Pull Up a Chair: A Conversation with Marco Giordano

Speaker A: Foreign. So good morning, good evening or good afternoon wherever you are. Uh, thank God. We are on the Analytics Friday podcast and today we have an awesome guest, Marco Giordano. Marco is uh, going to introduce yourself. It's a great analyst. Marco, glad to meet you here. Thank you. Uh, to have accept our invitation to our podcast. Please introduce uh, yourself to our audience.

Speaker B: No problem. And actually thank you Orge. So I'm Marco, I'm the founder of Statistics. I mean it's a brand where I teach people how to do web analytics or analyzing this type of data. Ah a little bit differently and going beyond for example in terms of workflows, in terms of processes and the business impact. So my specialty is mostly well of course the Google Data, but also BigQuery data form and now with the recent changes, LLMs as well. Because my background before was in uh, I mean it still kind of is in SEO and content. So that's how I started and then I pivoted into, well more progressively into these type of things.

Speaker A: Yeah, very good. So um, today what you find in the common uh, customers that you have, um, more that side of analytics and even fco. Because SEO is, has moved a lot since uh, the beginning of these AI things and AI revolution.

Speaker B: I would say that the basics and the core ideas are more or less the same because many brands still are still working on the foundations. For example, having proper tagging, having all of this stuff. I mean I don't do it, I don't offer tagging, it's not my specialty. But of course you still need that and especially if you can afford a server side solution in general it would be better. I don't see many downsides apart from the cost maybe for some brands. And then of course the other thing I notice is that most of them, like most companies I talk to, regardless of the area, geography and you name it, is that they are used. I think that LLMs are getting used improperly for some things but they are still very good to work on some specific workflows. They're much better.

Speaker A: Yeah. But um, if you see nowadays, um, this week I'm seeing everybody spending too much money on the AI because it's uncontrolled thing in the companies and they do for everything. And for AI it's not for everything. I think. What's your take on that? Because we are doing a lot of automation. That is pure logic. You don't need uh, in some sense you don't need AI for that.

Speaker B: Exactly, exactly. Pure logic would mean rules, rule based approach that you Have a set of rules, you define them. So LLMs would be needed. They give you a use case, a very valid use case that I'm currently working on. It's one of the best ones I think one of the most powerful content. So content is very underrated especially when you talk with people in analytics, in SEO, in marketing or if you do social media it's a different story. But if you talk to pure analyst, content is underrated because most of them think about maybe products, e commerce, catalog. But if you work in B2B especially B2B SaaS or competitive B2B, you sell based on trust and you get trust by making people aware. Uh, you exist and you make people aware you exist by having quality, actual quality and unique content. LLMs, they don't help you with writing content like actually writing it. It's bad but they help you in all the steps that come before that are very annoying. For example doing the research, proposing ideas or different angles, repurposing or distributing the content. So LLMs can help you because LLMs are nothing but language models. So true. What's the best use case for a language model if not language itself? So automations like those that are based on input that is for example web data.

Speaker A: Uh okay, so um, I agree with you because for uh content it's LLMs, it's a language model. Um mostly language model with reasoning or inference are a little different because they can do higher, higher job and higher cost as well. M what do you think about these main things that people are doing? Uh everything in AI like we said and some things are rule based or logic, pure logic that exists a lot of years for now in my history in my professional um take. But um nowadays um. So people are exhausting all the budget but not doing what supposed to do but for what you see for content but for analysis and for B2B SaaS what you think it's more um profitable. Because profitable in the takeoff uh you already said it's the content. But what you can do more with that uh in that take.

Speaker B: So the, the thing is that you don't actually for this specific use case you don't actually need to. Well per se you don't make more revenue directly if you know what they mean. It's not uh a direct connection to revenue because it's not like you write a great article then you sell via that article but it compounds over time. So the idea is that you have content that is original meaning that you are the only one covering it or you have enough material to convince people you're an authority. And the keyword here is to convince, convince, persuade people that you are an authority. This is the keyword. And then of course this stuff, you don't need to overdo it. You can also have a mixed approach which connects to what you just said, which is what they do. If you want to integrate a rule based approach, which doesn't need AI with this approach that is more chaotic, not deterministic, which is content or language, which is very unstructured let's say compared to numbers at least you will need to use for example skills. I give an example, I'm working on a project or claude project. You can define skills in a skill. You can instruct the model, for example CLAUDE to do something with a script or by using a certain routine you say hey, if the user or if I'm asking you to, I don't know, extract the data from BigQuery, then don't do it your way. Use this script that is here, which is a script, it's a rule. If you think about it, it's always like that. And then the output, once you have the output, execute these commands. And this is the non deterministic part, the LLM part. So in theory you can have a mix of both, you can do both of them while not overspending money. But the core is that it's not that the automation, the LLM automation per se is profitable or not. It's more about the people actually doing the actions. Because I think this is a big issue in analytics in general. Like I give you advice, I don't have the political power or maybe I'm not the owner, you disregard it or maybe uh, uh, you listen to it but then you don't have resources to execute it. I think this is the issue where automation and everything else fails. Not the automation is that people don't want to do it or maybe they don't have the resources or maybe the execution is lacking.

