Episode Transcript
Kyle James (00:01.122)
Hey, welcome to the AI Chronicles podcast. I'm your host, Kyle James. Today we're going be diving in headfirst to how a marketing research company called Tether Insights is using AI inside of their own business. And we'll share the exact steps that you can take in order to implement AI for yourself. Now, before I dive into that, listen closely. Are you looking to implement AI inside of your own company? Or maybe just struggling to get your AI to stop hallucinating? Speak to GPT Trainer.
GPT trainer literally builds out and manages your AI agent for you, eliminating hallucinations for good. Go to gpt-trainer.com. I promise you it'll be the biggest time saving decision that you've made all year. Trying to set up AI on your own is like trying to build a house from scratch. Sure, you could do it, but the time and frustration that the AI may cause to get it finished, it may not be worth it. Instead, it's better to schedule time with our team. It's a thousand times faster and safer to hire professionals.
Once again, that's gpt-trainer.com. To have with me Sarah Levenger, who is the founder and CEO of Tether Insights. is a Forbes featured behavior strategist and direct to consumer performance consultant. Using behavioral science and consumer psychology, she helps brands uncover what actually drives conversions. So you can make ads, messaging, and creative content that finally clicks.
Sarah is also, I might say, the host of Brain Driven Brands podcast. So excited for this show today. Hey, Sarah, welcome.
Sarah Levinger (01:35.032)
Thank you so much for having me. This is going to be so fun.
Kyle James (01:37.326)
Yeah, yeah. We're gonna have a good conversation. So tell us a little bit like what, how did you find Tether Insights? Like what was, what's the origin story with it? How did come to be? And like, exactly do you, do you offer a lot of your clients?
Sarah Levinger (01:48.664)
Heather was kind of born out of a very interesting problem that I kept noticing with some of the brands that I was working with. I had been in marketing for like 15, 16 years, and then I fell into paid advertising on accident, just because I kept having people ask me, you know how to run Facebook ads? So I ended up actually getting on Twitter in 2020, right before COVID hit. And I started interacting with all the people on there because I was trying to learn more about like, what's more paid advertising, like what is the industry like? And I started connecting with a bunch of people over there.
Right as iOS 14 hit. as soon as that hit, it was interesting because I noticed you can't track anymore. Everybody switched their conversation to like from over over where we were talking about metrics and trying to understand like what tactics to put in the ad account over to more creative strategy. What does your creative look like? Who are you targeting in the ad? Those types of things. So I started kind of posting some things that I had learned over the course of about 10 years of studying more psychology focused things.
And again, that was on accident as well. just, had a bunch of people asking me, you know how to do WordPress websites, these type of things, email marketing, like ad campaigns. So I'd go to the library of all places to try and learn that information and trying to upgrade my skills. Right next to all the marketing was this giant section on neuroscience and neuro marketing and psychology, early childhood development. So I just got kind of hooked on the books and I just devoured them over the course of about 10 years. So.
Fast forward to 2020, iOS 14 is hit. Nobody has any sort of idea of like, what should we do? We don't have account tactics anymore. How should we go about making our ads continue to work for these e-commerce brands? So I started sharing a lot of like, well, there's this psychology hack that you could use in your ads. And we could try this particular heuristic. We could try all of these psychology models. And I started just sharing a bunch of the things that I knew because it was a hobby.
And it just kind of blew up from there. So for the last four or five years, will be five years now, I've been kind of developing all these different models based on psychology and neuroscience that can apply specifically to paid advertising. And that was basically just born out of necessity. The brands didn't really have anywhere else to go when it came to learning. Really what it is is message to market matching is what we're really good at together. So that's kind of a history of where we came from.
Sarah Levinger (04:02.76)
Now, we predominantly work with e-commerce brands, but I have worked with B2B, SaaS, service-based, all kinds of different businesses. And we specifically provide market research to help people uncover the emotional, like, psychographic identity and behavior of their customers so they can match messages that will convert.
