This FinTech did 560m profit | Fabrizio Ballarini, Head of Organic Growth @ Wise

00:00:00: Wyse is moving a hundred forty five billion pounds across the globe, serving fifteen million customers across a hundred sixty countries and forty currencies.

00:00:10: They've been profitable for years with their latest financial accounts, posting a five hundred sixty million pound profit before tax, while saving customers two billion pounds annually on hidden banking fees.

00:00:23: And although the absolute number is pretty high in relation to their scale, they are just spending about fifty million pounds a year on marketing.

00:00:32: One part of this remarkable customer acquisition efficiency is a well-oiled organic growth engine.

00:00:38: And to understand how wise built this growth machine, I'm excited to have Fabrizio Ballarini on the podcast today.

00:00:44: As Head of Organic Growth Advice, Fabrizio oversees a cornerstone of their customer acquisition, and he's already been on board for ten years.

00:00:54: We'll dive into programmatic SEO at scale, the real challenges of operating in a hundred and sixty countries, and why sometimes the best AI SEO hack is just improving pages the way he always did.

00:01:06: So, welcome to the podcast, Fabrizio.

00:01:09: Hello.

00:01:09: Hi, everyone.

00:01:09: Thanks for having me, Nicholas.

00:01:11: Thanks for coming on.

00:01:13: Let's start with the easy one.

00:01:14: How is it possible to stay ten years at a company that has grown so massively, like, wise?

00:01:20: Yeah, very good question.

00:01:22: Sometimes I ask myself, right, time flies and it's hard to go back.

00:01:26: I mean, originally, just for context, I joined wise from consulting to enterprise clients through agency.

00:01:34: At the time, my main frustration was that working on mostly on technical project, the implementation was really slow.

00:01:41: on some of these very corporate companies.

00:01:43: and then I said okay let me join a kind of startup.

00:01:46: right at the time we were at series C that you know is growing.

00:01:51: we had some engineers in the team early days even though there was no SEO team.

00:01:56: so I think that was the original context to it.

00:01:58: and then the whole aspect of you know building a web presence and doing organic activity has now gone to roughly next level compared to what it used to be.

00:02:11: Also, probably my view on some of the topics.

00:02:14: Early days, we were slowly focusing on some aspect of SEO.

00:02:19: The original problem was literally going from spending maybe, or for growing, ninety something percent paid the marketing to increase the organic market share with the organic share of marketing.

00:02:33: And then like we, on many topics, we have evolved to you know, to next level in the sense that, you know, we continue to run the SEO program that we initiated like almost ten years ago now, but also on top of it.

00:02:46: we diversified across domains, we diversified across platforms, right?

00:02:52: We touched on area that back in time were completely out of my scope personally, but also, you know, the team.

00:02:58: And usually you go with what makes more sense or what is more obvious to begin with.

00:03:04: maybe you don't know necessarily what is going to be valuable in ten years time.

00:03:08: So there are many things that we built, literally the first two quarters that I joined that now are either extremely valuable or we have to change a bit of opinion on the value of that.

00:03:18: but then like initially you know also when you start in a company at the time we were only in few markets compared to now we are present globally.

00:03:28: at the time we're only serving B to C. these days wise is three companies right this we have the B to C. why is the account that you and I have, we have a bit to be offering for businesses from the freelancer to the large business, and then we also have an enterprise offering.

00:03:44: This is an area that we generally do less organic growth through my teams, because this is where we partner with banks and other payment corporations like Google Pay, and number twenty six, Bank, Monzo, all this kind of fintech.

00:04:01: like out why why like.

00:04:03: this journey has been quite quite crazy right also?

00:04:06: we went through IPO.

00:04:07: we changed the domain because obviously at some point we decided to rebrand from transfer wise to wise.

00:04:14: you know these are all decisions that come throughout the years that when you start early days you don't.

00:04:20: they don't forecast to happen right.

00:04:21: so the goal the early goal was simply to grow web presence and trying to be Very good that SEO something that back in time we were not when when I started pretty much was the home page and a couple of articles on the absentee.

00:04:36: that was all a traffic.

00:04:38: The only good thing that we had was maybe an okay Kind of link domain because we are quite popular in the fintech space one of the early fintech back in time and We had a lot of brand search demand because most of our growth even today is true or the mouth, right?

00:04:53: So it's basically customers that use wise product.

00:04:55: They're extremely satisfied and then they tell their friend who Google wise, right?

00:04:59: So we had these two positive things already there, but then everything else was quite was quite like to be built right like from the engineering side of things to other aspect of content production and other aspect of building building content on the web right.

00:05:16: so I was mostly focusing on the engineering part and myself.

00:05:21: someone else was helping me on the editorial and now the SEO program is roughly these two area plus a bunch of other domains that wise own on top of us.com.

00:05:32: Crazy journey.

00:05:32: thanks for the overview.

00:05:35: You already talked about organic growth, but you also talked about SEO.

00:05:38: So can you give us a quick behind the scenes like what does organic growth include for you from the vice definition?

00:05:45: Yeah, so we try to group together all the activities where we have to build something, right?

00:05:55: We have to build some content, some assets, some tools, any sort of you know, also collaboration with other brands or anything where there's a process of building some value for the customers and often this is like account resource intensive exercise as opposed to media spend, right?

00:06:17: So these are teams that are relatively at count heavy where we need more people for more languages, for more markets, for more disciplines.

00:06:25: And then we try to view the value of this and assets that we build organically on the longer run, and as opposed to maybe, you know, my colleagues on paid media that focus on their investment this month and the return on ad spend this month, right?

00:06:44: So this is roughly how we differentiate organic from, from like, from paid, and then within organic, between organic and SEO, we have roughly three, three SEO teams, right?

00:06:57: The people that work on wise.com, both product and content.

00:07:01: The people that work with affiliate partners and the people that work on some other domains, right?

00:07:05: So this is what we class as SEO.

00:07:07: and then within the organic family we have other activities like CRM.

00:07:12: content creation on social media, all sorts of other platforms that might be a blend between social and search.

00:07:18: For instance, in Singapore, we have a team that looks after WeChat, Alipay, Red, that is the Chinese Instagram doing Chinese TikTok.

00:07:28: Some of those platforms, there's a bit of a blend of the two, right?

00:07:30: It's not pure SEO, it's also not pure social, it's a mix of all the things.

00:07:35: And then sometimes also collaboration with other with other partners and brands in the sense that these at times in taste.

00:07:43: content creation from influencers where we create content together.

00:07:46: so there's an element of like you know, commercials as well.

00:07:49: It's not just purely organic, but it's generally an exercise where people from our team create some value for our customers on the longer term as opposed to just paying like an ad partner to have some visibility and serve some creatives to those customers in this short term.

00:08:10: And do you also consider PR part of organic growth or is it a disadvantage?

00:08:15: No, historically, PR has been separate from my team.

00:08:21: Back in time, we... So, like, the early PR of Wise was... Wise getting noticed, right?

00:08:29: Because at the time, like, let's talk about two thousand... Before even I joined, before two thousand and fifteen, when Wise started, at the time, the PR team, and we had a team... I don't remember the exact name, but I think it was something around brand engagement.

