven times as high. So they're actually smaller and better customers. And now I'm like, "Okay, I need to ramp that up. I'm still going to do the business, but that was counterintuitive." And their tool would have never been able to tell me that, but I was able to do that with an MCP and a proper prompt using their data. >> Yeah. No, we we see this all the time with uh like pop-up providers that do like uh SMS and email signups, etc. all of their like internal AB testing tools are just crap, right? Because all they optimize for is getting more emails or more uh SMS numbers because they make money on the sends. So like the incentives are totally misaligned. So like what we'll actually do oftent times is people will use our software where they'll deploy like the the AB test like within a Clavio for example, right? And uh our software actually has like a little script that goes and like pulls the ID numbers uh when those pop-ups like appear in the DOM uh on the site. >> And then we can look at real cohorts and look beyond what they're measuring, which is they're like, "Oh, people open the popup. They submitted the popup. You got the email. Here's all the conversions from the emails that were submitted. That is a stupid number to measure. Who cares? I want to know how many people actually bought things on the website. if they didn't need the popup and they didn't take the discount and bought it, that's better for me than if they gave me their stupid email address, right? And so we do these like kind of like macro things where we can look and say, "Hey, you acquired 10% fewer email addresses, but you converted 3% more people directly on the site exactly in that session. And we can break those email cohorts out and like look at the quality of exactly what you're saying, Kevin, which is that like people will be like, "Oh, this is amazing. So many people interact with my popup. I get all these email addresses. look at all this revenue I'm generating from the email. But when you look at the revenue per email address, it's much much lower. And they murder the conversion rate on their own website just to have to like wait to try to nurture people a week later. It's phenomenally stupid. Hey, Kevin King and Norm Ferrar here. If you've been enjoying this episode of Marketing Misfits, thanks for listening this far. Continue listening. We got some more valuable stuff coming up. Be sure to hit that subscribe button if you're listening to this on your favorite podcast player or if you're watching this on YouTube or Spotify. Make sure you subscribe to our channel because you don't want to miss a single episode of The Marketing Misfits. Have you subscribed yet, Norm? >> Well, this is an old guy alert. Should I subscribe to my own podcast? >> Yeah, but what if you forget to show up one time and it's just me on here? You're not going to know what I say. >> I'll I'll buy you a beard and you can sit in my chair, too. We'll just You can go back and forth with one another. Yikes. But that being said, don't forget to subscribe, share it. Oh, and if you really like this content, somewhere up there, there's a banner. Click on it, and you'll go to another episode of The Marketing Misfits. >> Make sure you don't miss a single episode because you don't want to be like Norm. So, we've heard you talk about this incredible software, but we've never mentioned >> Oh, yes. >> What is it? >> Okay. So, uh here. So, >> he's got to run up the wall first, Norm. He's got to run up a wall. >> Yeah. Hold on. I'm just going to backwards over this cash. Uh so uh so uh Proteius Digital Lab we were we internally as we were building this uh we kept saying like oh we should name the software what's its name? Uh and I was like oh no no no we're not naming it. And someone's like why not? And I was like ah because language is necessarily figurative. Once you name something it tells you what it does. We don't know exactly what this does yet because we want to do the thing that generates the most like value. And people are like that's the most pretentious crap I've ever heard. What's it called? And I was like it's called the tool. Uh, and they're like, "That's the stupidest name I've ever heard." Ha, got them. All the people that use it now refer to it as the tool. And I was like, "Oh crap, now we have to call it the tool." So, uh, we we have a tool uh that does uh business intelligence and testing and uh and customer identity resolution. Uh, and it's called the tool uh by Produce Digital Lab. And uh we are uh now actually opening up uh allowing folks to uh to to to use the software in more broad uh context. We did it with a a a group of those brands that you see like on our our website uh essentially as we were like piloting it with them and kind of building it over the the years and doing those services with them etc. Uh but it now exists um by the time some you're probably listening to this dear listener uh in a standalone uh capacity. So you can take >> referral only available by referral only. >> By referral only. Yes. Yes. Uh so uh if you're uh this is the best bit about reading really really good is you just only have to do stuff you like. Uh someone was like how do people work with you? And I was like somebody introduces me. I don't have to deal with people I don't like. who are you talking? >> But uh if you uh if you are listening to said podcast, you come to the the website, uh there'll be a form there you for you to uh fill out and we'd love for you to to take a spin and uh let us know how it is. >> All right. Fantastic. >> And that's just to spell that for those listening. That's P R O T EU S P R O Tus Digital.digital.com. >> All right. Okay, Matthew, thank you for coming on. This has been fantastic. >> Thank you guys. It was a joy. >> It's a lot of fun. You can tell you got a little bit of passion, >> casual. >> Hey, I got one question for you. I'd like to ask this to all of our guests or our misfits. Do you happen to know a misfit? >> I do. I think that you would uh really enjoy talking to my friend Taylor. Uh so, we'll see if we can get him on the show. >> Fantastic. All right, sir. You have a good one. >> This has been great. Thanks for coming on. Cheers, guys. Appreciate it. >> All right. >> Cool stuff, man. That's probably over the pay grade of a few people, but uh but but it's uh that's really cool stuff and it's really fascinating what you can do and where this is going and how optimization has gone from just little split testing to now this very scientific engineered approach. Um uh it's awesome. Awesome. I love I I love I like I love getting my head into that that kind of stuff. >> I I got You know what I got out of that? A new name for you. >> A new name? >> Yeah. >> Uh oh. A new name. What's the new name? >> The tool. >> The tool. I think you mean that differently though. >> I think you mean that not in a positive way and in a negative way. I'll print your badge at the marketing misfits. It'll be the tool >> in market masters. It'll be the tool. >> Yeah. Oh, that's right. Market master market. >> Oh, we can check. Yeah. On the misfits, it'll say Norman and the tool. >> No, it won't. >> All right. So, Kev, how do people see this? >> Oh, well. Well, if you want to see this one and about 10 and something more, we come out every single Tuesday since 2024. Hey, look, I just made a rhyme. Um, since uh since April 2024, every single Tuesday, Norman had been here where your buddies in the car, on the subway, on your run, uh, on whatever you may be doing. Uh, new episodes come out at marketingmisfits. Is it It's doco still, right? It's not.com. >> It'sco.co.co. >> Marketingmisfits.co. Uh th those are also posted on the on the uh the video sites. What what's that? The the tubers and the >> those tubers on YouTube. We're marketing Misfits podcast. And if you just like the the short 3 minutes and under uh clips, we have Marketing Misfits clips. And we also have a newsletter. I think it's uh what two months old maybe. >> Uh yeah, about about two month about three months old now. But um misfits.news is where you go to get that every Wednesday. a brand new edition of the email newsletter comes out. People are saying uh it's pretty good. I I I think so. Um myself biased a >> little bias there, but a little lots of actionable stuff from the podcast. Uh a little stuff about Norm and I. Norm has a cigar whiskey suggestion or something in there. So uh it's uh check it out at misfits.news and then we'll be back here again next Tuesday. Right, Norm? >> That's it. >> All right, we'll see everybody then. >> I'll see you the tool.