Speaker A: Okay, so um, uh, just to wrap up so everybody, what needs to do, uh, what is the goal and the end result. But we have to be consistent, uh, convenience, get trust on the audience and then construct all the content by being um, an alternative person and content, uh, making content what you think it's um, now changing a bit a little the subjects. So um, I think integration between all data sources that you can join together are very useful to get to know the customer. What's your take on that? Leveraging the AI tools that we have at our disposal nowadays, what you can

Speaker B: do with that so in terms of integrations, actually, uh, this is a great point because I always complain that especially in web, people only use GFR as a data Source. There are 1 billion data sources out there, and yet people talk about tracking gfr. There is no gray, it's only black or white. There is no in between. There cannot be anything in the middle, which is false because there is a lot. There are a lot of gradations, a lot of different colors, for example. Sure. Um, I gave you a very stupid thing I'm testing Now. It's combining GA4 data with GSC data, which actually looks dumb. You say it should be easy. It's not easy. Trust me. Many engineers tried. They do it incorrectly or it's not proper. And this is one. I wouldn't say AI helps with integration because this is mostly engineering. This is data engineering. Regardless of AI, this is purely engineering. But AI, of course, like everything is, uh, another layer. And it can help you for some things, if you're technical or if you do these integrations yourself. Of course, the time to code them, to prepare them without paying for a tool, it's quicker because, for example, before you had to pay for a connector. Now you don't necessarily need to pay for a connector. If you have a decent engineer, they can reverse engineer it or find a solution much more easily. Because if even I did it for some things, I guess a full engineer can do it much better than me. With AI today, no, you have a lot of tools. And also in terms of, well, not integration, but on, um, what comes after everything is integrated, I think that AI, well, if not abused, if used within the boundaries, if used properly, can be good to expect to, uh, inspect angles, which is what they said before about content. It's the same idea. I think that AI is very good for finding new angles and challenging you. What I mean is that the integration part, I think it's kind of easy to pull off because it's engineering, you can expect some stuff. It's data modeling. The usual rules that you should know by now. If you do that. Of course, if you do that, sure. Uh, what I like about AI is, for example, that I usually challenge myself with AI. Uh, I ask very specific questions to Claude, for example, to say, okay, but what I built like, does it make sense? How would you do it differently? Can you prepare a plan which is different or where you challenge me? Stuff like that. I would say it's very valuable. Even though people, naysayers, we like to say, uh, but AI is not a human. You don't have to outsource your thinking. I'll tell you what most people have never thought anyway. So outsourcing your thinking is not a big deal. In some cases, people do it anyway. I think it's one of the most underrated points. Uh, I recently tried it with agents because many people don't know it because it's. I don't think this was properly explained while you use Claude or these tools. I don't think they never explained it properly. But without doing crazy stuff, sure. If you ask Claude today, if you go to Claude or any provider and you ask, hey, I want to achieve this thing, I have this task, okay, I have this task instead. Most people just say Claude, they want to do this. Can you do it? And this is a quick way. There is also another way which is more sneaky and they would say more effective, which is, hey, Claude, I need to achieve X, I need to do this. Can you run five agents or you name them with these characteristics and then give me some verdict, a feedback. So for example, you create one agent.

Speaker A: Good one.

Speaker B: Yeah. Ah, that is more conservative. An agent that is more, well, innovative. You build a different perspective as separate agents in the prompt. So you don't even need to make it crazy complex to start. If it's a simple project, you don't need to. If it's more complex, you would have to create separate markdown files or do it more properly. But you use this method a lot. Sounds so dumb, but it's so effective because these agents run in parallel, which means they don't talk to each other. They are independent, they are independent voters. And then at the end you will have a verdict. For example, 3 out of 5 voted in favor of the solution. 2 out of 5 voted with this other solution, and so on.

Speaker A: Yeah, Mark, that is a very, uh, very good uh, example that you gave. Um, and that is go to my second, uh, my, my follow up question. That is, uh, if we, that analysts, um, have our job more easily because of AI, uh, what, we become more strategists. Right?