00:08:44: They were doing some of our most famous crazy stunts as in, you know, getting naked on bank in front of a bank station, organizing a funeral for the bank with a parade in the central, in the city.

00:08:59: So all sorts of like activity to get noticed, right?

00:09:01: And all sorts of tactical PR to bring some customer acquisition.

00:09:07: This has evolved quite a bit.

00:09:09: These days, also, we are a public company, so our PR team has got quite a large remit in managing also our reputation and the aspect of being a public company.

00:09:20: So it's a little bit different from what we do for organic growth.

00:09:25: There are some elements of collaboration.

00:09:28: At times, there are certain PR campaigns where we also engage with some of our team, mostly on the social side, more than more than search right.

00:09:37: historically on search we have not engaged and people sometimes on LinkedIn find this controversial.

00:09:43: but I think that in the last ten years we plan every quarter there's only been one to two quarter max where some people in one of the SEO team had a small test about you know link building NPRs but in practice ended up being nothing.

00:10:03: and then so pretty much we never buildings, a little bit because we didn't have to, a little bit because we were always so kind of, I was always so paranoid about not having enough web presence and pages on the website, because you know, it's easy to say, okay, let's optimize what we have.

00:10:21: When you have a relatively large website of an existing brand since we come from literally on page and that's it My main worry was like, you know, do we have enough articles pages and lending pages and and enough traffic?

00:10:36: by building more pages less so by trying to Optimize I mean we do optimization on content existing content that we have an existing pages, but we never try to work on the link aspect that much.

00:10:50: A little bit because in some area, some of our competitors, for instance, xc.com, they had the link profile of Wikipedia.

00:10:58: They've been on the internet since probably the dot com time.

00:11:04: And so it's not that we could outrank them by doing a couple more links.

00:11:11: And so we will probably never have more links than xc.

00:11:14: in some aspect, but we are competing with them quite well now, something that was not the case many years ago.

00:11:22: And so like we said, okay, since we are like the new kids on the block, let's try to build the best possible pages with something that is really unique on those pages that you will find only with us.

00:11:37: One example that I make is that It's not mandatory to surface price comparison data on the pages.

00:11:45: You could build pages without data of the comparative advertising, but we invest quite a lot to do it.

00:11:55: And this is not trivial because... and differently from other industry where you may pull an inventory of availability and data.

00:12:06: It's not that you can pull an API with how much banks charge for transfers.

00:12:11: We have to do all sorts of tricks to collect that data, make sure that it's compliant and do a bunch of things.

00:12:17: There are areas where we say, okay, how can we be the best in this type of intent for the user in answering this intent?

00:12:27: And we historically never solved it with links.

00:12:29: We solved it by making these pages better.

00:12:33: Because we come from a scenario where the baseline was zero.

00:12:37: So we said, OK, let's make the page eighty percent better.

00:12:40: But then once we were there, we said, OK, let's try to make it hundred percent better as opposed to try to build some links to it.

00:12:47: Again, this is very specific to us.

00:12:49: It's also the fact that we were a fast growing tech company.

00:12:54: we received.

00:12:55: investment and then we invested in a bunch of other advertising like.

00:13:00: but then I think that the argument of that people sometimes make is that you know yeah this is very easy when you are unicorn to you know say that you don't need to do link building but still the same people who are pitching to me to do link building throughout these years right.

00:13:14: so to some extent right.

00:13:16: I challenge them back and say you know like it's fine but you told me that I had to do it right.

00:13:21: and so I think there are many situations like these where we try to be very practical about what is helping our customers or not, right?

00:13:30: And if something is kind of favoring our customers, we invest into that.

00:13:34: If something is not, we generally are a bit more reluctant to do that.

00:13:40: And I think this is probably becoming more true in the age of AI in the sense that I often discuss with my team.

00:13:52: what if Tomorrow the beautiful search traffic that we receive every month goes away?

00:13:58: You know because you know many times we we collected million of visits on on wise.com and With the with the view that even though we are not doing a ten out of ten job in converting those customer into wise customer Next month they would have come back and search again.

00:14:16: You know and maybe that was the wrong attitude right like at times some of our you know our chief product officer at time was coming to us and saying, you know, why can't you do a better job on those on those pages, right?

00:14:29: So probably we were missing out as a CEO throughout the years, but now it's more obvious that investing in things that are valuable to your customer.

00:14:37: And again, why is now has got, you know, more than more than fifty million customers.

00:14:41: So it's relatively straightforward to say, OK, we're just going to invest in what is on our properties, because at least to our a million of customers, this is valuable, as opposed to investing in things that don't necessarily, you know, if tomorrow the internet disappears and we're just left with our customers, what do we give them, right?

00:15:00: And that's generally what we try to do.

00:15:02: That doesn't mean that obviously we will need to work out distribution channels, right, in maybe slightly different shape than what they have been throughout the years, right?

00:15:13: Like for sure, on the SEO side, we have benefit from from that distribution channel up to up to now in the sense that you know like even some of the you know more popular calculators and tools that we have.

00:15:28: you know we have invested in them throughout the years.

00:15:30: you know recurring investment quarter on quarter with engineering time design and and user research because eventually distribution was allowing to fund some of these right wise being kind of well funded.

00:15:45: you know unicorn didn't mean that we were allowed to burn money.

00:15:49: in fact we are probably one of the most diligent you know.

00:15:53: and the number that you see also on spend that you mentioned earlier goes back to the fact that we run the marketing sheep on an extremely tight set of constraints.

00:16:04: that to the outside they seem almost crazy, right?

00:16:07: In the sense that when we speak to other operators that run similar companies, they don't understand why we are so strict in evaluating our investment.

00:16:16: On the inside, this makes more sense because eventually our mission is to drop price to zero for our customers and And, you know, you can only do that if you become extremely diligent on the marketing spend, right?

00:16:29: And a little bit on the paid media.

00:16:31: and the other half is on the on the organic side of things where basically, you know, we have to try to constantly increase the share of organic against paid or colleague running paid media.

00:16:43: I'm not going on holiday anytime soon.

00:16:45: They're very busy and they're increasing their investment.

00:16:48: But, you know, if we want to continuously drop price for a customer, we need to drop cost of acquisition, right?

00:16:54: So it's quite a unique situation to be, you know, if I were at Apple right now, the next year the iPhone is not becoming cheaper, and therefore I might have a different set of constraints on in terms of like what the lifetime value of a customer would be for us, at least in the original product line, which is the money transfer product, we try to consolidate our price for our customer.

00:17:18: Therefore, We have to become better and better at acquiring customer at less at less, right?

00:17:24: Which is kind of led to a bunch of these investment probably led to being a bit more bullish on some of the organic investment at time but still in a relatively conservative way in the sense that Even though organic is an extremely good investment in term of payback on investment.

00:17:40: We still don't You know, we haven't like bear money for the sake of it.

00:17:45: Everything is run pretty pretty diligently.

00:17:48: You know, it's just that We probably more bullish than other company even today try to invest in into into things like content creation.

00:17:57: Sorry.

00:17:58: Another type of activity because you know for us at least at this point the bull is rolling like it's not that we are building wise presence from zero right.

00:18:08: we are sitting on on millions and millions of visit every month therefore you know.