Kyle James (04:19.95)
Yeah. So I'm curious, like when you, when you like come back to your library phase there, when that happened, like, was that like, when you started sharing content, like, was it just purely just to help people or like, were you thinking, okay, maybe if I start sharing content learning, and maybe it can be like a content builder to bring in more like leads, like, was that initially like what, like, how did the, how did that transition like, guess like happen when you figured out, okay.
Sarah Levinger (04:25.336)
you
Kyle James (04:42.67)
People are actually saying yes to this and then like walk me through like what was going on through your head when you were making that change. And it's been a lot of time studying and learning about psychology behind it.
Sarah Levinger (04:51.678)
I did not mean to be here at all. I tell people that every day. like, I didn't mean for this to happen. I was a freelancer for a consultant for a very, very long time. I didn't have a team. I didn't have a firm. Like I didn't, I didn't know how to do any of what I'm doing today. I had enough people coming in through my content being like, this is really interesting. I studied psychology in college or like, you know, I had an undergrad in this. had a lot of people who were just as passionate about it as I was.
And so I think they really resonated with that message of psychology first, people like you're selling to a brain is what you're selling to. So you have to make sure you're hitting these points correctly. Otherwise people just scroll and it's not your fault. It's not a bad message. It's not a bad offer. It's mostly just you didn't really resonate deeply enough with the subconscious mind. So yeah, I didn't mean to just accidentally kind of happen, which I tell people every day. I'm so grateful that it did. So.
Kyle James (05:44.972)
Yeah. Yeah. I love, I love that phrase. Like selling, selling to a brain. think that's like, we're to get back to it, right? Like it is people for sure. And it's like, but it's also like, we've got to understand the brain because if we understand the brain, then like, we can understand how the human works and how the mind works. And they typically at least clues to like, how do we get the best traction as far as, know, so now you're using AI over at Tether insights. And I love what you guys w why did you decide to start using AI in the first place? Like, what are you doing? And like, what challenges, I know this is a
Three for our question here. So you don't answer all three of them, but like what, what challenges were you trying to solve? Like by using AI within, within Tether.
Sarah Levinger (06:15.671)
You
Sarah Levinger (06:21.336)
I started out in a very strange way because I actually have, well, I didn't even know how to say this. It's not even coming out correctly. So I had a weird startup to this, weird lead up to what Tether was because I actually built kind of like a GPT-ish model inside a Google Sheet before AI was even a thing. So what I was doing was I had a lot of customers coming in asking the same question over and over, which was, how do we know which ad to run?
and what message to run at what time and all these different things. So I went out and I did minimal amounts of research just trying to understand how do I know which ad to run and when. One of the things that I was finding was when I went out and did this kind of like scraping kind of research was if I went on the Instagram comments of the brand that I was marketing for, those Instagram comments were kind of chock full of interesting things that were not related to the product or the brand itself. It was related more to who these people were.
So for instance, I did this first for an Italian leather like handbag company, specifically for men. So we went out and I scraped all these comments off of Instagram and everything I kept seeing was related to why they purchased the bag. So everybody was talking about, I just graduated law school. My son just graduated from high school. Like my dad just retired, those types of things. They would have specific instances of why they were actually coming on and commenting. So I would take all those reviews, stick them into a sheet and then I would go through one by one and.
read every single review and try and make a good judgment call as to which emotion was coming up strongest. And this was coming out of a book that I actually read in 2020 called Marketing to Mind States by Will Leach. Really good book. I always tell everybody go read that because it's so, good. But I took his psychology model of nine different emotions that people usually go through while they purchase. And then I just applied it to this model. So pull down reviews, categorize them into emotions. And then from there, based on frequency numbers, I could tell
Here's the emotion that's coming up the strongest for this brand. And then I would build ads towards that. Now it was tough in 2020 because some of these brands had 5,000 to 50,000 reviews. So I was categorizing all day, all night. It was so much work. As soon as Jeep, like, Chat in particular came out, ChatGPT, all of a sudden it became 32nd project for me. It was not like a three hour long endeavor or a week long endeavor, depending on the brand.
