#127 Ensuring robust data collection – lessons from Trackingplan (with Josele Perez)

[00:00:00] Dara: Hello and welcome back to the Measure Pods. I’m Dara joined by my established and trusty co-host, Matthew. Matthew, how are you doing? I’m very well. How are you? I’m also, well, I’m now in Spain. Yeah. Finally, Yeah. I could tell by the decor Yes. For anyone who, you know, watches, podcasts, weirdos, for anyone who watches us. Yeah. I only mean weird if you watch us not watching podcasts. 

[00:00:41] Matt: I do. I also watch podcasts. Yeah. So I’m keeping myself kind of mixed between it, but I do for some reason like watching people talk. Yeah. I don’t know. 

[00:00:49] Dara: No, fair enough. So I’m weird. So if you are watching this, yeah. You will see my traditional Spanish.

[00:00:57] Dara: Is that it? That’s all, all I’ve got to say about it. No. Yeah, settling in. I shared a picture with you. I don’t think I’ll share it in the show notes. So I want to try and maintain some element of professionalism, but my setup here is not exactly high tech. It’s a pilot Mac. It’s MacGyver. It is very MacGyver.

[00:01:17] Dara: Yeah. Yeah. So still on, yeah. Shades of MacGyver, load of box. So behind you can just see some Spanish colors to prove I am here. Although I could be anywhere really, I could just have a very Spanish taste in paint colors and furniture. The other way is all boxes that are yet to be unpacked. So yeah, the new life in Spain has started.

[00:01:42] Dara: Good. I’m not jealous in the slightest. No. See, see, see, see as though you’ve asked. I’ll say one more thing about it and then I’ll, and then I’ll stop. But we had a festival here in the town that I live in called Nche Delino, which is night of the wine on Friday night, which is, you know, just an excuse to, to drink wine, really, which I’m gone down for. 

[00:02:03] Matt: Which I did counter with the fact that in my village in the UK in Stockport, we had the whole village gather around in the precinct for an Elvis impersonator.

[00:02:14] Matt: So, you know, who’s winning really, I think probably should dove, you know, but whatever. It’s close. I tried. It’s close. It was fun. 

[00:02:23] Dara: Yeah. I would’ve liked to have seen the obvious impersonator, but I had, flamenco did flamenco instead. Good. Alright. Enough about our silly personal lives. Let’s get down to it, yeah. Nitty gritty of data and ai. 

[00:02:40] Matt: Yes. In the Salty News, last time out, I accidentally predicted that. 

[00:02:48] Dara: Sorry, sorry. Did you say it accidentally? Or accidentally? Because it sounded like you could have said either you hedge new, be accidentally, accidentally accident. Accidentally predicted 

[00:03:01] Matt: great journalistic integrity, and thoroughness. I said Chatty PT five will be out in August. And as that podcast came out, it had come out the day before. So to the listener, that would’ve just been the worst possible prediction. ’cause it would’ve just sounded like I said, it’ll come out this month when it had already come out. Yeah. but we do record those before Friday just to try and defend myself. Yes. GPT five was released along, well weighted. 

[00:03:29] Dara: And so far all the reports are, it’s, it’s brilliant. It’s everything everybody hoped for. And, a GI is here. 

[00:03:36] Matt: Yeah. 

[00:03:37] Dara: Not quite. 

[00:03:38] Matt: No. It’s been a bit bumpy. Been a bit bumpy. So they’ve consolidated everything into one, which is the first thing people didn’t like.

[00:03:46] Matt: So they don’t have, rather than, which I kind of get and I kind of don’t, but they originally you had oh 3, 0 1, 0 1 light, 4 0 4 0.14, mini four flash, whatever, thinking reasoning, this, that, and the other. They kind of just consolidate down to GPT five, GPT five nano, and another one. So three sorts of levels of the model.

[00:04:09] Matt: And the idea is decided based on your query, what it’s going to do. Like, oh, this, this needs a bit deeper thought, or I can answer this really quickly. And it sort of selects the model behind the scenes. People aren’t very keen on it. I mean, I did see a lot of people who were keen on it because it sounded like they’d formed some unhealthy relationships with chap pt, GPT-4 oh, or whatever it was.

[00:04:35] Matt: And like. They lost their AI girlfriend or something and yeah, was replaced by the harsh G PT five mistress. 

[00:04:46] Dara: Not sure where to go with that. Here we are. I was, this was dangerous. Just really carrying on with this analogy. So take it up and run with I’m just gonna, I’m just gonna stop that one right there. I did see that.

[00:04:58] Matt: I did See, 

[00:04:59] Dara: I think you’re right there. I think that I was picking up a bit of that as well, where people were kind of griping. It’s the usual thing though, isn’t it? It’s like, oh, you’ve changed it. The thing that I knew is not there now, therefore I don’t like it. Yeah. Even without actually properly trying it, whatever.

[00:05:12] Dara: So I was keen to ask you that ’cause we haven’t actually talked about it. People probably think we just talk about this stuff all the time, but, you know, believe it or not, we actually haven’t talked about this before recording this. No. Mostly this is actually the only time we speak. No. Even then it’s hard work.

[00:05:29] Dara: But I was keen to ask your take. ’cause I, you know, I, I obviously saw the, saw the kind of backlash, but it did seem like a lot of it was, I wanna call it, well, maybe superficial is the right word, but it was kind of like people were griping about things that, that, you know, were just certain aspects of it rather than it being, you know, clear cut that this isn’t any good and it was, I know it is better at certain tasks.

[00:05:54] Dara: Maybe there are some things that aren’t quite right. Yeah. So what, like, what’s your take on, where’s the truth? What, you know, how good is it or not compared to chat for, 

[00:06:05] Matt: I think part of, part of the problem here is that you, everyone was expecting some huge leap, like a 3.5 to four, but I feel like we’ve had, so it’s really entered into a, a, a very incremental sort of update phase at the minute.

[00:06:23] Matt: Yeah. Whereas I think on a, on an annual biannual basis, if you compare where we are now to where we were exactly one year ago. It’s still a massive change, but the change between four points, whatever they got up to, or oh, three to five, isn’t that giant leap forward? Like there was from 3.5 to four? Sorry, three points.

[00:06:46] Matt: Yeah. 3.5 to four. From a coding perspective, I think I mentioned this in our company Slack. I feel in the last six months the changes in its coding capabilities or multiple different models like coding capabilities are off the scale, compared to six months ago. So I, I dunno, I think it’s people expecting the world.

[00:07:07] Matt: Yeah. And it’s not quite living up to that incremental change going on all the time. Kind of, we’re kind of in the pot getting boiled a little bit. Mm-hmm. We’re just getting these improvements all the time that we’re not really noticing as much. and maybe just specific domains find it easier to notice than other domains because if you’re just chatting to it.

[00:07:29] Matt: I guess it’s difficult to sort of, yeah. How can you gauge what, yeah, what’s a generational leap in chatting to a chat Bott, you can see a generational leap in image generation or in code generation, but it’s trickier in just day-to-day conversations. 

[00:07:46] Dara: But on the expectation Yeah. Points. I feel like they shot themselves in the foot with that a little bit because I, I, I think you’re right.

[00:07:53] Dara: And I think if, if it had been presented as an incremental improvement or, you know, if it had been messaged differently than maybe it would, I mean, there’s still gonna be backlash, isn’t there? It 

[00:08:06] Matt: hmm. 

[00:08:07] Dara: It was teed up as being like, this is gonna be a big leap. Yeah. And it kind of wasn’t. So I can’t help but feel like they’ve shot themselves in the foot a little bit with that.

[00:08:18] Matt: Yeah. It’s almost really a consolidation effort and a rebrand dressed up as GPT five. Yeah. yeah. Which is probably, probably not the way they should have gone. The other thing they’ve done with it is they’ve kind of, they’ve tightened it up in terms of what it will and won’t answer. I’ve seen some in, I’ve seen some instances where it’s just sort of said, I don’t know the answer to that.