00:18:13: There is definitely a shift in certain user intent.

00:18:19: We will find out how much the consumer behavior will shift towards the new platform.

00:18:24: But then, even though that was pretty much one of my talk recently, even though we lose thirty percent of click through, we still have to do extremely.

00:18:35: the next job on the remaining one.

00:18:37: We're not going to throw it in the bin at the scale at which we operate.

00:18:41: If I would be a startup and I have zero.

00:18:44: organic traffic on ten blue links, and I were to try to build my web presence.

00:18:49: I would probably try to be more creative than what we are now, but as pragmatically, we're still sitting on a lot of traffic that we cannot discount tomorrow, just because some user intents are slightly different now in nature, and also the platform mix might be slightly different.

00:19:08: Yeah,

00:19:10: so much in this response we have to dive into.

00:19:15: Let's start with a first one.

00:19:17: The reason I was asking about PR is because there's a lot of talk currently about PR now being so super important with influencing LLMs because there are these authoritative sources that I use as citations, et cetera.

00:19:31: I'd like to get your perspective on that and also on generally the AI search landscape.

00:19:38: Yeah, yeah.

00:19:40: At the moment, I think that, so there is a lot of focus on the AI visibility aspect, right?

00:19:49: So what I mean by that is kind of checking if within, like I assume, depending on the tool, a given set of prompts, or in a given set of conversation, you are mentioned or not, whether you have a link or not, and so on.

00:20:07: The challenge to that with that is definitely useful exercise.

00:20:13: And whatever tool someone used to do that, at this point in time, I haven't seen a clear, like someone that got it particularly different from each other in some aspect.

00:20:28: To some extent, our internal team was building an LLM agent, could be building this.

00:20:34: So I understand that the bakery shop will not be able to do that, but we have resources.

00:20:39: We were building SEO tools even before.

00:20:42: The bit that I'm probably more curious about and this aspect of what the influence, the citations and aspect is, is the aspect of the way that these agents are crawling our properties and trying to build the relationship between between that and what they know about us, our content and our brand.

00:21:09: And then, like, start from there before going to third-party sources in the sense that, at least in the case of Ys, you know, globally, for at least the FX vertical, we are one of the leaders, right?

00:21:29: It's less likely that LLMs don't know about us.

00:21:33: It's more likely that they might get a fee for a certain transfer corridor wrong.

00:21:40: It's more likely that they may misunderstand the payment method with another payment method and they might compare us wrongly with other people.

00:21:47: So like again like similar to the previous topic of literally traditionally SEO we are more focused on what how we are understood by models, right?

00:22:00: Then, like, how can we build additional understanding of why it's outside our properties and what I mean by crowding and, you know, like, throughout the time with SEO, we have learned... quite well, you know, how crawling and indexing works and how this influenced the content that is on the pages and the intent and how we associate index pages with intents that are answered.

00:22:27: I think this exercise with the LEM has not been fully debugged, right?

00:22:31: So if you open your log file, you will see that there is like open AI agents that My my guess is that they're more similar to a Google bot type of behavior But they're also.

00:22:46: there is the charge APT agent for instance, right?

00:22:48: This one of the most prominent at least in terms of like quantity of According that they do.

00:22:54: that is probably triggered by user User engagement on the platform and then pulling resources to answer those queries.

00:23:02: Yeah, I think in that case again, I'm less interested in at this point in time on.

00:23:09: If they go to a third party website to find out one fact about why is, uh, you know, because like often.

00:23:19: We should be mentioned and discovered also, right?

00:23:22: So again, if I would be the, uh, like, let's say a money transfer company that that is starting today, maybe I would be worried about.

00:23:31: you know being being known to the.

00:23:33: you know i would be maybe not mentioned on the newspaper as one of the leading player with a different set of changes but for us is at this point you know and a little bit was the same argument on old in building.

00:23:44: if we become so big then we will be mentioned to some extent.

00:23:48: to some extent and if you can see fun enough the report that you see often the studies and so on wise is among the top mentioned brands in fintech.

00:24:01: Even though, as I just told you, we don't do anything on third-party domains at this point to influence that, right?

00:24:09: Or we do activities through our social team, our own community, and activities probably even related to Reddit and so as part of our community engagement, but it's not for the sake of AI mentioned.

00:24:23: We do it because it was making sense to us.

00:24:25: Even before to do some of these engagements, right?

00:24:28: But we don't.

00:24:29: we don't try to connect the dot on this stuff at this point the big.

00:24:33: that is more interesting is Imagine that you have like, you know, thousand product lines a lot of different pricing for it a lot of this different feature attached.

00:24:44: You might have we know wiki articles.

00:24:46: that explains certain use cases.

00:24:49: how is this understood?

00:24:51: and how is this, you know, like discovered, right?

00:24:54: Like that's the bit where I think that in general, and obviously, you know, it's less likely that someone will share their log files as compared to someone trying to run a visibility report.

00:25:05: But that's, I think, where some of the source might be in terms of this level, this understanding of optimization, right?

00:25:12: Like that, as opposed to just being, hey, can I get a mention on... Another website to find out whether I will be mentioned on the back of it?

00:25:21: in theory right like you know if this these models become really smart and And you know they go beyond what you who used to do of trying to connect a couple of key word to a page and trying to broad they understand the content.

00:25:35: this was The Google of ten years ago right?

00:25:37: the Google of last year was already much better than that.

00:25:40: So like it's not this not as far At some point like I think that you know if they have the right information about wise and somebody is really poking you know whether wise is better than an alternative and they should be able to answer that that question when like and so what we've seen is that often the lm's does that?

00:26:04: like some of the chat assistant they do the homework that you didn't do.

00:26:09: Back in time the search engine was not doing back in time.

00:26:13: so you are searching a couple of queries.

00:26:16: you are collecting the links and then you are trying to make an evaluation on what to do what to buy what to use right.

00:26:22: If now this exercise can be done by the LLM agent.

00:26:28: There's still an element of crawling and indexing some of these resources.

00:26:32: right but then the decision making of what.

00:26:35: What what is good for the customer based on the query which is heavily personalized is another thing right.

00:26:42: You know if I was not trusting some of the ranking back in time or heavily personalized queries on a couple of.

00:26:49: Word type of query.

00:26:51: imagine now that I'm typing at twenty five meter sentence and there's a previous context to it and a lot of history to that right.

00:26:59: so.

00:26:59: Yeah, that's why I'm like, okay, right?

00:27:02: Like, is this exercise of trying to reverse engineer the query might become a bit trickier, right?

00:27:08: And then the second aspect is his search demand, right?

00:27:12: So by search demand, I mean, you know, one thing is to say, hey, I have as many clicks that come from a Google result because, and I can see the monthly search demand that is coming through.

00:27:25: this type of query in the form of impression, in the form of traffic that I get to the page, is the day that the agent, LLM agent, let's say HRGPT, doesn't provide you with any data on the search demand, and you don't see any click-through coming.

00:27:39: I think someone published some data that is extremely low.

00:27:41: That's pretty much what they would expect in general.

00:27:44: At that point, all the aspect of trying to attribute the value of any of these activity becomes extremely hard.

