He Analyzed The Entire Website Visitor ID Industry And Now Thinks It's A Complete Scam — Larry Kim | How Visitor ID Tech Works, Why Bad Visitor Data Is A Disaster For Your Marketing, The Impact Of Bad Data, How To Test Your Visitor ID Provider (#381)

Enjoying the Ecommerce Coffee Break Podcast? Here are a few ways to grow your business: https://ecommercecoffeebreak.com/level-up/ 
---

In this podcast episode, we discuss the truth behind website visitor identification accuracy. 

Website visitor identification technology claims to turn anonymous visitors into sales opportunities, but is it actually effective? Larry Kim, CEO of Customers.AI, has analyzed the entire visitor ID industry and concluded it's largely fraudulent. He reveals how most providers deliver data with only 3-25% accuracy, potentially poisoning your email marketing and destroying deliverability.  

Learn how to test your visitor ID data's accuracy, understand the catastrophic risks of using low-quality data, and discover what truly effective visitor identification technology looks like. 

Get a Free Data Accuracy Test -- If you have already using a Website Visitor ID solution, Customers.ai will run a side-by-side test to prove the accuracy of our data. New to Visitor ID? Run a free trial and we'll give you 500 free contacts. 

Topics discussed in this episode: 

• Why is website visitor ID a scam? 
• How does visitor ID tech work? 
• What makes visitor ID data so inaccurate? 
• Why is bad visitor data a disaster for email marketing? 
• How can you test your visitor ID provider? 
• What’s the real impact of bad visitor ID data? 
• Why do match rates lie? 
• How is Customers.AI different? 
• What should agencies and brands do? 
• Where to get a free accuracy test? 

Links & Resources 

Website: https://customers.ai/
LinkedIn: https://www.linkedin.com/in/larrykim/
X/Twitter: https://twitter.com/larrykim

Get access to more free resources by visiting the show notes at
https://tinyurl.com/3f7cu5db

MORE RESOURCES

Store Optimization Beginners Guide: Instant PDF Download! It's FREE. 👉 https://ecommercecoffeebreak.com/essential-conversion-optimization-guide/

Top Tools, Apps, and Services used by leading Ecommerce Brands: https://ecommercecoffeebreak.com/quiz-results/

Become A Smarter Online Seller In 5 Minutes
Get free ecommerce insights via email every Tuesday & Thursday. No cost, just value. Sign up now at https://newsletter.ecommercecoffeebreak.com

Rate, Review & Follow
Enjoying this episode? Help others like you by rating and reviewing my show on Apple Podcasts at https://podcasts.apple.com/us/podcast/ecommerce-coffee-break-digital-marketing-podcast-for/id1567749422

Follow the podcast to catch all the bonus episodes I am adding. Do not miss out. Hit that follow button now!

Transcript

[00:00:00] Hello everybody and welcome to the E Commerce Coffee Break Podcast. I'm Claus Lauter and you're listening to the podcast that helps you become a smarter online seller. In today's episode Larry Kim, CEO of Customers.AI joins me and he has analyzed the entire website visitor ID industry and now [00:00:20] thinks that it is a complete scam.

So, let's find out why.

Hello and welcome to another episode of the eCommerce Coffee Break Podcast. You likely have seen visitor ID tech that helps turn unknown website visitors into sales opportunities. But are these tools really effective as they claim? Today's guest, Larry Kim, has put them on [00:00:40] the test and uncovered some very surprising results.

Larry is the founder and CEO of Customers.AI, a marketing technology company specializing in website visitor identification and AI powered customer acquisition. He's also the founder of WordStream, a leading PPC software company, managing over 1 billion in annualized ad spend for tens of thousands of agencies and [00:01:00] businesses globally, which was acquired by Gannett in 2018 for 150 million.

Let's welcome Larry back to the show. He was here before and dive deep into this very interesting topic. Hi, Larry. How are you today? Doing great. Thanks for having me back. Can you explain for listeners who haven't heard about Website Identification, Visitor Identification, [00:01:20] what it is? It's a relatively new emerging technology that's been around for a few years.

Uh, and the, uh, the promise that they are making a claim that they can do. Uh, is that by installing a JavaScript code on your website, uh, that ID providers can claim to accurately [00:01:40] identify the identity of the person who is visiting your website. And of course the value prop here is. Potentially very high because, you know, conversion rates are stubbornly low.

Like they can be like mid single digits typically. Uh, and if you can know who, you know, people who didn't convert with, uh, you know, [00:02:00] organically, you could then, if, if you were able to identify those people, you could. You can pull up with email or ads, um, if, if you had their email addresses or, or other, uh, identifiers.