[00:08:40] Matt: And that’s not something I’ve really seen LLMs do before. They don’t know the answer to it. They tend to hallucinate an answer. Yeah. so they’ve really sort of tightened that up and you can tell it’s a little bit more forthright when it comes back to you. Oftentimes it’ll be like the blunt truth of it or something like that.

[00:08:55] Matt: Say sentences like that where it’s trying to be a bit more less chantic. Yeah. Less hallux, hallucinatory. Is that a word? It is now. It is now, and I think I was saying this in the slack conversation we were having as well, but to me it feels like reducing down hallucinations, losing down mistakes and making it a little less sycophantic is all in an effort to increase its utility with agentic applications.

[00:09:21] Matt: Yeah. ’cause. The less predictable an outcome you can get and the less it makes stuff up within these sort of agentic systems, the more powerful they’re gonna be. And it certainly feels like just the technology where it is now, their capabilities of these models, where they are now, that you can show them up and make them accurate, more accurate, there’s a hell of a lot left of road to go just to, yeah, get the most outta what’s already there.

[00:09:43] Matt: Yeah, totally agree. But 

[00:09:44] Dara: again, why not present it as that? Yeah. Because that, yeah, you’re right. I think you’re right. Like that’s where the real value’s gonna be, isn’t it? It’s like if you can have that agent to agent, you know, it goes off and does something and you don’t have to worry about it completely making something up, then that’s, that’s usually powerful.

[00:10:01] Dara: But I dunno, is it because of the hype around a GI Is it kind of, but then why would you, even if it is, why would you contribute to that hype if you know your model is not actually gonna be that. Yeah. Generationally forward. 

[00:10:14] Matt: Well, I was listening to a podcast v. A go to Spain about, it was with Claude, it was with Claude, no Anthropic.

[00:10:23] Matt: CEO. Mm-hmm. He was saying how they don’t, they just won’t use words like a GI, he believes there is no definition to that. And essentially a GI at this point is just becoming a marketing Yeah. Label. He’s just become marketing speak and he’s trying to say, you know, open AI and Google and all these people using GI terms and throwing ’em around is just don’t listen to it because nobody knows what it is.

[00:10:45] Matt: No one has a clear definition, nobody agrees on it. So they just strive for further, further utility checks basically, and, and get the most out of it. and open ai, I think are the most guilty of chucking those kinds of words around, well, they don’t, I don’t think they’re particularly helpful. There was another, I can’t remember where it was now.

[00:11:08] Matt: I wish I, I wish I’d grabbed this graph before it came on, but it was essentially a survey of people. And the question was, who will have the most powerful model at the end of 2025? And it’s been open AI steadily rising forever, and Google sitting well below underneath it, literally the day after chat, GT five came out, flipped around really completely.

[00:11:33] Matt: And Google is now well on top and open eyes lower down. It does feel like, again, we’ve said this before, but Google is still able to just keep flexing their muscle and just feels like their size might, and heritage in this industry is just meaning they’re gonna start pulling away a little bit. ’cause the philanthropics of the world don’t have the budget like open AI do or, or Google do even.

[00:12:00] Matt: And, Google has all that infrastructure and yeah. 

[00:12:03] Dara: Yeah. And do you think the, ’cause the other, I don’t, I guess this sentiment was maybe out there anyway, but it seems to have like the, you know, the, the backlash of chat GBT five has added fuel to the fire, but there’s a lot of talk around how, you know, those that, like the scaling laws that were in that paper back in 2020 or whenever it was, are now looking like they weren’t.

[00:12:25] Dara: And I, and I think actually I read earlier that even the authors of those papers didn’t say that this is a theoretical guarantee. They just said, you know, this is how it could scale. But it looks like now that’s tapering off. Yeah. And as talk around how maybe, you know, a lot of this has been hyped that it’s gonna just keep exploding exponentially.

[00:12:47] Dara: And actually it might level off now and it is gonna get into more of a, like fine tuning models, reducing hallucinations. Do, do you think, do you think that’s the case? Do you think it is gonna taper off now or. Will there be a chat GPT six or a, you know, Gemini, whatever, that’s, that’s gonna be that next big leap forward.

[00:13:09] Matt: Yeah. It’s so hard to tell because they have to put most of these models that they’re trained in to even test that law. Even though, like you say, it’s not, it’s not a law. It’s proving potentially it’s not a law even to test that. The amount of time and compute you’ve gotta put to try and push and see what the next load is, is like, it’s like a two year lifecycle.

[00:13:30] Matt: You’ve gotta build data centers, you’ve gotta pull together more computers than has been pulled together before. So it’s, it’s a tricky one to guess at. It does seem like recently a lot of the focus has been on that pre-training, post-training steps, the, the reasoning, and thinking, improvements and, and.

[00:13:54] Matt: Like, say reducing hallucinations and all that sort of stuff. Maybe that it’s a different technology that ultimately gets us past that marketing term. A GI. Mm-hmm. and LMS have a theoretical limit. dunno. I dunno. 

[00:14:09] Dara: No. Well, nobody, nobody does. That’s the, that’s kind of the point of it, isn’t it?

[00:14:12] Dara: It’s all, a lot of, a lot of the future around it is speculation. But one interesting thing, it was in an article that was shared internally at Measure Lab, but it was, written by Cal Newport who’s written a bunch of productivity books, but it was in the New York Times. Yeah. And he was kind of presenting, you know, a positive if things are tapering off, it gives regulation a chance to catch up.

[00:14:35] Dara: And that’s something maybe that some people would argue has kind of been lacking. And if it does slow down, maybe there’s two big advantages. One is it. Regulation, a chance to catch up. But the other one is that there’s a lot of potential that hasn’t been explored with what’s possible at the moment because everyone’s so caught up in what’s gonna be next.

[00:14:55] Dara: Actually, if it slows down, it might not be a bad thing ’cause it’ll give people a chance to kind of push what’s there at the moment to its potential, rather just always moving on to the next shiny thing. 

[00:15:05] Matt: Yeah, a hundred percent. I think there’s tons of incremental little gains just from the power. Just from the power that’s already there.

[00:15:13] Matt: You’ve gotta think there’s plans afoot for potentially other, for still pushing the LMS and training other types of models potentially. Because half of the world is building huge data centers which are costing tens and hundreds of billions of pounds. Which if it’s just, if it’s all based on that one paper about scaling laws seems absolutely insane.

[00:15:36] Matt: So maybe they know something we don’t. But yeah, a hundred percent Claude Claude Code bangs on it at bat a lot. But just what that can do now Yeah. It is unbelievable. Finding more uses and organizations being able to ingratiate it and get it as part of their day to day and increase their productivity.

[00:15:54] Matt: There’s, there’s, yeah, there’s so much more that comes out of right now. Yeah. We’ll see. Yeah. Time will tell. Yeah. So a little while ago as well, we mentioned this a couple of times about the AI mode for Google. Yeah. We were always theorizing like, how are they gonna do this? They need, how are they gonna monetize this?

[00:16:13] Matt: Blah, blah, blah, blah. And then just saw on LinkedIn recently, somebody shared some communication from Google that adverts are coming to AI mode. what they look like, I’m not exactly sure, but sort of adv, it’s relating to the original user request and the subsequent generated text by the LLM. So presumably a customer could ask, what are some nice shorts?

[00:16:40] Matt: What, what are some nice shorts to wear in southern Spain? And then it might. Back and then people can inject shorts and maybe, oh, have you thought about sandals? I dunno. I dunno where I’m going with this. You should work in marketing. I should, shouldn’t I? Yeah, I have underpants or shorts or socks. I seem to be focused on that area. Yes. But that seems to be coming soon, which I guess we kind of were pretty convinced was going to happen anyway. 

[00:17:09] Dara: Yeah. Any word on the ’cause I would imagine if as soon as that, as soon as that’s available, then there’s gonna be some analytics to go with it. So maybe this is gonna be the trailblazer for LLM analytics, potentially.