Sarah Levinger (08:44.088)
So I now can take all of my reviews down and run what I now know was a natural language processing analysis on this. So at the time I was like doing NLP work by hand. Now I can do it with chat GBT in 30 seconds. So I started developing more models like this where I was gathering more information using all kinds of different interesting like research formats. The one that I use today is actually with pictures. So I run picture based surveys so that people can have emotional reactions to the images instead of having just like logical.
reactions to words. So we run picture based surveys now and run it through chat GPT to do analysis on correlations between behavior and emotion.
Kyle James (09:22.286)
So, so when you went through the, the, you know, the comments and like the feedback and seeing like going back, cause it's a really interesting, cause like the feedback or the comments like reveal almost like, I guess the emotional part of like, when you went through and saw all that data and obviously even you probably learned a lot doing it manually. And I'm sure you're thankful that now that you're having it, you can automate it, but like, like what, what did that change? Like when you saw these comments and you saw the categories went into, like when you, when you saw that data, what did that change?
Sarah Levinger (09:41.144)
Yeah.
Kyle James (09:50.646)
With your business moving forward to where we're like, okay, Hey, we're going to target more on the emotional side. Like what did you specifically start doing?
Sarah Levinger (09:57.398)
Yeah. So one of the things that I saw most often, because I ran this for that Italian leather kind of handbag company. And then the second one I ran it for was actually a beverage company, non-alcoholic. The third one I ran it for was a beauty brand. The fourth one I ran it for was like a home and gardens kind of thing. So I ran it in about four different industries within the span of about six weeks. And it was really interesting because every single time I pulled down this data, drafted my ads and ran it into the ad account, the same thing kept coming up.
And it was really interesting because I didn't mean again lots of things Sarah didn't mean to do it was just really interesting that they have it. I didn't mean for this to happen but uh CPA was one of the biggest metrics that we were able to mess with which is one of the hardest metrics to actually do anything with so if you're not paid advertising CPA stands for um like um oh geez it just blanked out of my head thank you customer acquisition it kept coming up customer my brain kept
Kyle James (10:48.814)
Cost Brad, Cost Brad. Yes. Look at that. That's a hundred. I, I let's go, man. I got you. I got you.
Sarah Levinger (10:56.918)
My brain kept saying customer and I was like, no, it's not customers. Cost, thank you. Cost per acquisition, geez, now I sound like I know what I'm doing. This is what happens when you run through too much data on the weekly. So my brain is dead. Cost per acquisition, CPA. So CPA has to do with how expensive is it to get a customer in the door, right? So for ads, we track this very closely.
Kyle James (11:00.43)
I thought you were gonna say like a accountant, like a CPA accountant. like, I don't know.
Sarah Levinger (11:20.576)
And CPA is one of the hardest metrics to move, predominantly because it has to do with, is your message really resonating well with that customer type, especially in meta? So I would run ads directly towards the same customers that they were marketing to the last time they ran a few ads. But mine in particular usually came about 40 % cheaper. That's how we get people in. So I could drop acquisition costs by 40 % just by running it towards better emotions based on this research.
Kyle James (11:48.462)
And so when, when like a customer, you know, clients you're working with, what is that something that's facing on the AI side? guess like, walk me through the AI step by step. is it, and when I say that, I mean, like, is it just like internally that you're doing all in the backend? That's what customers, that's what the kind of result they're getting and why it's so quick or like even like on the customer facing the side, like, they, are they using an AI tool at all? Or is it just like, walk me through that.
Sarah Levinger (11:58.039)
Yeah.
Sarah Levinger (12:13.206)
Yeah. So nowadays I tell everybody nowadays I don't really run the NLP as much anymore because you can do it yourself. Like with AI with ChatGBT, you don't have to hire me to do it. You can scrape down all of your reviews, any sort of like customer emails, anything that you have that's customer facing that came directly from a customer, zero party data. You can pull that into ChatGBT and just ask chat, can you categorize these into emotional categories for me and give me the top three or the top five, however many want to track? Very, very simple.