00:27:52: And I'm not saying that we will never do it.

00:27:54: I'm just saying that at this point in time, especially us that we have a rolling program with a lot of investment in SEO, it's not that we can say, hey, click of a button, we shift all our effort in this stuff.

00:28:07: that is most of the time not measured.

00:28:11: and most of the time like a matter of feelings right at this point right.

00:28:15: so I think this is the two.

00:28:18: one is the crawling and indexing and not indexing but the retrieval of the information and the way that they are digested.

00:28:27: and two is the value attribution of doing activities around that.

00:28:32: on the one I think we will get there easily on the value attribution is probably a bit harder on the number one in some of like activity that I envision company doing more than more than necessarily.

00:28:47: The aspect of like trying to not manipulate but try to improve citations through by citation on other party sources is more the aspect of fact checking.

00:28:59: the models, right?

00:29:00: So you might be doing this with an external tool.

00:29:03: I'm pretty sure that our guys who are building our LLM agent for content production, they're already in a position to do that.

00:29:13: We, you know, it probably would make sense that you make sure that all the content that you have is understood in the way that you want it to be understood across all models and you carry the models in multiple ways to validate that this is the case.

00:29:28: Yeah, this is one aspect and the other aspect is how do you probably build your own model or your own kind of customization of a chat assistant.

00:29:40: And on that one, I'm more bullish in the sense that, you know, is it true that is a better experience to ask a chat assistant to do the homework for you in comparing five articles and telling you which one, what's the summary of that intent?

00:29:58: Yes, but then like for instance for us, especially in finance, you have compliance, you have another bunch of aspect that fall into it, legal consideration with comparative advertising.

00:30:10: I think some companies would probably be better off to have that experience on their own properties than completely delegating that to the agent.

00:30:24: Obviously, you need to have customer in order to do that, right?

00:30:27: So, because by default, people will not wake up and come to your website to do it.

00:30:31: And they will first go and charge you before coming to your properties.

00:30:34: But the day that they have engaged with you, or even the day that they lend to one of your lending pages, you know, I think it would be better if the company would be able to train like an agent.

00:30:48: with their data, not just the product data on like how to use the product, but also the marketing data, right?

00:30:53: Like our content on the blog, our like content on landing pages, our comparative, you know, our price comparison data.

00:31:01: And so if this agent would have all this data, probably for some customer would be a better experience.

00:31:06: that's crawling a page, right?

00:31:08: Like, and they would feel more comfortable to continue that conversation, you know, with you, right?

00:31:14: In the same way they were having that conversation on.

00:31:16: on charge pt if they get used over time to have those conversations.

00:31:21: i think this is like an aspect that many companies i can see they're already investing or is already live on their help center right with you literally chatting.

00:31:30: To the wiki right rather than consulting the wiki.

00:31:34: i don't think it's very difficult to do it on.

00:31:37: your own marketing material and content, right?

00:31:40: With an understanding, obviously, that who is a customer versus who is not, you might have different, like, you know, engagement with them, but I don't think it's that far away.

00:31:48: Some companies, even with some of the shared plugin, they're already providing the chat experience on their own website on that.

00:31:56: And I think that, again, back to what the team does in SEO, the day that we need to train the model that we give to the customer, Probably we want to train that model with our own content, not with who is what is on the web, right?

00:32:11: Because God knows what that place that is cited wise, what they're going to write next to it.

00:32:18: God knows whether they understood the competitor correctly.

00:32:20: You know, we try to take that risk on ourselves, right?

00:32:23: And then delegating that task, which doesn't mean that obviously the mainstream model, we necessarily listen to our version of the fact, but at least for our customers, we have.

00:32:35: we have an experience that is like safeguarded by us.

00:32:40: right because it's very easy to say hey let's delegate the.

00:32:43: you know the consideration stage to the model is harder to then try to change it where you think that the model is getting it wrong.

00:32:51: right i'm not saying that all the models we get is wrong but there might be one model that might like you more than another or there might be.

00:32:58: and also We have seen another thing is that when you go to languages as well, so obviously why operate across multiple countries, even on pure localization and translation of certain resources?

00:33:11: on content side, the models are particularly different from each other.

00:33:15: So we don't trust the CharGPT as well in all the countries, in terms of what they're doing, and same in other models.

00:33:25: I think Gemini is the one probably from our internal study that does a bit better on some aspect of language localization.

00:33:33: But still, it's not a hundred percent.

00:33:35: And you have some situation where, depending on the language, you might have different level of... English is generally easier.

00:33:44: It was a little bit like the same in the past where people were complaining to Google.

00:33:49: about the quality of the result in Norway, I think, or one of these countries.

00:33:54: And Gary is, or somewhat the representative in Google, just suggested to make more babies in the country so that they could make more websites and more content.

00:34:03: I think it's a little bit that the problem is that, obviously, the models are as smart as the corpus of the web, I assume, without being so sophisticated on it.

00:34:12: Back in time you could see obviously obvious difference in the even in the quality of search results in Google and the understanding of context on pages by language, right?

00:34:21: And this is still the case and which again Kind of you know.

00:34:26: back to the original answer question.

00:34:27: like you know, I would I would focus on the crawling aspect heavily.

00:34:31: I would focus on training my own data sources.

00:34:36: But then there's other aspect of the content creation that we were doing for SEO that we're gonna keep doing you know, at this point.

00:34:44: We can do it differently in the sense that, you know, the ability to automate certain processes through LLM agent is like crazy mental, right?

00:34:57: So it's like compared to, even since one year ago, the team is making massive improvement on that.

00:35:04: You know, I'm not saying that we are producing the content end to end with LLM.

00:35:08: It's far from that.

00:35:10: have still a very large editorial team, any article that goes out goes through people, not just purely agents, but their day-to-day life has been, you know, speed up significantly and also the way that they engage with the agent is quite useful.

00:35:28: That aspect I find it more interesting, which again is down to doing a little bit more of the same, better and faster.

00:35:38: rather than doing something different because of then the way that the data is showed to the customer through LLM assistance like ChargeEPT, if that makes sense.

00:35:49: And can you walk us through how you're approaching building your own LLM agent for content production?

00:35:55: Yeah, yeah.

00:35:56: So we have, broadly, we have two... two reasons why we have such approach.

00:36:07: One is on the data security.

00:36:13: And this has been always the case, right?

00:36:14: Like, why doesn't use Google Analytics as primary analytics source?

00:36:20: Why doesn't use, even though we have licenses on various tools, it doesn't mean that we connect our own data to the tools.

00:36:27: Often we buy API and we ingest these into.

00:36:30: our are stuck and the same goes with with LLMs in the sense that you know we as you mentioned earlier we transfer and we hold you know billions of pounds for our customers.

00:36:43: we hold their financial life.

00:36:45: we have a lot of sensitive data so across the company there's like a lot of level of security around around usable LLMs which means that you know we The team operates a bit differently from maybe someone being a startup that you buy a subscription of charge and you try to build a bunch of prompts and the agents in that way.

00:37:10: We have to do it in a different way in the sense that we basically have shares where the data is secluded and then we query models.

00:37:20: to do different tasks and then we end up building agents that probably depending on the use case they have different set of models and tasks that they perform not just querying a model but also you know going to scribe some data going to collect and do some other stuff right.