[00:17:21] Matt: Yeah. Or at least it’s gonna have to sit in. Yeah, it’s gonna have to sit in Google ads and things you’d assume. And, and, yeah. Yeah. Imagine it. Yeah. Yeah. 

[00:17:31] Dara: It’d be managed through Google Ads and then there’s gonna be some data in ga you would imagine. 

[00:17:36] Matt: Yeah. Particularly if these things get tighter, tighter, and tighter together, like we said before, where you could start having complete.

[00:17:43] Matt: Shopping journeys or e-commerce journeys within these chat interfaces or within Google AI mode. There’s gotta be something to, I dunno, to pass around and understand. I suppose you’re not really, if you’re appearing in searches, as an e-commerce company and people are converting happy days, it’s the content folks that maybe have more of a struggle I guess.

[00:18:03] Matt: But yeah, I assume it’s gonna drop into Google ads, GA four, those kinds of ecosystems. Then the citing new world of AI analytics begins or grows. Yeah. Okay. 

[00:18:14] Dara: We’ve got a bit of our own news, haven’t we this week as well. 

[00:18:18] Matt: Yeah. What 

[00:18:19] Dara: is 

[00:18:20] Matt: it? Oh, 

[00:18:22] Dara: such, such good as you’ve forgotten? 

[00:18:24] Matt: Yes. We are now officially Google Cloud Marketing Analytics specialized partners, which is really cool.

[00:18:34] Matt: So that essentially means can Google Cloud partners for a while, and to be a Google Cloud partner, you have to have a certain number of professional certified Google Cloud engineers and machine learning engineers on staff. but then the specialization, you show the work you’ve done with clients and your customers and you show the way you’ve got about it and the, the results and you go through all these various assessments and self checklists and it’s months and months of work and hours and hours of interviews with assessors and Google ’cause they wanna be as thorough as they can to sort of give you that badge and say that you are, you know, the creme de la creme in marketing analytics specialization in Google Cloud.

[00:19:17] Matt: And we are officially now creme de la creme Google’s words, I should say, for legal reasons. I don’t Google’s words,

[00:19:27] Matt: but yes, it’s exciting. A lot of hard work. On internally to get it all pulled together. working with clients in Google Cloud and growing that capability there has been really good. So excited to see what that brings moving forward. 

[00:19:39] Dara: Definitely. And a final piece of news, again, bit of a plug here, but we’re allowed to do that. It’s our podcast. So we’ve got our second webinar coming up on the 11th of September. It’s gonna be from, it’s gonna be live from 1:00 PM and it is around data form. So the title of it is, Meet Data Form, the Smart Solution to Fragile SQL Setups. Do you wanna give a bit of context? I’ve just read the title so you can actually say what it’s all about.

[00:20:07] Matt: Yeah, it’s, we’ve done a lot of, a lot of work with Data Form recently and we’ve built out some quite transformative changes in architecture for clients. We had Veronica on the podcast talking about one such project recently, which was, which was really good, but it’s kind of just. Introducing people to it a bit more and, and showing, you know, if you’re doing things in this way in the old way of stack SQL queries and things like this, it’s brittle and it can be prone to breakage and silent breakages and all these kinds of things.

[00:20:36] Matt: And how modernizing your infrastructure in the way you work a little bit with data form can really future proof you and allow better versioning, better collaboration, better data quality, all of these good things. We’ve gotta take you through what a typical stack looks like now, what data form is, and how we can help solve all of these, these various problems. So it’s gonna be, it’s gonna be really good. 

[00:20:57] Dara: So we’ll share a link in the show notes to the, to the signup. But again, yeah, it’s the 11th of September at 1:00 PM and it fits with our kind of theme, our recurring theme on the podcast around data quality and making sure you know that you’ve got good foundational data collection and that you don’t have these kinds of fragile systems.

[00:21:15] Dara: And what a segue this is. So joining us on the show today, we have Jose Perez from Tracking Plan, and this again, fits in with this recurring theme of making sure that you’ve got the most robust data collection in place and that you’re able to flag any issues before they cause too many problems downstream.

[00:21:36] Dara: So we have a really good chat, we’re big fans of Tracking Plan. We’re partners with them. Yeah. Do you wanna add anything else to that, Matthew? 

[00:21:44] Matt: No, no. I think you said it all. We’ve, it’s exactly in the kind of theme we’ve been talking about, about making sure everything is robust from the very beginning of collection, right the way through to getting it anywhere else and doing all those.

[00:21:56] Matt: Cool and exciting things with it, with AI or whatever else you’re doing with it. It’s a great conversation just to fit, fit with what we’ve been talking about for quite a while. Enjoy and enjoy. 

[00:22:07] Dara: Very warm, welcome Josee Perez to the Measure Pod. It’s great to have you. Welcome. 

[00:22:12] Josele: Thanks. It’s a pleasure to be here, big fan of the book podcast myself, so. 

[00:22:17] Dara: Even better. It’s always great to get a fan on. So if you’ve listened before, you’ll know how we start. It’s over to you to introduce yourself initially. Feel free to use as much time as you like. Give us a good, good bit of background. Maybe start with you individually and then, ’cause obviously we’re gonna talk a lot about tracking plans.

[00:22:35] Dara: So maybe a little bit about your bio, what’s your background and, and what’s got you to where you are today. 

[00:22:41] Josele: Sure. I think that my story is typical, and especially at, at the same time, this kind of person who had his first computer, in the nineties when I was nine. And all love with it started like coding at the age of 10 because my father taught me some, some of it and had not stopped doing things until now.

[00:23:07] Josele: So I obviously studied computer science. There was no option for me and I always knew that I wanted to build something like, like a company or so, so I got involved in that Kindle American startup mood garage thing. It’s things that have changed a lot. The time was just, you just put your head down, go things hard, launch it to the market in your modern or.

[00:23:32] Josele: Four cable connections as is. One came, so I had my, we built, our first company, launched with investors and all that stuff when I was 21 and it was Geomatic. There was a travel guide or electronic travel guide curated by the computer then. It was a good company. We sold a lot of travel guides at the time, but five days later we ended up with that endeavor and we started another company, the stem, the B2C space with the notifications app called Hooks.

[00:24:01] Josele: We went up, me and my partners to travel to the states to live a little bit of the American style. I’m from Spain by the way. That’s why it feels far away when you are young to do business at that level. And for that thing we went to, you know, COVID arrived. We, I was working like helping a company with the role of a C, chief product officer.

[00:24:28] Josele: COVID arrived and we felt that we had time. And we found a thing about, we did a lot of, of program analytics was part of my job. I love that. That you’ve had, I have used all the tools out there. I am decent, that engineer tool that I love in that field, but never have, never thought about how hard this was.

[00:24:49] Josele: It was to implement all that stuff when it’s not only like using mixed panel and amplitude and your, you know, your purchase event and your like monitoring products easier than when you are doing marketing. And we found that it was like a space with no tooling. It felt identical to this, but it felt like people were doing things in that space like in the nineties and we were talking about 2020, like then like the tooling, they were like throwing things to their websites and hoping for the best and relying on the partner a little.

[00:25:22] Josele: So they were nothing. We jump into this tracking plant sink right now, like five years ago, and we are here like working and, and rolling and all that stuff, so. 

[00:25:32] Dara: Brilliant. Well, listen, we’re gonna like, we’re gonna dive into that a lot, but I have a very important question to ask you first. What was the first computer 

[00:25:39] Josele: I had? My first computer, let’s, for my first PC was a 4 8 6 dx, two 50 megahertz. We wrote that when I was, I started coding and playing with, say, a Spectra video that was, like a clone of a spectrum that my father has. Oh man. At this basic stuff I fall in love. The first time I saw that, like drawing, I mean, at the time I was like, like just drawing in circles and some conversational stuff.