Speaker B: Yeah, yeah. So I was waiting for this question, actually. I was waiting for it. I was there, wait. So let's start. I want to give the, you know, the philosophical or the cool answer here. So analysis as a word comes from the Greek. It means analuo, which means to dissolve into essential components. Meaning you take something, you physically dissolve it, you make it, you break it down, you make it easier. I think that before analytics was the opposite. It was sold as making it more complex rather than dissolving it. Making it easier, making it smaller to digest. And the thing is that as you said, now we are closer and I like it to strategist, but more than strategist, even more. We are like decision optimizer. Or if you prefer, we make people think we should by asking the right questions. For example, a bad question would be for a business, I don't know why is like why revenue is not growing. Like, this would look like a normal question. You would say, well, of course we have to make more money. No, why is this a bad question then? Well, it is a bad question. First of all, because it doesn't have, you know, a proper and simple answer. It's normal. And second, because there may be better questions. For example, as an analyst, I often ask, okay, why do you say you want more traffic? And people are like, well, because I want to make more. Okay, so you want to make more money. You don't want to make more traffic. And then I ask, what do you mean by making more money? How do you quantify it? And they say, for example, revenue. And they meant like, but you sure it's revenue? I'm not sure. Maybe you mean contribution margin. If you're an E commerce, maybe you mean net profit. Because there is a. All of these things. People, uh, were not doing them. I think in web in particular, but in other branches of analytics, I think so. But in webinar, I think we have

Speaker A: been towards tagging like you are saying, uh, I think what you are saying and I agree with that. Um, the, the pure. I um, don't know the pure jump that we should do. It's to um, the way that you answer the questions, the more simple way that everybody can understand that it's not, um, it's not uh, applied to everybody. It's a very difficult to say, um, complex terms or complex things in the simple way that everybody can understand. One that I can do, example, it's Avinash Kazik, that is you can challenge and say things that are very complex in the simple way that everybody can understand. And that is the um, like our uh, mantra to go. It's to have simple way. But. But the question that um. And the take that I want you to answer, it's uh. So the data analysis, go to data strategists, go to the business side of the questions to then translate to actions and insights. Right? It's. So that's, that's the take. That's the take that I always, uh. Because I always work with the customer inside. So my focus is the customer it's always the customer and I think most of the web analytics guys are more go to more technical part than the business side of the question. So what I uh think you are on the same same angle. You see on the same angle because of what you're saying. What, what we can expect the guys that see the things more on the business side leveraging all these tools. What is if we have a crystal ball, what is the next two years? Not five years, only two years. I think it's enough to see what is coming.

Speaker B: I'm curious when I make predictions because well of course we don't know but one of the things that I predict because we already have hints today is that most jobs especially in web will be more specialized or they will change. Which is what I just described before. I hinted at it kinda because years ago it was mostly about measurement, tagging. Ah, uh, you want to do analysis. Cool. You have a marketing specialist for that. Don't bother with it. It's another guy now. I think it will change because especially in advanced economies countries because in the end we have to make a difference in some countries some industries are not as developed as other countries. Naturally. Yeah, that's. It doesn't have to be related because for example Scandinavia, the Scandinavian region, which in my opinion I deem to be the most advanced or one of the most advanced. Web analytics for example is not as good for business. The usa, UK are decades beyond because they are more entrepreneurial, they're more aggressive in their spirit. Now they approach these topics. But for example, in terms of web, if I have to look at innovation, I would rather look at Denmark or Sweden for example because from what I noticed they're more innovative. And if you look at these countries you see that something is changing or even what they offer, the topics they talk about are different compared to 2 or 3 years ago and not only because of LLMs. For example BigQuery, data form, DBT, data modeling. All of these topics before they were existing like data warehouses and data modeling are not new topics. But they were not as popular or needed as of 2026 or today. So this is a change I'm seeing that you will have more specialized positions or roles requiring these new skills. Engineering, engineering in particular, because now the entry barrier is much lower. Second for LLMs, I don't think, because it's very unlikely and I will tell you why that LLMs will completely cut the jobs. I don't think so. One reason, number one, companies and all the layoffs that are happening worldwide, pretty much everywhere. LLMs are an excuse because they don't fire people because of LLMs. It was because of over hiring during COVID or the wrong periods or you know, because they need to make cuts and they need an excuse for the lawmakers or for someone for reputation. This is number one. Number two, which is very, very simple and not many people think about it, but stop one second, think about it and you will realize I'm right, is that LLMs are not self sufficient. Machines cannot automatically start. They don't have, you know, this free will, let's call it this way. I give you an example. If you're a company, you're a manager. Imagine you're a manager and you say I don't need a person to do this because now I have AI. And let's imagine that the AI output is actually good. Let's imagine in a utopic world, utopian world, that everything is perfect and the output you get is as good as a human. Okay?