00:37:38: so that's the way that we approach approach building these agents.

00:37:42: And obviously it's the most painful because you need some level of machine learning engineering support.

00:37:48: We have dedicated people who have become specialists in building agents, but then obviously the advantage is that they will roll out any of these agents through the editorial process, for instance.

00:38:05: We try to, with a lot of variances by country, by language, by content types and so on, but we try to have a relatively standard approach, as opposed to having fifty people for style, you know, now they're gonna use LLMs.

00:38:22: And we have seen this when we roll out agent to teams, right?

00:38:27: By agent, I don't mean, you know, you press a button and the article comes out.

00:38:31: We can technically do that, I think, and it's quite good, but we don't do it.

00:38:36: So, like, technically, the guys who have been the agent, I think that they within maybe a year, they could be getting there.

00:38:43: But we have a lot of reason not to.

00:38:48: And what I mean by agent is maybe like a task within the content creation process that we have speed up or did an additional check or automated because of that agent.

00:39:02: So one example is pure clustering.

00:39:03: Another example is a brief creation, right?

00:39:07: Like there are certain aspects that we automated.

00:39:08: And when we roll out this specific agent to the team, it's not that the team is generally saying, amazing, you know, I will use it from our own.

00:39:20: This is perfect.

00:39:21: We have a different level of adoption across the editorial team.

00:39:26: And this is not because people are reluctant of, you know, dedicating some of their job to machines.

00:39:33: It's because the actual output for our customers is not always the same by market, by type of content, by and so on.

00:39:42: So I think that's where the standardizing.

00:39:45: a little bit approach across them helps also to have more significant feedback that comes back.

00:39:51: We run constant checks and surveys on our adoption on what is increasing, what is not.

00:39:59: We let people create a little bit bespoke From within with some model just to let them experiments, but also then we try to Then take someone with expert in building agent to standardize Some of the processes.

00:40:13: just avoid that again, you know who does content in Japan?

00:40:17: We use a completely different set of instruction.

00:40:20: Then who does it in another country or maybe completely different model?

00:40:23: You know, maybe they you know just just because they not the lazy but because they used to one model as opposed to another As opposed to the people who are building the agents are testing different tasks with different models and trying to stress the boundaries at scale on what the difference is.

00:40:40: This is how we approach this.

00:40:42: And we mostly do this on editorial content.

00:40:47: We do a bit less of it on the kind of landing pages, calculator, or the programmatic stuff.

00:40:53: But even there, we will do more.

00:40:57: Some of the collection of data, for instance, price comparison data, can be helped a lot with some of the agents, right?

00:41:04: In kind of building knowledge, right?

00:41:06: So sometimes it's more important to speed up the knowledge building exercise inside the team than the creation exercise, right?

00:41:15: You know, often it wasn't necessarily like a writing problem, right?

00:41:20: So we have a lot of network of freelancers that write the content, the internal team produce the content, and then they own the performance.

00:41:28: So it's not that... writing with a scale even with freelancer would have worked completely fine.

00:41:34: Overall, the payback of that investment is still relatively low even by using real content writers.

00:41:41: And so it's not a matter of like cost reduction or that is more about like.

00:41:48: how can we, you know, have at times faster speed, right?

00:41:54: Like, so it's not that all the tasks done with agents are cheaper.

00:41:58: I think that when we do the math on building the agent plus the cost of running might be as expensive as someone doing it, plus you still have reviews, plus you still have, you know, it's only the speed aspect that has been interesting.

00:42:14: And then, you know, and then at times is the ability maybe to scale a little bit faster in some instances, or maybe a bit more flexible on resources, right, when we miss certain people to do certain tasks.

00:42:29: But then, yeah, part of the process is still running the same way that we used to, if that makes sense, at least for now on the editorial side.

00:42:41: I'm less worried about there being some discussion also about whether content that has been written with or with or by AI can perform or not, even in the traditional Google search.

00:42:56: I'm less worried about that.

00:42:59: Not that I advocate for making it completely with LLM agents, which is something that we don't even do.

00:43:05: But depending on how big you are, in terms of resources and the company, you might need to leverage more or less this, right?

00:43:14: So even in the past, even without LLMs, they somewhat wanted to go from zero to five thousand article on a blog.

00:43:24: They would have definitely compromised on the quality of the content, right?

00:43:28: So I think there was no scenario where they could have done a one to thousand with the same quality of maybe a large team like Wyze is doing the next thousand.

00:43:42: So in those cases, and that was equally bad content that was even written by people but equally bad or there was some level of shuffling the content around in some way or maybe some templating right.

00:43:53: so I'm just worried about that right like at some point maybe some some explanation to customers are the same no matter how you write them right whether it's written by air or not if you teach it how to boil an egg, how far can you go, right?

00:44:06: Like at some point in changing that, right?

00:44:08: So, and our fear remains the same, our, you know, our features that remain the same.

00:44:14: So, um, yeah, so a certain aspect of them on less work, there is more about, you know, making sure that then we maintain our quality, you know, and also we maintain control at times on what the, what the process looks like.

00:44:29: And definitely what, what not, what has not been helping on the internal like LLMs is the ideation process, right?

00:44:40: What I mean by that is that, yeah, sure, you can make agents spit out a lot of ideas, but then even before, whether you've been throwing a lot of people at it, we are still informing the next content production with our own data and the past performance.

00:45:00: It was less an exercise that now, thanks to the agents, we are doing this better.

00:45:06: So there are aspects of the production that were already quite good with the data pipeline that we had to inform.

00:45:15: next article, next article, next article.

00:45:17: It's not that having the agent made it better.

00:45:20: Yesterday, I was laughing, someone posted on, I think it was on Twitter, X. at that perplexity or one of these platforms has got when they do the kind of share price pages that now they probably pull with some API data and so on.

00:45:40: They tell you inside our trading, right?

00:45:43: At the bottom, they let you know if people from Google have bought or sold their own shares and they announced this as a feature.

00:45:52: And then someone commented that Yahoo finance at this in nineteen ninety seven, right?

00:45:58: So maybe I don't know the specific right, but it's just like make you realize that some stuff it was possible even before, right?

00:46:05: Are you doing it differently?

00:46:07: Yes, right?

00:46:07: Like sometimes it's quite dramatic how differently can you do it, right?

00:46:13: And it might be that the new way of doing it is better than the one before.

00:46:18: I saw a teenager once asking charge EPT to tell them which one of the friends were following her, of the one that she was following on Instagram.

00:46:30: And my answer would have been, let's try to scrape your follower, let's try to scrape the follower, you know, whether the friend is following or not, and let's try to give the answer to it.

00:46:46: JGPD asked to scroll through the follower, record in a video.

00:46:52: She's mental right like five years ago.

00:46:54: you would never have thought that you could answer that problem by taking a video of it right because obviously the ability to read the images was was pretty dead that the ability to understand the task was pretty dead right.

00:47:05: so maybe there are new ways of doing certain things that we will find out right that That are better than the one before or maybe they are not.

00:47:14: Not not as not as good right in the center.