So that's sort of the, the premise of, of, of Visitor ID Tech. There's quite a few vendors in the space today, uh, and, and Customers. ai is one of them. [00:02:20] Now, this is a very promising technology, um, finding out more about your website visitors, but you found out or you realize that there is a problem there and we want to dive into this.

What brought you to the conclusion that you might need to dive deeper into it and find out what is going on? [00:02:40] Um, so recently, well, about a year ago, we, uh, it came to our attention that the way that we were sourcing data was meaningfully different from how other vendors were We're doing it. Uh, and that just [00:03:00] uncovered a kind of a rabbit hole of, of us wanting to do, uh, data accuracy tests on not just our data, but also the other vendors in the space.

And what we found was that, um, the, the whole industry is basically a big fraud. Uh, this is like [00:03:20] seriously that the, these ID vendors. Are just producing crappy shit, garbage data and selling that to you. It's not 0 percent correct. It's, it's, it's correct. Maybe 3, 4, 5 percent on the low end, uh, to as high as like 25 percent on the high [00:03:40] end.

Um, these things are wrong, like 70 to 95 percent of the time. Uh, like when they give you an ID and say like, this is so and so on your website. It's actually the wrong identifier. Uh, the vast majority of the time is, is, is what we found. And we repeated these tests across like every possible data vendor that we could, [00:04:00] we could try.

Uh, and we, we believe they're all. Pretty awful. Uh, you know, it's not a hundred percent wrong. Like there's a small amount of signal in there, uh, but it's, it's mostly wrong. Like, like it's one in 20 is correct. Or one in 10 is correct. Like it's, it's, it's, it's pretty bad. [00:04:20] And we. I believe that the issue here is that these companies in this space are all sourcing their data from the same, uh, you know, toxic, toxic waste pile data companies.

Um, and, uh, it doesn't even matter like they're the, the, the, the data companies are being used are like [00:04:40] typically like core edge networks or DSP data, um, companies. And they're just kind of trying to work, work with that data, uh, that data. And for, you know, high accuracy type, um, utilization and, and, um, you know, they're rife with, you know, fraud and [00:05:00] bots and fake, fake, uh, you know, shenanigans and, uh, and so it's just kind of de anonymizing the toxic waste that, that they're, they're licensing is something.

So, um, It was, uh, you know, on the other hand, customers AI where we're the [00:05:20] opposite. We are mostly right, like it's nothing is 100%. But we do see that when we run the same data tests on our data offering, uh, that the, the accuracy rates are mostly right. So it goes from like mostly wrong to mostly right. Uh, it's typically like, you know, [00:05:40] 65, 75, 80%, uh, you know.

Correct. And, uh, the, the, the important thing is you don't actually have to just believe me, uh, just verbatim. Uh, there is a way to, uh, you know, do a little bit of analysis on the data being provided by whatever data provider you're using. Uh, and, [00:06:00] and you can kind of arrive at the truth. Um, so. I want to dive a little bit into the implications that that can have for a online seller or somebody who has an online store of things.

Now, he has the correct data from website visitors and obviously gets data delivered email addresses. Um, [00:06:20] what can happen to your own email marketing, for instance, or generally in marketing, if you're using. I think it's catastrophic, Claus. The issue here is that, um, if you're just using crap data for ads, I mean, it's not necessarily the end of the world, uh, because [00:06:40] like Facebook can kind of, can kind of do some, um, You know, auditioning of your ad to see who's interested, but if you're using it from an email perspective, um, the, the risk is catastrophic, like, like an existential risk to your, to your business, essentially, uh, you know, [00:07:00] clearly email marketing is a very, very important channel for, uh, DTC e commerce, uh, retailers and, uh, If you're using these, you know, ID providers that have like, you know, 25 percent accuracy, that means that the inverse of that, you know, 95, 90 percent [00:07:20] of this data is wrong.

That is, is going to torture your email deliverability, your, your, your email reputation, the, um, current spam, you know. Threshold for, uh, being, you know, classified as a bulk sender is a 0. 3%. So like, you know, imagine if you're, you're [00:07:40] randomly emailing people, I mean, this is not an ID provider. These are worse than cold emailing systems.

Like even with cold emailing systems, like what you can do is you can. Get certain people with certain job titles or certain demographics. You see what I'm saying? And then just email them out of the blue here when these data providers give you a wrong ID, [00:08:00] it's not just like, Oh, it's another buyer, you know, and we're close.