[00:26:06] Josele: It was really fun. I mean, I, it’s been, I’m four years old, right? Right now. It’s been a very long time and for the good or for the bad, I’m still in love with this thing that you touch the give and they are being on the screen. I think I’m so naive. It still impresses me like every day that things are working with cables and chips. You know, that’s what I like. 

[00:26:29] Dara: I remember my first Commodore 64 and it was, yeah, it was mind blowing. And when I think back to it now and having to wait for the tapes to load, it’s just these things. Things have moved on a little bit. Things have changed a lot. 

[00:26:42] Josele: Otherwise, I’m the same. I think it’s the same stuff. There’s just more, more code and code made by other people. That same stuff is like playing the game again, just by using your pen for rewinding the, the, the, the tape I was quite experienced with. I’m gonna blow. So 

[00:27:00] Dara: just before we kind of get more into track plan, just one of the questions, are you getting a chance still to, do you still code or in your role now?

[00:27:07] Dara: Do you not get much time to do that? 

[00:27:09] Josele: Yes. That’s my therapy. Sometimes it goes by my face. Sometimes I have a lot of marketing work and legal work and going to, talking to clients, part of my, my job, but just things first. Just for my mental health. I’d take like 20% time. I, I can call for 20% of the time.

[00:27:29] Josele: That being a CEO means probably it will be on Saturdays and Sundays doesn’t take that. But I do that a lot and I do it linked to a tracking plan. And something interesting from recent times is that the product, the product, got too big somehow. We all like convex tools, like huge lots of features, lot of processes.

[00:27:52] Josele: Like how would this framework be? At some point you go like, you cannot do anything in two hours. Jump in two hours. Just, just giving work to your mates to fix your stuff, taking your stuff back. And I think it’s an interesting topic too. With the rise of AI, AI coding, I’m back into that because of all that pipeline and all that time that you need to, now I can do things again.

[00:28:16] Josele: I can draft something, have an idea, let the thing do the first version and do the cleanup offer so that two hours are again productive. So I’m like, say, having a good month in terms of going Yeah. Versus yes. And I don’t if my team likes it or not, but. They don’t tell me that. They’ll never say, they’ll never say, they never say two sides of the same coin I get.

[00:28:42] Dara: Exactly, exactly. So listen, we know the tracking plan, we work with you guys, but for our listeners who maybe don’t know, do you want to give your kind of summary of what the tracking plan offers? 

[00:28:53] Josele: Absolutely. Let’s say tracking plan exists to mental life of the deed analyst issue by automatically monitoring their assets and the websites, their mobile apps, making sure that all the events you have, all the integrations you have, all the details you have are always working, telling you when they break and helping you to fix them.

[00:29:17] Josele: So the core value for us is that we give analysts that usually live in the g Google Tech money world. The integration world that we talked about before was really hard. We give them light. We show them what’s going on, we give them the tools that developers have for their work. But from our experience, analysts don’t have more than the idea. 

[00:29:42] Matt: Well, from our experience, we know this to be true, but I, I guess you will see this as well, but there’s so many silent issues that people don’t know about that are just, they’re, they’re probably in a boardroom reporting on something that they have no idea is nonsense because it’s just silently broken somewhere within that chain of events. That’s got to 

[00:30:00] Josele: Absolutely. That’s, I must, I’m still surprised. And, by the time I think we were with around a thousand and we jumped into there, and you just used the things from, from outside, say, okay, everything looks good. These people look really professional like this. They have this first call with, and then you jump in.

[00:30:18] Josele: You start tracking plans. When you start a tracking plan, you actually map everything, all your events, properties and values, et cetera. All that is relevant and I have not found any correlation, between how good is a brand? You go like a super brand. These people are investing millions of dollars in this platform.

[00:30:34] Josele: So the analytics should be amazing. No, it doesn’t work. And then you go to a small client and they have a perfect implementation, even if it’s complex. So there’s a lot of hidden problems. And I would say that’s the nature of what we are doing here. So you remember how the manuals talk about how you build analytics or how you build things that you start with, with what do you want to get?

[00:30:56] Josele: Then they find the metric, then build below that. These people are only looking at one number, but for building the number, there are a lot of inputs below and are hidden. So no one is looking at the, that you can define this as an equation that turns into the KPI you’re looking for. Any factor of the equation can be wrong and you don’t notice because it’s just a.

[00:31:17] Josele: 10%. So that’s what you consist see and that’s what with trucking plan, we have learned to make our clients like solve that from the root cause so they can actually trust their data and their conversions and privacy stuff is also related and all this kind of stuff that it’s hidden because the platform don’t, don’t show you that.

[00:31:38] Josele: They don’t tell you that you are passing wrong information. They just give you the apps or they just give you the APIs you’re talking about. 

[00:31:45] Matt: What levels do you work at then? So what parts of that? So I’m thinking of a, I dunno, a typical marketing analytics stack where you’ve got a producer of the data, I suppose the website or the application, and then you’ve got some abstractions off top of that with GTM and creating some downstream additional metrics.

[00:32:03] Matt: You then got stuff in GA four where you probably. Building out custom audiences or whatever and then maybe you’re sending it off to BigQuery, so like where does tracking plan coming across that spectrum of different technologies? 

[00:32:14] Josele: We found that that is the core problem, and we studied this a lot because we were probably able to pick up at any level if not, the integrations are not the problem and we found that this was a collection problem that this probably almost solved.

[00:32:29] Josele: You are thinking about what’s downstream. Okay. It’s always upstream where the problems happen. Where all the mess is, where, where there are a lot of players doing that. In a normal website, you have the medium size e-commerce, for example, you have one or two agencies doing performance that one, one or two agencies doing analytics, some of them doing just data.

[00:32:51] Josele: Then the CRM on the actual like internal marketing people is also playing with that website code, the Google Tech manager. And then you have the coders and all and the internal coders of the application. That’s where all the mess happens and all that happens in the connection phase. And that’s where things broke.

[00:33:07] Josele: You know, I know I want to sound typical, you’re not very reaching or reaching out the topic. I’ve seen a lot of companies having that DBT solution like this come pretty common. We fix things in DT Knowledge, you cannot fix things in DV t because it’s too late. Deal with that, the data’s already collected. So we focus on monitoring the collection.

[00:33:26] Josele: So we leave talking to them. Technically we live in the websites we run with each time by using navigates to a website. We are there listening to all the network requests that show the actual production of the website. And that’s what we collect. We, and that’s what we monitor. And when we on, on top of what, we build things so that anything before collection could be that layer.

[00:33:49] Josele: We see that because that’s before the, that collection, DTM transformations, we see that because it’s before that collection. It’s a network request. It’s what we look at and what actually is, what should be wrong because you cannot modify things inside Google Analytics. That’s, you can modify the data there, what we might data wise and statistically.

[00:34:07] Dara: So who would own it then? Because if it’s, this actually has reminded me of a conversation we had with another guest recently on this kind of shift, left paradigm of going back to the source and, and making sure that everything is kind of tracked correctly at the source rather than waiting on things on something and fixing them.

[00:34:24] Dara: So who kind of, with tracking plan, who do you tend to aim it at? Like who within a company should be, should it be a collaborative effort or should it sit with a particular team or like what’s, what’s your kind of general view on that? 

[00:34:36] Josele: I think this is, we won’t be able to change the dynamics of the politics of the company.

[00:34:42] Josele: So you have, the marketing team will be, will still want to do their integrations and we will, and they will still want to. Mess with rules that manager they own or the good one for the bad, developers will be still needed to do the data layer pushes at the very least, and the data team will still be asking, the product team will still be asking for, I need information about where they used to clicks and we need a pixel or what we set, is that all of them or do you have a way, or the company itself had a way to see what’s actually going on and we’re responsible for that.

[00:35:16] Josele: I think one should be responsible for its part. It’s pretty common that when we work with, company, we have different stakeholders that the marketing team monitoring their Google ads and meta TikTok pixels and the conversion and the campaigns and it builds things, monitoring amplitude and with analytics and the data team is monitoring their own collector.