Right? And chat, obviously, is getting better as the years go by. When I first did it with chat, it was very inaccurate. It was probably like 70 % accurate most of time. Now I can get it to be probably 90, 95 % accurate when it comes to pulling out what types of emotions are being, I don't know, talked about the most with these particular pieces of data. The secondary thing I would say, though, if you want to get more robust with AI, like I said, we're using it to actually analyze images. Now this is something that's kind of new for chat GPT.
We just saw that it started to generate its own images at the beginning of this year, which is really interesting. They're not all that great yet. It's working on it. It's getting better as it But analyzing images is interesting. So for those of you who are super tech-driven in here, we actually take the images down that we provide inside these surveys now, throw them into a JSON language so that the computer can actually read it from an actual computational language.
Kyle James (13:18.541)
Mm-hmm.
Sarah Levinger (13:36.856)
Once it's in a JSON language, then we can actually pull it in and give it the spreadsheet of all of the answers that came out of our CIM, out of the survey itself, and tell it to make correlations between what the image looks like or what was in the image and what's in the actual text itself. Now again, these are all correlations. Causation is very, very, very difficult to nail down. It needs a lot of double-blind studies. You've got to be very careful with how you run those. But I can give you with high confidence a correlation that exists inside your business that you did not know existed before.
based upon the fact that we are giving ChatGPT a way to experience the outside world without actually having any experience with it, because that's one of the issues with AI these days is chat's got a lot of words and it's got a lot of knowledge because it can scrape from the internet, but it doesn't have context. It doesn't have experience and only humans have that currently. So with this, we're trying, we're working on it.
Kyle James (14:26.478)
Yeah. Yeah. No. Yeah, no, think it's, crazy how far it's come to, mean, taking it to the next level, the image generation, like that's it's taking data and squeezing every single piece of juice out of it, out of the fruit to get what you're looking for. And like, even though like sometimes I mean, I'd be okay. It's, it's a fully increased, like with sales, but it is giving that revealing data. I'm sure it does increase sales, but like the main thing is like, Hey, getting that revelation is like, what is the data behind it that can, you know,
the side on where, which direction to push towards to in the future. So what types of like, have you seen, you know, since you've been best way I guess, phrasing this like, it's gonna be on both customer side, but even on the AI side internally from what you've been using it for, like what types of results have you been seeing so far since you've been using AI?
Sarah Levinger (15:12.28)
The AI itself is really, really helpful because we can pull out some things that I just didn't expect. And this is something I'm trying to work towards, especially for Tether. We're correcting for marketer bias because marketers are biased. Like we're human just like everybody else. And so we have our belief systems about what our customers want, what they don't want, what their problems are. Usually in my experience, all of those are wrong. Usually we think we know, but we only have like one small view of what's actually happening underneath there.
For instance, one of the best brands that we ran was actually recently, a couple of weeks. They had a beauty product. They were doing really well with this makeup kit. And it was really interesting because they were trying really hard to figure out which direction should we go with our messaging. Because we tried to go this way, more high elevated, influencer-y, like that type of thing. And it did not work. It was not the greatest for them. And then they were trying to figure out which direction should we go now because our customers didn't seem to resonate really well with those kind of high-end influencers. Who should we become as a brand this year?
So we ran a CIM, pulled down the data, analyzed it using AI, same exact system that I just shared. And we found out pretty quickly that these customers weren't interested in what we call esteem-focused purchasing. So they weren't really resonating with like the new year, new you kind of a thing. They don't really want like to be glamorized or to look different. A lot of these women were actually like, I just want to feel like me, but just put together. So it was more composure than confidence.
Kyle James (16:36.087)
Hmm.
Sarah Levinger (16:40.982)
And that was a very distinct shift because for this brand who had been pushing like confidence for the longest time, because they're in beauty, why would you not push confidence? This is a very, very key finding that we would not have been able to find had we not ran it through AI. Cause the AI can spot things that I can't. So that one in particular, we'll see how their ads run pretty soon, but they're going more for a feel composed, feel like yourself look put together. And that should do wonders for their ads and especially for cost. Yeah.