Like this is somebody who, uh, I don't know, was on the same IP address as someone else, you know, at the same time and the, uh, the data provider that they're licensing, you know. Has a thousand of these people on the same I IP address, and they're saying like, oh, well, let's just, let's just include them all.[00:08:20] 

Okay. Uh, . So, so, you know, it's, it's like a, it's like going to, to the wrong person in line at a, like a Starbucks or something like this, , and like, it's, it's not just a little bit wrong. It's, it's completely wrong. And so, you know, the, the, uh, negative, uh. Uh, attribution, sorry, the negative email engagement rates was like unsubscribes, [00:08:40] complaints bounces.

It's going to be inversely proportional to the accuracy. So if the accuracy is like, you know, single digits or low double digits, those complaint rates are going to be sky high. And you might be wondering, well, why, if this is true, why hasn't everyone been kicked off their clavia? Well, first of [00:09:00] all, we've heard many stories of people being, uh, you know, Kicked off of Klaviyo, uh, because of, of, of, um, you know, compliance and deliverability issues, uh, when you, when you start pumping this toxic waste in, in, in, into your ESP.

Uh, but secondly, it doesn't happen all at once. It's like being poisoned slowly. [00:09:20] So imagine you're The thing about these visitor ID solutions is they actually provide a lot of IDs, like you might be getting like a thousand new emails a day or something like this at, say, 10 percent accuracy. So that 90 percent error.

Okay. Um, that's off of the thousand new. Um, and [00:09:40] an e commerce store might have, you know, half a million emails in their Clavia or, or ESP like on day one. Okay. So that 90 percent error is then diluted by the fact that their domain is doing a lot of first party emailing that that's kind of like, uh, diluting away the toxic, [00:10:00] toxic waste.

The problem with this is like, if you. Stay a customer of one of these crappy shit data, uh, ID providers, um, the, the, the problem kind of just, it just poisons your ESP over time, like, so at one month, you [00:10:20] know, in this example. You're at 30, 000 emails at, at three months, you're at 90, 000 emails. Like that's, you know, percent of your, of, of your half a million emails.

It seems like, and if, if, uh, 90 percent of that's wrong, like that, it's just like, it's just, Uh, chipping away at [00:10:40] your deliverability and, and it's just a matter of when, like, maybe it's like three, four months down the road, like at some point, or if you're a smaller, uh, you know, brand that's a starting and you're using this, like that, that tipping point from when you're, when you're ESP is mostly like garbage, um, that'll happen very quickly.

Like it could happen in the first week. [00:11:00] Um, so. It's just, um, it's just a very, uh, really dangerous situation, uh, that all these vendors are using, um, uh, you know, toxic, uh, visitor ID data, not really having an understanding of like how that data is created or how to [00:11:20] test and understand what the accuracy of that data is.

No, that makes perfect sense. And I like the example that you mentioned with Starbucks. I work a lot of coffee shops. So basically I go to an online store and then everyone sitting around me, because we're all on the same IP address, one of a sudden shows up in your database with their email address as far as they can reach through obviously very bad implication.

[00:11:40] So Obviously, these providers and these database providers give the whole industry a very bad rep. And we were mentioned in the pre call also that you can see it from your conversion rates because you're not really targeting the right clients. Tell me a little bit on how that reflects in conversion rates.

Um, so we've done, uh, case studies with major brands. [00:12:00] Uh, and, uh, one, one such case study is with, uh, a retailer called Jordan Craig and, um, you know, one vendor. Um, you know, gave, you know, 14, 000 emails, uh, and generated 20, 24 purchases. So that's like a very, very tiny [00:12:20] conversion rate. Like remember, it's not all wrong, but it's mostly wrong.

Uh, so 24 out of 14, 000, that's like. Point Oh, you know, something like point Oh 4 percent or something. It's a fraction of a fraction of a fraction of a percentage. Um, customers. ai gave approximately 6, 000 emails. So like less than half the number of emails, but it generated like [00:12:40] 224 purchases. Uh, so that's.

Eight times more purchases with half as many emails being produced. Uh, so, you know, it's 16, 20 times higher, uh, conversion rates. That's, it's not just a little, it's, it's like an, more than [00:13:00] an order of magnitude difference here. Uh, another vendor was even worse. Um, they, they did like, you know, a similar number of emails, like 8, 000 emails, uh, and it, and it produced one sale.

So that's, that's like 0. 01%. Uh, and then another one did, um, you know, three sales off of, off of 10, 000 emails or so. And [00:13:20] it's almost worse than just randomly generating emails and, and, and, and, and sending, sending them out. Uh, there's just a lot of, a lot of error. Can we talk about like how to test the data at all?