00:47:17: Maybe it's more efficient to to crawl and maybe it's enslaved and then actually doing computer vision and the custom processing that to achieve the same goal.

00:47:26: A little bit we will see maybe with the, for instance now, I'm not sure if HRGPT, they did a bunch of partnership with maybe Zillow or Expedia to basically get their inventory data.

00:47:40: This is a classic example that, you know, if you have a billion of properties with pricing updating every second, how expensive it must be to do this agentic experience if you don't have API connection to it, right?

00:47:55: So at times maybe we will move from like scraping and then like doing other tactic to maybe connect better certain data sources, but then the upside of doing this is that if you were already investing in your own data, infrastructure, then it's not a big difference.

00:48:17: For instance, whether charge EPT will collect our price comparison data by doing computer vision on the pages, by scraping the pages, by API, to us it will be less relevant if you have invested in that data and the structure of the data and those tables and the accuracy and all that kind of stuff.

00:48:42: We are ready to give them an API if they want to.

00:48:45: If they don't want to, they will need to spend more resources.

00:48:47: That's what probably they're doing right now and trying to check our pages often.

00:48:53: But then the underlying job that we are doing in terms of the creation of the data pipeline is kind of the same, right?

00:49:00: At least the guys in the SEO team who who build the data pipeline.

00:49:04: Then there's the other people who do the visualization on the web pages that they maybe might be different, right?

00:49:09: Maybe we need to invest less in that.

00:49:11: Maybe this will be a chart rather than a landing page.

00:49:14: Maybe we will not need to paint every table as much as we paint them now in terms of like UI and design.

00:49:21: That is questionable.

00:49:23: But then the underlying data, or if you eventually want to press the button to move the money from A to B, this will have to pass through Y's API anyway.

00:49:32: So I think there are aspects of the journey that we keep building anyway, regardless of how the user interface will look like.

00:49:42: And for sure, there will be short-term optimization of the user interface.

00:49:47: So for instance, every time there's a new user interface, if you are relatively early adopter, you can squeeze some gains maybe compared to reasons right.

00:49:56: this may be something one reason for some companies to look into Trying to optimize for for certain models.

00:50:06: But then like often we have seen that on the longer run right like it becomes a bit more You know like the experience of user becomes a bit more Standardized and then in that case you you you will probably have less benefit from.

00:50:23: I'm doing certain stuff right and other but we see I think is it's something time right.

00:50:27: I think the fact that we are the first time in the.

00:50:32: In like at least I've been doing a CEO for fifteen years now is the first time that something could change quite dramatically is definitely definitely interesting.

00:50:42: how much will it change?

00:50:43: probably you know we have to see on our side we are less opinionated about that in the sense that.

00:50:50: you know, we are not trying to sell you the new AI tricks, right?

00:50:55: So whether this will be a hundred percent, fifty percent or zero percent of the future, we are fine from that perspective, right?

00:51:04: Obviously, we don't want to miss out on things, but but we always have to be pragmatic about what is there.

00:51:10: One thing that we have done already since a bunch of time is, you know, and that's why also personally, I'm less worried about things is that We have been the risk in search against other organic activity since a while.

00:51:27: And the reason why we did it was before child GPT was at some point, you create a lot of traffic, a lot of pages that do a lot of traffic, and then you rank first.

00:51:37: There's nothing else to do.

00:51:40: It's like unless the wise product team decide to launch a new product line, And it's not that we can rank better from what we already do, right?

00:51:48: So that's where we say, OK, from now on to drive more growth in some area, we either have to diversify like cluster of keyword that we go after.

00:51:59: But the more far away you go from your core intent, the less you will have in term of return.

00:52:06: Another thing is to expand vertically on the page.

00:52:08: You say, OK, do we have more domains on that page?

00:52:12: And then the other is to be completely off the page, right?

00:52:14: Like so, you know, it's YouTube is like content that we don't have a platform.

00:52:19: to some extent now I can see that even in some of the eye overviews they can pull snippet of YouTube video something that in the past they were not By no means that would have been able to understand between the lines of a twenty minute YouTube video.

00:52:32: What was important to you for that query?

00:52:35: I think this quite insane, right?

00:52:36: But we have been diversifying across other way to distribute content since a while.

00:52:43: I'm not saying that obviously this is as successful as our SEO efforts.

00:52:48: SEO is still a gigantic mammoth that is driving growth for a while.

00:52:54: But it's something that we thought even before AI overviews, which eventually is the only thing that at the moment is under pressure.

00:53:13: As opposed to maybe, you know, perplexity, HRGPT and flow, yeah, we don't have the perception, despite the insane growth that this is threatening that much performance on Google search, at least at our scale, right?

00:53:30: Like maybe our company with the less advanced web presence, maybe this is more important to them, but to us, it's still quite small existence here.

00:53:42: You already touched a little bit on SEO tools and programmatic.

00:53:46: And I think it's also a super important aspects of your whole organic growth motion.

00:53:53: So can you walk us through a little bit how you think about SEO tools?

00:53:58: You also called it SEO products in a talk you gave at SamRush Spotlight and how this all ties into this programmatic bucket.

00:54:09: Yeah, yeah, so when I when I started wise At the time, you know, we were trying to Build the pages on the website and get the index as fast as possible and Imagine you have at the time wise it was not even profitable as far as I remember.

00:54:29: so imagine you have a clock ticking and that is telling you, you know, at Christmas, we shut down, right?

00:54:37: And what can you do, right?

00:54:40: If in the short of time, obviously, SEO is not a channel that is generally short term, right?

00:54:46: Like in the sense that we'll be investing in SEO, you know, we build, you know, the day that the money goes out, the return on that money is zero, right?

00:54:54: Then you let it stay out, especially forever green query, and then you weekend that money back.

00:54:59: but also regardless of whether you get the money back or not in terms of investment, it needs a bunch of time to be there.

00:55:06: Something that I understood pretty soon is that we were in the middle of building our technical stack because at the time, Wise website was a monolithic AngularJS application and that application was bundled with the core product.

00:55:30: just to understand the craziness of it, if you were changing a futile link on the homepage, you could, and this was going wrong, you could stop payments.

00:55:40: otherwise, right?

00:55:42: So, like, not to that extent, but... Handoff right like so and this is obviously quite risky these days would be mental to have anything like this in the sense that if we stop even for ten minutes on a fifteen plus billion a month is quite a lot of money.

00:55:58: But but but back in time.

00:56:01: this was the case right.

00:56:02: so we had engineers in the team early days to rebuild our tech stock.

00:56:07: so we build a CMS Fully now and at the time and we wanted to decouple from this monolithic application on the home page.

00:56:17: So we had engineers in the team.

00:56:19: And then we started to publish content on that CMS.

00:56:23: And there were two content types.

00:56:25: There was a CMS for landing pages and a CMS for editorial content, like the blog.

00:56:32: And what we realized, obviously, is that the page creation process was going as fast as we were hiring people.

00:56:40: To make more pages and so like the blog.

00:56:43: scaling mechanics are.

00:56:45: you are one more person that does one more language in one more country that will produce another X many articles a month with the help of freelance writers.

00:56:55: Even though that person improved their efficiency is impossible that they will go from like.

00:57:02: thirty to ten thousand articles a month, right?