[00:35:35] Josele: So that’s a shared thing. I love to have one person in the company, so some, some of them have. Like, the figure of the, sorry, the role of implementing something like a marketing integrator. They have different names. That’s something that is in the suite that’s in charge of that. But usually there’s always alien integrations out there, you see done a lot.

[00:35:59] Josele: I mean, I’m pretty sure that when you jump into a client, you see that all the time and there’s no solution for that. We just help them to communicate the things that we’re tracking them. We have the truth of what’s going on and they can mention each other within the platform. They can share the problems, they can share the specifications.

[00:36:18] Josele: So that’s, in that helping with him. Now for me, that is like asking who owns GitHub in another company. Okay. Tech team. This is a short answer, but no, there’s a lot of people. Messing with it, 

[00:36:31] Matt: I suppose to a certain extent, the very fact that it exists, like for most of these teams, the marketing team or the, the engineering team or whoever, they, they want a certain level of autonomy, don’t they?

[00:36:41] Matt: To be able to go and to be able to go and set their tracking up in GTM and not have to go and beg hat in hand to the dev team to go and add in their tracking pixel or whatever. So having something like a tracking plan in place to be able to trust and see where errors may occur is almost freeing people up to, to own their own stuff a little bit and spread that responsibility out a little bit.

[00:37:03] Josele: Totally agree that we, without observability, if you cannot block what people are doing, let’s call it away, you are, you won’t be able to collaborate. So what we are trying to give is light to that. Is that like, going to use again the analogy with development. Imagine like all these people that are working in code, imagine like people pushing everything to production through FTP.

[00:37:27] Josele: Without asking all the time, you will say no. At some point you will say, no, no, only Jake can push code to production, only, and then do happen. Test the mechanism. One of them is always observability. It’s just to see what happened. See how things start, how things change. Seeing how things are right now, how they went, how they were yesterday is one of the basics, and that’s what we are trying to build here.

[00:37:52] Josele: So yes, it’s enabling people to be more comfortable. Like in like when I’ve seen it a lot is that, and I hope this is changing with our help and others, it’s like a digital analyst within companies being a little bit cornered. Like, where is the ugly duck? I dunno, I’m saying that in, in, in English, it is wrong.

[00:38:11] Josele: Like, like, yeah, they ask for things, but the developers don’t ask, don’t answer. You open a ticket for fixing them, you fix your tracking, but maybe at the end of a sprint or. It’ll be done. And part of it is just because they are not in the workflow, you need to be in the workflow. You are outta the workflow.

[00:38:31] Josele: I’m sorry, we don’t play that thing. So we try to build a new workflow where everyone sees what’s going on and the key thing is when developers in the case of tracking implant started believing in tracking implant saying when the data analyst can say, Hey, we have this problem. It’s not a hallucination.

[00:38:49] Josele: I had the data here that you’ll release a thing and it’s broken and you can see the graph of how you broke this in this release. Then things change. Then your analyst now is empowered and believes and in the end he can grow within the organization. We can press it just by saying and doing the navigation.

[00:39:06] Josele: And I remember that the purchase had 10 parameters, now it has nine. That’s not there. Something a developer is going to believe. 

[00:39:15] Matt: Even be able to point to a number and say like, I know that this thing is going wrong here. Cost X, like having the empowerment to draw that line is gonna be a much more powerful message.

[00:39:26] Josele: Yeah, I mean, you, you know, I mean, that’s a common topic. I’ve heard of you talking about this in the podcast before, but I’m not checking like the developer, like the developer part has lost, because they are learning and they learn to value this. But I think our messages were like a digital analyst.

[00:39:42] Josele: Messages in the past were a little bit weak because they don’t have an impact with it. They don’t have the looks like this, this is broken and it’s always broken. So let’s not tell me that you are now letting you invest 50 K per month in Google Analytics, sorry, in Google Ads, and you’re wasting 15 K.

[00:40:06] Josele: Let’s talk. But before you haven’t even demonstrated that. And platforms don’t help, I mean. Facebook tells you everything is all right, Google that tells you everything is all right. That’s, their job is not to tell you that you are doing wrong. They want to pull money. They’re doing a great job. They do a great job, like full money in this star, in this thing and turn out the thing.

[00:40:25] Josele: And I understand them. I would be doing the same if I were, if I were them. Like a, 

[00:40:30] Dara: does it only look for errors and broken elements or does it actually look at the data as well and detect anomalies and say, you know, for example, you know, usually you have this much traffic and it dropped by 50% or something like that.

[00:40:44] Josele: Absolutely. That, that, that’s the second thing. While we actually do excel, the Volkswagen plan is that we are able to also monitor almost everything, meaning that, let me give you some examples. If one event, like your purchase event drops in the page X, we will tell you immediately if the parameter value of your purchase event.

[00:41:08] Josele: Yesterday was one decimal. And now with two decimals, we will tell you if you have page category, and that was the values, A, B, C, D error. Now you have an e value. We will tell you, so we know how to monitor everything. We found that the rules are liquid. like even if I ask the best team in the world what the specifications of their things are, they will only be right 90% of the time.

[00:41:38] Josele: And those are the things. But we found that monitoring the changes, monitoring what, what was yesterday and what today’s notice is the best way to find things. Yes, we will, we will be really deep into the data, deeper into everything and trying to make everything automatic so you don’t have to tell us the rules because somehow we will, we may not believe you or you may not, may not, may not know the right rule.

[00:42:01] Josele: The second thing is too much effort. So we will, when we’re working with agencies like yours, we know that you will have mass time. You make, for example, like mostly plug and play, you install it and one week later we’ll be telling you if you are not passing the currency with the add to cart event, if that is a problem, 

[00:42:21] Dara: let’s follow the, I know this isn’t really a, you know, a fair comparison, but I think back to like when you try and set alerts in Google Analytics and they would become just noise after a while because it would tell you, you know, your traffic’s increased by 20% or dropped or whatever, or would do it every Monday because your weekends were quieter or something like that.

[00:42:39] Dara: And you’d get to the point where you’d just ignore the emails rather than, you know, do anything about it, which is a confession of extreme laziness. But is there any risk of that where if, you know, we’re almost like people miss the message because there becomes too much noise, too many different alerts about too many different changes.

[00:42:58] Josele: Absolutely. I mean, that’s. Typical noise, expressor signal problem. We had some experience with that. My previous company was about notifications. So we learned a lot about how this works and how the mental model of the users and how they react to a notification. And we learned a lot. And we were with the, at the time I had the, opportunity to work with the best design agencies thinking about how to solve this problem.

[00:43:21] Josele: And in the end it turned into a notification only when you are sure that it’s relevant. And luckily what we are doing, and I don’t know why Google isn’t doing that in their alerts. That’s the thing by conveying the different medics. So when we tell you that your participant is down, we are not just telling you that your participant is down, but telling you that your participant is down, all the traffic you have on your page, usually for the number of users you have right now on your page.

[00:43:49] Josele: So it’s like making things a little bit more complex so it’s not a 20% job compared to yesterday. That will be really easy. It’s far more complex. We found that we don’t have so many false positives and that’s around maths and, and having other information, how the website works. It’s hard. I think we are able to do that because we are adapting this exactly to this problem.

[00:44:13] Josele: If you apply the best mathematical model, just that model to the participant traffic, you will have many more false positives than we do. It’s about understanding what a purchase, not only, and it’s not that just a time series of on on time matter Compass users. It’s a session, it’s a user navigating.

[00:44:34] Josele: When you put that into the equation, then you cut down the noise. 

[00:44:38] Matt: I, I suppose Google Analytics four is a, I mean, it’s a very generic tool, isn’t it? It, it’s, it’s made to spread as wide to, to go across many different vectors and sorry verticals as it can. To be as appealing to as many people as possible.

[00:44:53] Matt: So their ability to track very specific patterns of traffic for very specific sort of e-commerce use cases and, and different business is gonna be limited compared to a specialist tool that’s aimed at just that. 