Oh, yeah, of course. Yeah, sure. I would, I would be interested to hear. Uh, so. I think that the challenge for all these unsuspecting customers that are using all these and [00:13:40] agencies like, Oh, my God, like agencies like using this stuff and not even knowing what it is, um, the, the, the issue here is, uh, it's, it's a little tricky at first to understand if the ID is being provided by the ID vendor is correct or not.

Like there's an assumption that it's true because that's what they're. Um, but like [00:14:00] most of the discourse in the, in the, in the space has been revolved around like match rates. So like, Oh, we can match 20 percent of your visitor visitors and someone else comes up, uh, another startup comes up and says, we can do 40%, like it's the match rate is, is, it's kind of a joke because like I can, Don't [00:14:20] maturate just by, you know, generating more, uh, phony, you know, fraudulent emails.

Uh, and, and, and the way that you can check to see if the guess provided by the ID provider is correct or not is to establish a baseline of truth. Uh, that baseline of truth is typically your first party data. So if, if somebody [00:14:40] actually purchases from your store, uh, at that instant in time, you should have a very, very high conviction.

This particular session. Is Claus or this is Larry, like you'll know exactly who that person is because they just purchased. And so what we've done is we've just made it easier for people to then rewind that [00:15:00] session to see what the guesses were by, you know, any of the available, uh, visor ID. Providers, uh, and they're usually wrong.

Like, like it's, it's just, you know, and you look at the, you look at the emails of who they thought it was and who it really was, they're not, [00:15:20] not even close. It's just like, what, you know, it's just, it's, it's, it's a random person who was routed through the same, uh, you know, network, uh, at, at the same time, or, you know, has a similar browser configuration.

Like it's, it's just completely wrong. The vast majority of the time. Uh, and, uh, and this [00:15:40] is a, this is a big deal, Claus. Like people aren't testing it. And what we've done is we're making it easier for people to see this, uh, by offering like a free data testing, uh, service. Well, obviously at customers. ai you do something different because your results are far better than [00:16:00] anyone else in the market.

So what, what's, where's your approach different? Uh, well, um, we. Do this fundamentally differently from, from everyone else. Like just the whole premise of how, um, our ID graph is, is constructed is, is completely different from, uh, how everyone else is doing it. [00:16:20] And without going out into all the details, what I can tell you is that the founders of the companies who do this kind of ID graphs are definitely like, yeah, they have that panache and they're like, uh, you've, you've met some of them, uh, they have this panache and they're like.

Really just like good at hyping [00:16:40] things. Like, like really, uh, none of them that we've looked at have any. Serious engineering, uh, abilities. Like there's mostly like a, like a sales and marketing, um, kind of go to market. Uh, where we're different is that customers AI is an engineering [00:17:00] company. Like we, you know, most of our company are engineers.

Um, my background is, is, is in electrical engineering and we just have a. Approach the problem from, uh, I would say in the last 18 months or so, we've been approaching it from a meaningfully different, um, uh, angle for then from, from, from what [00:17:20] the industry is doing. And, uh, I think that the way that the industry is doing it right now is ridiculous because, uh, none of those sources that vendors are licensing are just black boxes.

They don't even know, like, uh, like they don't even know what's. In that data, they're licensing because, um, it's just like a feed, like you, you just, [00:17:40] you didn't have anything to do with how the ingredients were packed, packaged or anything like it's just like, here it is. It's like, okay, I'll, I'll use it.

Like we, we control, uh, Uh, the, the, um, kind of the build, uh, and the sources at, at, at, at every stage, uh, it's, this is, this is very different. [00:18:00] Uh, and, um, I, I, I, I, I think that, uh, the fact that none of these other companies have. You know, done anything to fix their data, uh, leads me to believe that, like, it's just not possible for them.

Now, obviously, somebody who's using these systems and you [00:18:20] said you're providing a free data accuracy test so people can go and really check on the quality of the data that get from any provider out there. How does it work? Somebody comes to you, how can they run the test with you? It's pretty simple.

Like most of these ID providers have a way of like either exporting or, or sending the data [00:18:40] somewhere. Uh, and, and so, you know, just if, if you happen to have, uh, you know, some other ID provider, we can just, uh, you know, work the data and, and, and, and kind of connect the dots. It's important to understand that the data tests that we're providing here isn't necessarily something that.

Couldn't be [00:19:00] done by any individual brand because like the brand knows who the real purchases are and they know what, what data is being provided by these, these ID vendors, we're just doing a better job of matching them up, uh, to, to, to see like, you know, you know, were they right or wrong? And this is, this is something that they [00:19:20] could do if, if they had the technical wherewithal, um, and, and that they could reproduce like on their end independently, but.