00:57:04: Like they will get better at doing it, they will find more serious that can scale, they can become faster at briefing freelancer, they can have more freelancer.

00:57:12: But eventually there's a constraint on people, right?

00:57:14: We should need to hire on board and scale, right?

00:57:16: So we kept doing that, right?

00:57:18: So we kept hiring people to create more pages through CMS.

00:57:22: But what it was obvious is that there was a lot of intents that were repeatable that we didn't need to hire people to create pages through a CMS.

00:57:30: Early days, we were validating the need of those page types in the CMS.

00:57:35: So the CMS had capability to scale page types and even with some API code to build some of it.

00:57:43: So this could have been already a small programmatic experiments.

00:57:47: But then with engineers, the data something was validated in terms of, OK, can we get it indexed?

00:57:52: Can we get traffic?

00:57:53: Does it drive revenue?

00:57:55: Roughly what is it going to be worth?

00:57:56: Before we created the next twenty languages, you know, other five to a million pages worth of those, those keyword cluster, less you work with engineers to make it.

00:58:08: That was the primary reason to do that.

00:58:10: And obviously, again, in wise periods, you know, the more we can scale, the cheaper our acquisition is and the more we can drop price, right?

00:58:16: So I think that's why we, we went down the path.

00:58:18: But then in general, some queries are prone to, to be solved in that way, right?

00:58:23: You're not going to answer the price of pound to euro in articles, right?

00:58:28: You can do that, right?

00:58:29: We even tested that.

00:58:30: We started to at some point to create fresh content on a daily basis, talking about currencies, but it's very costly because you have to, you know, also at the time you couldn't use AI to make any of it, so LLMs to make any of it.

00:58:43: It was very costly and also it wasn't kind of answering the query of having a reliable product to work on that.

00:58:50: So probably years we have built a certain amount of you know, landing page types, calculator products that our customer use in relation to their user-wise.

00:59:02: Some of them are purely product landing pages, right?

00:59:04: So one of our products is sending money and we have landing pages for sending money.

00:59:08: And some other are products that our customer use in the process of doing some of the intent that are part of our product, right?

00:59:16: So our currency calculator is not something that you... You absolutely need to use from wise, right?

00:59:22: You can check the rate on xc Google and whatever and then you can see come to wise to transfer money.

00:59:28: But we thought that was better to do it on our property and also we thought that You know since there was such demand why don't they do it on wise, right?

00:59:37: And obviously this product they have very different intent than than some of the product landing pages, right?

00:59:43: So there are acquisition products that is product that are tangentially irrelevant To your business that they can help you acquire your customer and then there are product landing pages that this is what they actually do right.

00:59:53: this is our service right.

00:59:55: And obviously I'm pretty many pages.

00:59:56: your conversion rate is quite stellar on this.

00:59:59: acquisition products is often.

01:00:01: a volume gain in the sense that the conversion rate on those products is extremely low, because it's not that my father, who is in Italy right now, he might not need to transfer money, but if on the news they talk about the pound going higher or lower against the euro in my Google, it's just to check the weather.

01:00:20: So there's a lot of intents that go literally in the bin from a usage of wise standpoint, but at the same time, there's a lot of other intents that we can add.

01:00:30: customer and then again like you know historically in our industry most of the people that do price comparison they they have a pretty you know not transparent experience right because they they optimize table for pays more on the table not for which cheaper.

01:00:48: and all this behavior.

01:00:49: so we had more intent in like reason to build some of these ourselves.

01:00:54: right.

01:00:55: the reason why we are on price comparison wise.com rather than letting the comparison happen on independent websites, only is that we also want customers to hear it for a stand for months, right?

01:01:08: And we always done this very transparently.

01:01:11: It's not that when WISE is cheaper, it's not cheaper that we remove the other provider, right?

01:01:16: There have been months where Western Union, for instance, does price promotion on certain corridors and they are cheaper than wise on wise on page which is kind of mental right to think about that.

01:01:29: right they're.

01:01:29: usually no brand would put their main competitors on their own page or one of the key lending pages.

01:01:35: but we you know we have educated our customers that on our side we'll find the transparent comparison and then you know.

01:01:42: so this is the kind of work that that we've been doing throughout the years.

01:01:46: right?

01:01:46: some of it is very SEO heavy.

01:01:48: right in the sense that some of this product they exist because of SEO.

01:01:54: Some of these products, they exist with some SEO angle, but also they are products that our customers use.

01:02:02: This is roughly the same principle as the blog.

01:02:07: Our kind of goal is to educate the customer on why they should shop around when choosing a provider.

01:02:15: If we do that and we increase the number of people who shop around, we're going to be fine.

01:02:20: And this is different maybe from other industry.

01:02:22: Often when you are disrupting an industry, search demand is not there.

01:02:26: So if tomorrow you want to compete with Booking.com, you don't need to teach customers to Google hotels.

01:02:34: There's already a ton of search volume.

01:02:36: That's why Google makes billions on paid search ads and so on.

01:02:40: So you don't need to teach people out to Google hotels, but in the case of At least our original product money transfer now we have a lot of other product line you can spend.

01:02:48: you can you can you know kind of all their money wise account and so on.

01:02:52: but the early early day product There was not a lot of demand because most people didn't even know that they had to shop around.

01:02:58: right.

01:02:58: they were You know using their bank.

01:03:01: They were sending money and then they were using money with the exchange rate and so on.

01:03:05: And this is very different from other industry.

01:03:07: Maybe I don't know car insurance, right?

01:03:09: And it's very likely that you change car insurance without comparing On one of the comparison sites right users.

01:03:15: very likely that you get a credit card without checking which which you know allowances and rate they give you.

01:03:22: right so.

01:03:23: Some of these industry also regulated.

01:03:24: right.

01:03:25: the money transfer industry is not regulated from a comparative advertising standpoint so obviously we are regulated financial company.

01:03:33: that is beyond beyond the case but from how the customer fear displayed.

01:03:39: in to customer.

01:03:40: there's not very strict regulation which means that only European Union.

01:03:44: in a couple of jurisdictions the regulator tells the banks and the financial institution how to show that data.

01:03:51: but in other places you can show the data as you want right which would be quite weird that you.

01:03:57: try to get a mortgage and the mortgage that is more expensive for you is ranked on top as the best mortgage.

01:04:05: But this happens in our industry, right?

01:04:06: So we have a lot of this situation where we say, okay, our job is not to necessarily, you know, capitalize on the demand that already exists, but we need to do everything that we can to have the customer find out that when you convert money between countries and currencies, you should shop around, right?

01:04:23: If this happens, then Wise mission is to over time drop the price become more instance, right?

01:04:29: And so on.

01:04:29: so you generally will end up using wise and this is roughly the case right now.

01:04:33: We are I don't know the latest report.

01:04:35: Maybe, you know, almost sixty to seventy percent instant, right?

01:04:40: Transfer instance mean under twenty second Which is almost faster than some of our banking apps refresh their their interface, right?

01:04:48: And so like back in time was was zero right like when when it started was zero.

01:04:54: and so we we improve our product over time and so we are less less worried about you know on the long run having the best payment infrastructure the best product.