[00:45:06] Josele: Absolutely. I mean, that’s an innovative dilemma all the time. They have it. I mean, my post company was around Gmail and we made our lives like making Gmail a little bit better.

[00:45:18] Josele: We had notifications when someone read your email. That’s the kind of stuff, or made a big basis and they are not going to do it. Google is not going to do it because they will, like 20% of the user base will get met and they don’t, they don’t take risks. And GA four is absolutely, you are, you are totally right.

[00:45:37] Josele: Too generic. It serves a small flower shop, in the corner. huge clients like courting less or like big brands like that work with us. So that’s a, there needs to be a more specialized thing and that’s why when, when we, they, they open BigQuery, like everyone went crazy. Okay, now we can do SQL queries from it.

[00:45:57] Josele: Like what, 2024. And you are so happy because you can actually query your data. That’s what I told about the v underserved in this industry. So yes, it’s too generic. They’re trying to do the same thing and that’s what, but they have to stay, they cannot tell the flower shop owner that they have some data privacy risk.

[00:46:20] Josele: What? Like, they don’t care about that. They just want to count the number of visitors they have on the website. That’s a combined product delivery. 

[00:46:27] Matt: And the irony is half the time when you’re in ga, when you’ve got your data out into Google, into BigQuery, you’re trying to figure out what the hell it means.

[00:46:35] Matt: Like what, what is, that’s why saying this Yeah. 

[00:46:39] Josele: Decided to, to cite a sword. That’s the thing is that like, they, they cannot build that professional product and they’re free. That’s, that’s the part that is, and even big companies are using it because it’s free or is free and they have the market and they have the, you as a professional, you can actually tell your clients you start this because it’s free.

[00:46:58] Josele: And we started with that so you know how to use it then. And it’s incredible what you do with Lucas Studio. And so we can be talking about that for a little while. Lucas Studio and Google Analytics and Database can do great stuff. But you are fighting with the wrong ui, the platform because of a big change.

[00:47:20] Josele: Do you remember that last couple of months or half a year ago, they added new graphs in the studio. They looked so slow to adopt because if they changed one old graph, it did change the color of one of the old graphs. Some of their big legacy clients will call the Google CEO and quit and go to Adobe with part of being new in industries that we can innovate, we can, things like pretty sure that in 15 years we won’t be able to change anything from our ui, but luckily we can do it now.

[00:47:54] Josele: So let’s, let’s take that advantage. Take advantage, 

[00:47:57] Dara: yeah. Yes. Are you, are you surprised though, I mean, it makes sense with GA four for the kind of reasons you’ve both said, like about it being a little bit too generic and it’s not really what Google are trying to do. Are you more surprised though that GTM hasn’t got, maybe not to the full extent of tracking plan?

[00:48:12] Dara: Why do you think GTM doesn’t have more of a kind of observability in a Tool built into it? Is it just again, that they’re just going really broad and they don’t really care and they’re just happy to have other companies out there doing that? Or is there some other reason, do you think? 

[00:48:27] Josele: I think Mark said it is like the percentage of users that are interested in this, it’s not as big. I love that they will never want me, I will be rich or more rich if I had everyone. But right now they’re adding some kind of monitoring that looks like you have installed your pixel correctly. That’s important. And we be, we have the trying to do the, so they, we layer out Y Combinator as an accelerator.

[00:48:52] Josele: We went through like the 80-20 solution. So they have like, by telling these three things, your constant mode is, looks wrong and something that, that’s quite a thing. Professionals need the rest. The other thing, Dara, is that it’s expensive. Google I is free. GDM is free. I mean, what we do, it’s a lot of infrastructure not probably cheap for the value.

[00:49:13] Josele: You still need to log all the information you need to give a lot of value. And then again, if you’re free tool, you’re going to have to do that for every client, even the ones that don’t use it or they don’t need it. So you select log requests and run reports every day and, and sending them, it’s optional.

[00:49:31] Josele: So that’s a, I think that they are not in the, not the market. 

[00:49:36] Dara: I think you’re right, by the way. Yeah. What’s the main blocker, if somebody’s interested or potentially interested, what do you find is the most common reason why people either wouldn’t? Is it an attitude of, oh, we’ll be fine, we can do that ourselves.

[00:49:48] Dara: We don’t need a tool for that? Or is it more around, you know, a lack of general awareness of needing, what’s the kind of typical, what do you think is stopping more people from using something like a tracking plan? 

[00:50:01] Josele: One of them for sure, is that they actually don’t care about this problem. That’s, and it’s easy to say you care.

[00:50:09] Josele: It’s like I asked you, Hey, do you care about yours. What do your kids wear? Yes, of course. I care about that. Are you up to go to the shop every two weeks to buy them new clothes? That’s not enough. So they are, they don’t actually care. Let me tell you. Like for every customer that actually reaches us, we convince them when we jump into them, when we, we, we use LinkedIn, we do outreach, and we write to them, only 20% actually go farther.

[00:50:38] Josele: And instead a ster free POC and so on. And you say, and you say, Hey, you’re a big company. You have this data. I can actually, we can actually scan your website from outside and say, you have this problem and this problem. And of course they tell you, yes, it’s really interesting. You have this problem. We need to fix it.

[00:50:55] Josele: We need to fix it now. And you wait and you wait and they don’t come back and they don’t fix it. The only reason can be that they don’t actually care about this problem. Right. But in stem, I mean, we are not the most important part of the business. We work, we work with huge businesses and they’re losing like millions of dollars in bad tracking on bad integrations.

[00:51:14] Josele: That’s real money in terms of performance. But if you’re making 1 billion per year, that millions is not what you want to optimize right now. So that’s a focus problem, but they don’t care about the problem. That’s something that we have to live with. That’s, and I’m sure that you find this too in, in, with your clients.

[00:51:30] Josele: The second thing would be they have a data team, the data department in the company that thinks they can do trackings and plan. You, you nailed that. Is that No, we can do that. it’s not as complex. We can create, it’s just alerting as we told before. And we know better. They come back like one year later or two years later, it happened.

[00:51:49] Josele: I mean, we have five years in the market. We have seen that. And big companies are pretty common that there’s always, I mean, they have the budget for a tracking plan or somewhere in the company that for the money tracking plan, maybe they can hire someone. Build that, and then one person is promoted. Now it’s not a, but they have a team of developers building that established.

[00:52:07] Josele: It’s actually super cool. I recommend everyone do it. They try to do it, it’s a fun product, a lot of data infrastructure, algorithms, fun product. They want to build it for fun, for politics, for growing aggression. So that’s it, that’s it. So I think that the two main, main blockers are, are the way, let me add one, is that the person who’s in charge of this.

[00:52:32] Josele: Is lower than it should be in the organization, but it has no access to budget. It has no broadband. They have the small opportunity to tell their bosses once a month we might try this thing. And even if the two would make their lives easier, no one is taking care of their happiness or things.

[00:52:52] Josele: That’s the third thing. 

[00:52:53] Dara: I guess the irony being that it could be actually costing a lot more than they realize if there’s, you know, broken tracking left, right and center, which there usually is. 

[00:53:02] Josele: That’s why we try to build like their oil use cases. That’s super obvious. Like, your participant was down in Dicto for two weeks.

[00:53:12] Josele: You invested $3,000 this month. That’s, that’s absolute. Or it’s 5% of it. That’s already, that’s already at the table. I mean, everyone knows that data quality exists. It’s not the tracking plan and fixing it, it’s just food efforts of fixing it and they’re not fixing it before. Why are they going to fix it?

[00:53:30] Josele: Because I exist. They have to be of some interest in fixing it. 

[00:53:34] Matt: We’ve found that it’s just sort of defining the problem upfront rather than going to them with the solution. Like if you go, if you just go in and say, Hey, you need a tracking plan, and they’re like, why? You know, like, well, because of this, this, and this.