Speaker A: Uh-huh.

Speaker B: Nothing would change because you still need to ask the LLM, like a human person. You still need to give the input, you still need to manage them, you still need to instruct them. So the problem is that when people talk about machines, they forget this bridge, the missing link, which is machines are our, well, you know, they do our stuff, they need to execute our orders. They have no free will, they are not self sufficient. And uh, you need to instruct them. And as an analyst you should know that distracting people or giving advice is an art itself. It's not easy, it requires skills. Asking the right questions is a skill. Instructing people is another skill. Understanding how you should structure a, uh, process is a skill. So will we be orchestrating agents? I would love to. It would be my dream. I would love to. It would be very nice actually to actually do it. It's as bad as they think. I'm not negative. I don't want to spread negativity and say, ah, uh, the jobs are finished, whatever the economy. No, I think the reasons are other reasons, simply companies wanting to make money using the scapegoats. So I'm very positive about LLMs for now.

Speaker A: Yeah. So my take, it's more or less like you. I think it's, it's more positive than negatives because you know, uh, you can do more things that you can do, you cannot do in the past. So I think companies that have fewer resources, human resources, can do more, uh, with less cost so they can do more things, more project more quality that uh, they can deliver to the customer. That they used to do it. That's my take on that. And I think um, for the ethics things that. Ethics, there's a lot of, of ethics here because of the geopolitics and so on.

Speaker B: Yes.

Speaker A: And the um, the sovereignty of Europe, that is uh, now it's a question that should be raised, uh, years ago, but it's.

Speaker B: Yeah, it's now.

Speaker A: It's better now than ever. Yeah, yeah. What's your take on that? Because, um, even Switzerland as ah, it's an island of independence. Even on Europe, like you said, the Nordics and the Scandinavian have on technical part. They are, I think they are very, very straightforward on the, on the front, on the frontier of new measurements, new techniques and so on. But I think like you said, I think business um, takes are if you aligned the data strategists, data analysts with the business outcomes for the customer, you get better serve the customer because you know what the things that should be there to get the outcomes that, that you need. So the question is what about data ethics? I. Ethics in the same way. And uh, the sovereignty of the European countries on that take as well because uh, for the future, I think you already have a very excellent take on that. Thank you for that. So all about sovereignty and ethics because it's a huge team now.

Speaker B: But I have a lot of. I'm very opinionated on this topic, so I would like to bless the audience with these ideas. The idea is that I think that the problem of the, well, of the EU in general is that, I mean we already know we are passive. Like there is this passivity. We are passive. We don't do stuff we're not active enough. It's like thinking about problems once they happen, not before they happen. Uh, it's a very passive mindset and there are many reasons for that. There are many, many reasons. Of course there are geopolitical powers at play because how can I say that in a very nice way? It's not like countries are actually independent. It's not like Spain, Germany, Italy can wake up tomorrow and say, cool, I want to do that. No, there are more equilibria that we normal people, we citizens ignore, we don't know about and we have to consider. So there is a lot of internal politics between inside the European Union, which means between members, and in Italy, especially from Italy, this was very common when I was young because I remember that every single time. And maybe you have seen it also with the recent um, conflicts. No, there are for example members, uh, arguing X and other members saying no, no, let's not do it. While the United States, despite being a federate country, despite being, well, 50 states, it's still one country that are very united. About that for some reason, I mean, we know why, historically speaking, compared to other countries. And this is one factor that we completely lack. You know, this is very important. We are not actually united because how can you unite countries with so much history? Completely different. You cannot compare. Yeah, Italy, Greece and Spain, which are closer to Sweden or Lithuania just because they're Europe. We are completely different histories, completely different cultures and perspectives on life compared to Americans. Even though, of course, the west, between east and West coast, there are differences. I'm not saying they are all the same. No, but these differences are small compared to the differences there are between a guy like me from southern Italy to a guy from eastern Germany. They are completely different. You know, we're completely different. And this is one, this is a big difference. And second, the fact that the eu, I wouldn't say it's a power player because we depend on the US if the US say, hey, you need to do X, most states, Germany, Italy, whatever, uh, have to follow. And France, no, France, it depends, but they have to follow. Number two, for the ethics. I think that ethics. How can I say that ethics is, well, subjective. Because it depends on the winner. It depends on what is the current framework. Because if tomorrow, let's say, if tomorrow they say that, I don't know, this idea is completely valid. It becomes widespread. This is the new ethics that is now considered ethical. If you look at 100 years ago, actually not even 100 years ago. If you look at 50 years ago. 50, not much.