Kyle James (16:57.07)
Wow.
Kyle James (17:08.492)
Yeah. And I, like, I imagine too, like the, just the shift between like the, you said it a confidence and the, what was the other one you said? Composure. Like those are, they're two, like, very, like very similar, but they're different and like in different ways. like, I'm always thinking like, like I play the guitar and like piano and stuff like that. And like, specifically on the guitar, like you can go like half a step down, but like that half a step down, like can completely throw out a tune where like people can pick up like, this is off. And even just that small trickle where like you just, just.
Sarah Levinger (17:14.39)
Composure, yeah, like put together.
Yep.
Sarah Levinger (17:32.31)
Yes. Yep.
Kyle James (17:37.004)
just a little bit of a half step, like it sounds that much better and it's way more, you know, rhythmic and, and, and, and so all the things, so like, I can see this, like it's so important to make sure that you have clarity as to like which direction to go to. Cause if you're just a few steps off, like long-term that might completely derail you. and I'm sure that's game changing. and, and like, as far as like upcoming, you know, AI is changing and you've seen it's been, getting better.
Sarah Levinger (17:39.064)
Exactly.
Sarah Levinger (17:56.6)
Mm-hmm, 100%.
Kyle James (18:05.294)
Like for you though, like in Tether Insights, like what are some of those kind of maybe upcoming AI initiatives and like, where do you see AI playing maybe some of the biggest role in your operations next, both internally and externally?
Sarah Levinger (18:13.964)
Yeah. There are some interesting things happening in e-commerce for sure. I think the e-commerce brands at this point have very quickly become experts, expert level users of the AI. The interesting part about it though is I'm noticing a lot of people that are on the fence about should we even be using it? Like what is this doing to our brain health? We're hearing a lot of like news about like, is this a good thing or a bad thing? One of the things that I see, at least on the e-commerce side, people are using it predominantly for copywriting.
like very heavily for copywriting and for data analysis. you know, unless you are a data analysis analyst, it's very difficult over here to know which pieces of data should we run? How do we analyze this? How should we prompt to do some analysis on this? So it's still evolving. I'm not entirely sure where the connection between the humans and the AI will go. But at this point, it's just cutting down, I think, on the time it takes to find stuff. It's almost like a library than anything else.
cause at this point we have so much data we can't, we don't even know what to do with it. So AI is just cutting down on the time it takes to find these gems.
Kyle James (19:20.78)
Yeah, absolutely. I love that. And Sarah, I definitely appreciate you having the show today with Eric Chronicles and for those listening and maybe they're just curious to learn a little bit more about you and about Tether Insights, where do you recommend that they go to learn more?
Sarah Levinger (19:33.992)
You guys can actually check out my newsletter, my podcast, like we said before, Brain Driven Brands is my podcast. Brain Driven Brands, it's hard to say. And then if you guys want to go to signal.tetherinsights.io, hopefully we can put that description in there for you. Yeah, that's the best place to go. Mostly because this is where I'm sharing all of these psychology breakdowns from the brand side. I'm also including trends. So when you do consumer trend research over there, you guys can come and just get free.
Kyle James (19:41.632)
Hahaha
Sarah Levinger (20:00.726)
access to consumer trends that are working right now. So yeah, either one
Kyle James (20:04.174)
Awesome. Love that. Love that so much. Thanks again, Sarah. It great having you on the show today. And thank you for everybody else who's listening in. And again, if you're looking to implement AI into your business today, please don't try and do it yourself. The time and stress that the AI may cause, it may not be worth it. Schedule a call with GPT trainer. Let them build out and manage your AI for you. Once again, that's gpt-trainer.com to schedule a consultation. Signing off for now, have a great rest of your day, everybody. And looking forward to seeing everyone on the next episode of the AI Chronicles.
Sarah Levinger (20:07.116)
Thank you.