Okay. We're just offering to give them a tutorial on, on, on how to, how to do that, uh, matching up of, of data just because it's, it's really not that, that complicated. I think it's super [00:19:40] important because as I said, you're poising your data over time and a lot of marketers are only interested in volume. So if you're sort of blind, if you see, Oh, I have a thousand new emails coming in per day, makes every marketer out there happy.

But if the data quality is just not on spot, you're really hurting yourself and you don't want to be kicked out of [00:20:00] Klaviyo for that. It's, it's, it's, it's happening. Uh, and, and, you know, the churn rate for a lot of these products is extremely high, it's as high as like, you know, customer life and value of like three, four, five, six months, um, which is, is, is extremely low.

Um, it's just really funny because [00:20:20] even when it's complete crappy shit garbage, like still get these 40%. Open rates and you still get these like 26 percent clicking rates. It's, it's these people saying like, what the heck is this? Like they open it and they click on it. You know, so, um, it's not immediately obvious, uh, what's happening [00:20:40] right now, but, but I would.

I would, I would be focusing on things like the, uh, the complaint rates, the, um, you know, which, which are always higher than, uh, you know, 0. 3%. I would be looking at the, um, uh, the, you know, the overall, overall accuracy and, and the, um, uh, the, the, [00:21:00] the, the, the conversion rate from, from, um, You know, like if you're getting, you're getting like 20,000 emails or something, like there should be more than, you know, a couple dozen purchases, one

Absolutely. Let's talk quickly about, um, customer ai. Who's your perfect customer? Uh, well, uh, [00:21:20] we love, uh, e-commerce companies. Uh, uh, 3 million GMV and up. Uh, is, is, is. We're focusing on. Okay. What's the onboarding process. If somebody really wants to move away from a, uh, provider that does not provide, um, so [00:21:40] it's, it's pretty simple.

We would want you to feel comfortable. So we would work with you to just do it, do a free data test and to see, see, uh, you know, what, what that is. It's. You know, spoiler alert, it's, it's going to be shocking, like it's going to be, you know, 5, it's going to be mostly [00:22:00] wrong. Um, and, um, and then, and then transparently, we, we can, we can offer to do the same thing with, with, with our data.

And, and, uh, you know, it's not, not perfect, but it'll, it'll be, uh, you know, it'll be a multiple better. Like we've seen as high as like, uh, you know, 16 times more accurate or two times more accurate. It's, it's somewhere [00:22:20] going to be on that spectrum. And, um, and then you can, um, decide for yourself, uh, you know, which way, which way to go.

That makes perfect sense. And I think it's definitely worth trying it out because you don't want to wreck all the data that you have and more coming in by the day. Yeah, like imagine if you're an agency and you're like [00:22:40] recommending the use of one of these ID providers to, you know, a portfolio of like 20 clients or something like this, like that's an existential risk to your agency.

Uh, because like, if you don't like, you should at least do a test to see what it is that you're recommending. You're pumping into your clients accounts like you don't have to do [00:23:00] anything with the information that you could say like, yeah, well, I'm happy with 25 percent accuracy or something like you can, you can, um, you can make that determination separately, but to not even know, uh, like what.

What the accuracy is of the data that you're using from your data provider is, uh, is, I think is, is, is quite irresponsible and, and, uh, like as a, you [00:23:20] know, you're, you're supposed to be helping them grow and looking out for them, your, your clients. And so I think, uh, this is, is just like a, it's worth getting a data checkup is, is what it's our say.

Absolutely. And I think the idea and the technology is there to make it work. And I appreciate that you're helping [00:23:40] our listeners in getting to the truth of it and hopefully getting better data out of it. Where can people go and find out more about you guys? It's customers. ai and, uh, if you go to customers.

ai slash test, uh, there's the, you can sign up for a free data evaluation. Okay. I will put the link in the show notes and you just want to click away. [00:24:00] I hope a lot of people will reach out to you. Not only the ones that are already using another provider, but also the ones that are generally interested in the topic of, um, finding who's on their website on their store and getting more data from that.

Yeah. Uh, in summary, I just think it's, it's, uh, something that. A lot of people may have had some suspicions around like, you know, like [00:24:20] it should be doing better. Why is that? Why doesn't everyone use this technology? Like, if it really is the correct ID of all these visitors and, and, um, you know, it's, it's, this is like a second generation, um, visit verification.

Like if we can actually do what we're saying, we're Doing. I think, I think the, the, [00:24:40] uh, the upside is tremendous. Thank you. Thanks so much. Um, let's keep in touch and I will be interested in how this goes into the future and how more accurate this can get. Thanks so much. [00:25:00] [00:25:20]