Speaker A: Yeah, yeah.

Speaker B: You will see that the concept of ethics, or, uh, what we consider normal, even in movies, even if you watch a movie from the 80s or the 70s, they have a completely different concept of what is politically correct, what is not really correct, what is ethics, exactly what is normal. That's why when we talk about ethics, I'm always a little bit skeptical because it can change so fast in like five or 10 years for no reason. Because now we're talking about data, sovereignty and all this stuff. I remember that 10 years ago, no one was caring. They were like, ah, uh, cool. My data go there. Ah, uh, interesting. Ah, uh, they know what is my favorite color. Very cool. But in the last year, in the last years, there was this obsession with privacy that people were like, ah, no, I don't like cookies. Ah, uh, no, no, no, no. You have to ask for my consent. I didn't give consent. But 10 years ago, the same people, including me, of course including me, including us, uh, most people were like, yeah, okay, accept all the cookies, I don't care, just let me use the website. Stuff like that was completely different. So that's why for AI is the same topic. I don't care about AI per se, knowing about my stuff. I am only worried when they triangulate this data. Because the dangerous thing is not giving your data to the US or giving your data to Google. The dangerous thing is that Google has, uh, all, you know, all your other data points. They can combine all the data points together. And this was also used in the latest wars or conflicts. I don't know if you, if you knew it, but they are able to study your behavior. For example, they were studying their behavior. I'm, um, not joking. Their behavior of enemy commanders. So just to say that even this stuff has moved on a cyber digital level, they can tell you, okay, Jorge at 1:20pm usually walks his dog. He usually, you know, follows this route. So if we have to attack him, this is the moment. This is like this. Five minutes per day. This is why data sovereignty is now important. Because people have realized that it's not just, uh, Jorge likes salmon. Very cool. No, it's more about your behavior. What, what they can expect from you. If you like a certain type of person, what are your hobbies so that this can be used against you. And another example is that if you Instagram. Instagram. How is it possible? Like, if you think about it, you mentioned to a friend, I would love to buy an air dryer, or I would love to buy a PS5. Ten minutes after you see an ad and you're like, eh, yeah, what's going on here?

Speaker A: Yeah, that is, that is true. Uh, so, um, to wrap up the ethics. The ethics? Yeah, it depends. What is the ethics? Because in the States you can buy, you can buy personal data. Uh, not in Europe for now, it's not official, but in the States you can buy personal data from the banks for everybody. Um, okay, so, um, we are almost in the end of our show, um, Marco, where people can find you on the, on the social media.

Speaker B: Yeah. So mostly LinkedIn will be my, my favorite or also my website, which is search e o tics.com. well, it's also on LinkedIn, so you can just go there and this, these are mostly my channels. I would say yes.

Speaker A: Okay, um, now for the final ask, uh, you for a recommendation for our audience. What is your recommendation? A book, A movie? What is.

Speaker B: Yeah, there is, there Is a book, uh, there is a book. Let me find the English name because I always forget it. Okay. I think I found, um, the name. So there is a nice book, uh, with a very weird title. It's a very weird title. It's not about what you think. This book is called the Sword that Kills and the Sword that Gives Life. This is the full title translated from Japanese. So this is a very simple book which is not about swords. It talks about swords, but it's not a book talking about fencing. It's a book on mindset of how to, well, approach, uh, life and how to think, for example, about problems. And I think that books like this one, I don't want to spoil or give you any detail, are important. They challenge you to think the status quo, which is what I did. I opened this well in this shot, because they make you think.

Speaker A: Yeah, yeah, yeah, that's, that's good.

Speaker B: For example, I give you an example, an example of how you can think in a different way. We need to make a decision. We need data to make this decision. This is what the majority of the people think and this is the mainstream thinking. And they tell you that sometimes the best sword is not having a sword, is no sword. Sometimes you don't need data to make a decision because there is no need for data. If you have bad data or if the decision is quick, you need to act in two minutes. You don't have the time to look at the data. So using other types of data or using your own experience can also be good. So this book will force you to think, uh, in this way, not to say more.

Speaker A: Okay, Marco, so thank you very much for um, for this, for this recording, recording our podcast. I will be putting all them, all these things that Marco told in the LinkedIn in the episode. So Marco, thank you very much, uh, for your time and attention to the podcast and uh, speak to you soon.

Speaker B: Bye bye.

Speaker A: Mhm. Analytics Friday podcast.