[00:53:44] Matt: Then half the time they’re like, I don’t wanna pay for anything else. But like to, to use sort of data engineering terminology for a second, like getting them to define their recovery point objective and their recovery time objective and say like, how long are you willing to accept? No, your tracking being completely broken and not working, and what time are you willing to accept?

[00:54:04] Matt: And most of the time you answer that question with zero. I don’t want that to happen at all. And then you go, well, in which case you need to go over here and, and fix things and make sure it doesn’t happen. ’cause until you get them to define it themselves, they don’t really see it as an, as an issue. 

[00:54:18] Josele: I define it.

[00:54:18] Josele: Well, you’re gonna be, if you had to explain the problem to your clients, you probably won’t be sorry. That’s true. Well, because you, you might be totally right. Won’t be at the bonus. It won’t be the same conversation, right? So, hey, you have a problem. You’re losing money. Have to interest you. They have to be interested in that.

[00:54:34] Josele: They have to believe that they have to research the problem. And then you do like great selling organizations are able to create their own demand like by voting an idea on your head. And we are, that is the market is evolving in that direction. That’s really good for us. Not evolving data quality as a topic.

[00:54:51] Josele: That helped us a lot, but we cannot explain the huge transactional website that data quality is a thing. If they don’t know that, forget about that, that deal. 

[00:55:00] Matt: I mean the reason we say that and the reason we frame it in that way is ’cause we believe it. Like it kind of leads me on to what was gonna be my next question.

[00:55:08] Matt: I think this is getting more important to my mind. Anyway, we’ve got all these new technologies and I dunno, four or five episodes ago, Dara and I went through the next 25 when they were showing all of these new automations. They were talking into BigQuery and, oh, this will just go off and roll your data and do this thing and this will go off and do this thing.

[00:55:25] Matt: What we both commented on was, you know, what if the data’s not right and it’s just going off and running and doing all these things. So it feels sort of on the right hand side, to use the shifting left analogy, a lot’s getting automated and a lot more accessible. So it’s gonna be a hell of a lot easier for people to just produce crap from crap data.

[00:55:46] Matt: you wanna shore this other side up feels like more than ever to make sure that as people are empowered within your company to go and produce stuff on the fly really quickly, like you said before at the top of the call, being able to quickly code things like that. Yeah. But doing it on good data, so, so yeah.

[00:56:03] Josele: Now you get asked this a lot for investors. I mean, we are, we have investors, we are this kind of startup company and everyone calls, okay now the way AI will be destroyed. And it’s like a. Using ai, but could be like any change in the, it could be like the management, no, everything’s going to be easier and then they won’t need you.

[00:56:23] Josele: Now. G four is easier to implement than yours, they won’t need you. Now they have Adobe, they have this kind of, they will need you exactly the opposite. You give someone a, like a trouble and 50 kilograms of cement, to build a wall and they will build a wall. Now you give them a bulldozer. What they will do, they will try to build a fucking building.

[00:56:45] Josele: So that’s human nature. And, buildings have more flow than walls. And I don’t see that actually we are still, yeah, we are already seeing like this AI giving us clients in terms of we are using that technology in the front end, that now it’s collecting more data than its following domain. That we have this tower on the website that’s doing magic here, but pushing a lot of automations.

[00:57:10] Josele: The more automations. The more things can break this, some counter it. It’s like you automate things for not, they’re broken, it’s the opposite. Automations are coded and are left. Another example would be when you, like you have a CRM. Okay, do you manually, everyone that jumps in the website, you read the logs, you check manually, ip, you put the country and you have a simple form that goes to your observer and you manually update, update, update your CRM.

[00:57:40] Josele: at work slow, it’s manual. Now you have a tool that automatically has a pixel connected to your Google Analytics and wherever. There are two side effects. The good thing is that you don’t have to do it manually. The bad one is, no one will be checking every interest. There’s no, there’s no validation.

[00:57:55] Josele: So more automations, more problems, more work for us. 

[00:57:58] Matt: Have you seen that with the rise of say, AI and are companies realizing that trend, that potentially this is going to lead to really massive buildings being built without a lift in it to torture the analogy or touch, but are people coming to you with that problem?

[00:58:13] Matt: And, and if so, what kind of businesses are larger businesses that have figured that out? Are they smaller businesses or 

[00:58:19] Josele: I’m seeing more brave people. That’s good. I love that meaning, let’s tell you that, that may be clear. I think that’s how things should be done with the right tooling. More brave people.

[00:58:29] Josele: Like I’m seeing recently, like for example, more custom collectors and they are built. I know that, like for example, handling snowplow schemas was a pain in the ass until the GBT was able to build a schema for you. We are seeing a lot of this, medicine. It’s a little bit soon for most of my clients.

[00:58:51] Josele: I mean, they are as usual, I mean, established companies are by definition laggards in terms of including new things. When we talk about American clients, I’m seeing more, more activity there and also with American agencies, like they’re being really proactive with these, using lms like modern, like many of them for example, are using JSO import export and gig tech manager using like TTM for creating the events like they are the T TM management.

[00:59:19] Josele: They’re doing crazy stuff. Right now. I think it’s inevitable. That’s my point of view. I dunno if it’s, it’s my observations or my intuition here was talking. 

[00:59:29] Dara: I’m skeptical and I wonder if you are seeing people, ’cause you call them brave and I wonder if it’s like people who do see this and do see that the quality of the data being collected is only gonna become more important with ai.

[00:59:43] Dara: And actually maybe a lot of people who aren’t gonna realize that and they’re gonna go ahead and they’re gonna take advantage of all these automations and all of these agents and they’re gonna realize the hard way that they should have cared about this. And they’ll learn it eventually and then come back.

[00:59:59] Dara: Because I am skeptical, I think people are gonna go for it, ’cause it’s something you mentioned earlier as well where you said, you know, like people might say they care, but there’s a difference between saying you care and actually really caring. And I think there’s probably a lot of people out there that are gonna think, oh, I know my data’s a little bit.

[01:00:15] Dara: Dodgy, but it’ll be fine because there’s all these AI tools now and I can just use them and everything will be okay. 

[01:00:20] Josele: That’s, you are absolutely right with this. I mean, I’m skeptical about data quality. I mean, let’s say I’m optimistic in the sense that I will have more business because things, things bridge a lot.

[01:00:31] Josele: Totally. Sorry, I’m the bad guy here. Look, no one builds things. I mean, you have to be a little bit insane to build things for the sake of data quality. You will do things because they’re cool. You will do things because they are, you save a lot of time and you will do things because they shine. But so when you build new things, when you get, when you can do that bulldozer, you will try to build a building.

[01:00:53] Josele: Your f my first building, everything will be for a long while, will be the, my first whatever, first, whatever time. We don’t want to play with this. We actually don’t know a lot about this in this industry. I’m tired of jumping into clients, tech managers or checking time monitors and saying things like, okay, every junior person of every agency in this country.

[01:01:14] Josele: Code in the CDF. I’ve seen things like that. S like people play and likely sometimes kind of a junior industry. And you, if you get a junior in terms of the person who’s, I guess, recent grad doing things in, in this industries, likely pretty common or, I mean, it’s a good size for sure, and if you give like, matching gun to a junior for a or recent scout, you will have a problem and that’s, everyone will, will be, that’s interesting from a startup standpoint.

[01:01:46] Josele: That’s why in the startup world things happen. Beautiful. If they break, no one cares. They fail and we don’t know about them. But we were talking about the data that a multimillion or multibillion dollar company cannot be as brave or responsible. 

[01:02:04] Dara: So we were talking about AI creating an opportunity, which I think we all kind of agree.

[01:02:08] Dara: It definitely will. But are you a tracking planner? Are you using any of the AI tech within, are you now or do you plan to incorporate any into it? ’cause like you said earlier, you know, you’re using kind of maths for the anomaly detection and setting correct thresholds that are useful and AI I guess is gonna be quite powerful for that.

[01:02:26] Dara: So how does it fit into your own kind of development plans? 

[01:02:30] Josele: Yes, absolutely. We are playing, I mean, we are like small things like processes, like handling how we handle problems. But what we are doing now is mostly like how we notify our success team. How we detect something is that the client needs, needs, attention, how we manage our tickets, everything that’s an ai.

[01:02:48] Josele: And so mostly internal things, not, I mean. We use things like open AI or lenses. Understand it’s not that we will lose our lens, like no one does that right now. but we are doing that for internal automation. Then in the product, what we are working right now is helping with the, one of the core features of the target plan is root cause analysis.

[01:03:06] Josele: Is that when a problem happens, what is the cost of the problem? So we have a lot of data. Let’s say in one example, your active event is missing a parameter and is missing the currency sometimes, and we have to find the customer to tell you how to fix it. So how we right now what we are doing, some, some ML techniques that are finding with all the variables that when this happens, it also happens that the data leader was at this value or the page of the content was that, or we are building on top of this, is just being able to actually humanize that reports.

[01:03:39] Josele: To tell you and to detect when it’s relevant, when it’s actually the cost or not to be able to tell you, Hey, there’s a problem. We already know the costs. Go to that problem. We will admire fixing it instead of giving you just the road. There’s a lot of effort being done right now and into this. And also more, more basic stuff, like automatic descriptions of the events, like automatic reporting.

[01:04:04] Josele: I’m not a strong believer in creating reports by text. I feel it’s like a little bit la as that’s not kind of reason, like a little bit key. The idea of a, I don’t want to click Google Analytics group by country. when the country is Germany. I prefer to write the text, give me Google Analytics. I don’t, for me it doesn’t make so much sense.

[01:04:29] Josele: But finally, we are playing with ways to interact with tracking plans with natural language. Be, there are some parts that produce, little bit too much information and we summarize things. We’ll be able to tell you in a more friendly way what’s going on on your analytics, what’s going on, on your tracking. 

[01:04:47] Dara: Are there any other, I appreciate, you’re obviously not gonna share, you know, your confidential roadmap, but is there anything else that’s on the horizon that you can share?

[01:04:54] Dara: Any part of the future plans or anything you, you know, that you, you can share? 

[01:04:59] Josele: Yes. We’ve done a lot of work like growing that product, like what we see, what we don’t like to think, like putting light to areas. We were not able to do it like one year ago. We were looking for the right way to data. For example, we weren’t collecting right now, we see that we have a lot of features like finding sessions, but you can, hey, you can go to a brand.

[01:05:18] Josele: It’s like, give me a session that has a purchase, an add to cart in Germany and they end up here. So the user don’t have to manually reproduce anything. They’ll go to tracking plan with and see everything that you don’t have to. Use the com extension anymore. That’s a way, but we thought we lost some basics in the, in the minute.

[01:05:36] Josele: For example, we are really seeing, I hope in early September, pretty confidential, because the team will tell me why do I give dates? I should not give dates, but my, that’s my initial estimation. Like finally, like real time alerts for everything. Like they want to tell you when, not at the end of the day, not when it is obviously a problem, but letting you set.

[01:05:57] Josele: I want to be notified that the purchase event goes down in this country at any time. Notify me within one hour. The second thing we are working on, and this will be really soon, is having a really good API, I know that’s the basics. There’s things that we were not doing, like growing on top of what we asked.

[01:06:18] Josele: We found that, like a really good API where our clients can ask everything and do their own integrations. We are working with more major agencies right now, like big brands, understand. You have right now the team to build whatever you want. I don’t need you to stick with our ui. So we have a looker connector, we have a Google specific connector.

[01:06:40] Josele: We have a power BI in the making. Something like that. Like that. We don’t want to lead that. We don’t want to create the report for you. So that’s opening all your data to you and you do whatever you want for this. And we found that for some, especially with the agencies, that’s a must. And we had that. I guess that, like for me, it is like super basic stuff.

[01:06:59] Josele: I mean, the kind of person who puts an API on every thing, the first thing I do, but you know, we forgot about that. I, that one, that’s the key thing. I think that, not fancy, but they are complex. Real time will change a lot of the probe, I think, and I think we’ll be really impactful. I mean, we, we see that a lot in conversations recently with clients, 

[01:07:20] Matt: so we’ll look forward that to be de definitely delivered in September 

[01:07:23] Josele: about that.

[01:07:24] Josele: Sorry about that team. 

[01:07:25] Matt: That’s, I’m joking, 

[01:07:26] Josele: I’m joking. 

[01:07:27] Matt: We 

[01:07:27] Dara: we’re, we’re gonna be, we’re gonna be really popular with your team now, after asking that question, just blame us. Say we forced you to tell us. Yeah. 

[01:07:36] Matt: I was gonna ask a bit of a wrap up question ’cause I’m just, that I asked to think, I asked everybody.

[01:07:41] Matt: I might have missed a couple of people, but it’s, it’s, it’s essentially a crystal ball, like everything that’s going on, and we’ve touched on a couple of these things already for our conversation, but be it with ai, be it with tracking, be it with cookies, be it with whatever within the industry. If you had to look into your crystal ball and say, you know, what’s coming, what’s the next big thing?

[01:07:58] Matt: What’s, what’s over the horizon? What, what would you say? 

[01:08:00] Josele: I think self-service or almost everything will be a thing. I’m pretty sure this will, this is a trend that will impact how people expect to interact with products. So, and we talk about our industry I that the, like the, like the managers don’t want reports anymore.

[01:08:20] Josele: They want to go to get it. Or again, too brave from the side. I think I would always prefer that you prefer the report from me, not me doing the report, but I expect like the manager wanting to, wanting to put the question in, interface and get the answer. That’s really tricky. That’s a huge problem. It won’t work.

[01:08:38] Josele: That’s my prediction too. That’s it. If I don’t worry, it will cause more problems. That’s and it will be unavoidable that they will ask that way for things. If we think about how the integration industry works and how we like doing that, all analytics or all pixels or attribution and all that stuff, I also see an important shift in how websites are built and that will have a huge, huge impact in this business.

[01:09:04] Josele: Like how do you define it? Search. Search searches are changing. For example, they’re going to navigate, I don’t know, but my bias that Google likes searching is going down and talking with the computer will go up, voicing interfaces and all that stuff. So think about how that changes in terms of attribution, how that changes in terms of what the pixel means, what’s a conversion, what’s a thing.

[01:09:30] Josele: Also like a bundling of solutions in terms of, of that unbundling of different things like how this bundle, how this one player winning this stuff. And there will be, there’s always one big one player winning this career and putting in their own rules. There will be a lot of changes there. And with that there will be some like massive simplification.

[01:09:47] Josele: There should be some kind of massive simplification for these because it will be, we are still dealing with Google tech manager and tax concept and all that stuff for a mostly conversational interface doesn’t make sense. I mean, I think that a new generation of smart people will jump in. And they find there are rules.

[01:10:06] Josele: Last week you were talking with a DE from the States. I mean they are like kinder, redefining this Google Ads integration shouldn’t be that hard. You know, that doesn’t make sense that it was a service and integration doing that little bit of CDPL dimension happening there. I think that some kind of amplification will need to happen and that’s more a desire than a forecast, I guess should happen.

[01:10:34] Josele: And for COVID, I’m not, I’m not going to say anything that you haven’t heard before that looks like an immense, like huge shift in how people think. 

[01:10:43] Dara: Okay, Jose, thank you again for joining us. It’s been a really, really great conversation. Appreciate your time. So thank you. Thank you. That’s it for this week’s episode of The Measure Pods.

[01:10:53] Dara: We hope you enjoyed it and picked up something useful along the way. If you haven’t already, make sure to subscribe on whatever platform you’re listening on so you don’t miss future episodes. 

[01:11:02] Matt: And if you’re enjoying the show, we’d really appreciate it if you left us a quick review. It really helps more people discover the pod and keeps us motivated to bring back more. So thanks for listening and we’ll catch you next time.

• Spotify
• YouTube
• RSS Feed