01:05:03: we are more worried that the fact that if people don't shop around they will always you know use the current provider and then be agnostic to it right and sometimes you have to.

01:05:14: it's not just with banks.

01:05:16: we had many times where new fintech they come into the market they price zero for a few months to acquire new customers, a little bit happening in other industry, assuming the utility bills happens.

01:05:28: There's always a new electricity company asking me to have a discounted month and then the month after becomes more expensive than my current one.

01:05:35: So these tactics are being applied.

01:05:38: So it's not that all the times you're going to be winning on price and service, but on the long run, if wise keeps stopping price, keep becoming faster, become more convenient with payment methods, then we should be okay.

01:05:55: At the condition that customers know how to search and know how to shop around for the product.

01:06:01: Maybe just asking the right question to judge GPT.

01:06:04: So just the fact that you know that transfer money is not free, as some of the people say with those big asterisks, will make you chance to judge GPT and say, hey, I know that spending money is not free.

01:06:16: How should I do it?

01:06:17: Who should I consider?

01:06:18: I think this is what we try to do with marketing rather than try to necessarily optimize the last mile, if that makes sense.

01:06:27: I can imagine a lot of people wondering if you're operating in a sensitive industry, let's say fintech, banking, maybe insurance, but maybe also something like health, medical industries.

01:06:42: You already also mentioned that you have to bring in compliance or you have to work with legal, you have to be compliant.

01:06:48: So you have to make sure that things are correct and that things are like legally sound.

01:06:53: And I wonder, and I would like to focus on this, this aspect, how do you organize this collaboration to not lose speed?

01:07:02: Because obviously speed is always something for a fast growing company we care about.

01:07:05: And then I think if you hear the legal team has to be involved, a lot of people will be like, Oh, damn, this will kill speed completely.

01:07:14: Yeah, yeah, yeah.

01:07:16: I would be honest on this, right?

01:07:18: Clearly, throughout the journey of any company, you learn the level of risk increase and the level of how serious you become about managing the risk gets better and better.

01:07:36: So, you know, clearly our process now is different.

01:07:40: from our process of ten years ago.

01:07:42: Now we are a public company.

01:07:44: We are regulated in multiple jurisdiction.

01:07:46: We have a different level of risk.

01:07:49: We hold, you know, billions of customer deposit that we didn't know the early days.

01:07:54: So it's normal to think that things slow down a little bit to some to that extent.

01:08:01: And, you know, on day one, there is frustrating, right?

01:08:04: Because obviously, as marketer, we want to go as fast as possible.

01:08:09: On day two you learn that maybe some of that risk management will help you save time later because obviously the bigger you get the more some problems might occur if you kind of disregard that risk.

01:08:21: So we time the team is trying to put more effort into safeguarding ourselves from a compliance and legal perspective.

01:08:31: But then in terms of like how do we keep it manageable for the teams?

01:08:36: So two things, one is Historically, we have tried to basically work with legal and compliance to build processes within the teams as opposed to having these processes run elsewhere by someone else with a different operating model.

01:08:56: And that is quite essential because my past experience, before wise, the reason why I left consulting is that I was working with HSBC.

01:09:05: And the incentive of the compliance team was to arm more compliance people and make it slower for you.

01:09:13: It was not to help things.

01:09:16: It was just a different incentive within the organization.

01:09:20: As we have a compliance team that is working within the marketing org and working team by team, obviously all teams have got different processes from paid media to editorial content creation and so on to try to Workout out to best integrate this so that a they spend more time together and collaborate and be that this fits our Our process and doesn't get completely disrupted at times also.

01:09:47: I think is the case of Taking decisions right on certain stands.

01:09:53: What I mean by that is you know we When it comes to risk management at least from from my personal experience There being times where There's one thing that the regulator has said, or at least what was said by the regulator could have been interpreted in different ways.

01:10:16: And obviously, there is always the safest way, which is do nothing.

01:10:21: So if you don't do anything, it's the safest way.

01:10:23: Clearly, this is not.

01:10:24: possible, right?

01:10:26: We have growth targets, we have certain ambition, right?

01:10:29: So the best is to try to work with the compliance team and legal team to really go deep on.

01:10:37: where do we set our risk appetite, right?

01:10:40: Which doesn't mean taking any risk for the customers themselves, right?

01:10:44: So this is like wise product compliance.

01:10:47: This is all sort of you know, it was sort of AML, KYC, the last product that we're talking about marketing material, we're talking about all the marketing real estate.

01:10:56: We take certain risks with the view that then we can back up those risks and we have the right rationale for doing some action, right, which doesn't go against what the regulations say, but takes one on the interpretation of the other, right, depending on what.

01:11:13: What.

01:11:14: what we think at times is valuable.

01:11:16: and I'm not saying that we know this means circumventing the rules and just saying that.

01:11:22: Unfortunately in some countries The interpretation of some rule can be quite open, you know, even in comparative advertising.

01:11:33: Sometimes you go through case studies of past experiences rather than having clear legislation about this.

01:11:39: And so we need to always have some level of work of interpretation with our legal team.

01:11:44: And obviously we try to find compromises, right, on how we adhere to the regulation and obviously ultimately not mislead customer, right?

01:11:52: So like if this is like by default advice, right?

01:11:55: So there's nothing that we do to mislead customer.

01:11:58: If you have confidence that we're not misleading customer, that we are accurate to customer, Then, like the interpretation of the regulation kind of goes in line with that, you know, unless very specific situation.

01:12:09: Then you have some markets where the work is particularly challenging.

01:12:13: So, for instance, in one market in the Middle East, it's not possible to even mention competitors on your websites, and clearly we do a lot of comparative advertising.

01:12:26: In other markets, there are restrictions on advertising against each other.

01:12:30: So there are market specific rules that are quite a blocker for the team.

01:12:34: It's not that we can do anything about that.

01:12:37: That's the playing ground that everyone has to play with.

01:12:41: But then in most of the other markets, this journey is not a massive blocker.

01:12:47: I would call it getting serious.

01:12:48: What I mean by that is why is... getting in trouble with the regulator, you know, fifteen years ago, because of certain activities was definitely less painful than what would be now.

01:13:03: Therefore, our level of risk, you know, risk appetite has decreased significantly, right?

01:13:09: And maybe made some process a bit slower, but not to the point where this is like, this is impacting the people in the team.

01:13:19: I hope that with LLM, Certain activity could be, to some extent, edged.

01:13:27: Not that we're going to delegate compliance to a legal LNM, apart from that.

01:13:32: But it's just the element of monitoring the accuracy of data, monitoring certain facts, going back to the sources of some of our claims and check the validity of the sources.

01:13:49: There's a lot of operation that was going around being compliant that I think we could automate and become to some extent even more reliable than what the team at time is doing, right, if we automate this well.

01:14:04: So again, it's not changing the interpretation of what is compliant and what is not, but is speeding up the operation on the team on making this course, right, by giving them faster comparison.

01:14:18: One thing is to is to, you know, if the source of this article is out of date, one thing is you having to figure it out, go to the source, check if it changed or not, find the new resource, and then try to update the article.

01:14:33: Another thing is that you get a notification and you get a side-by-side of the new versus the old one pointing you out the difference, and then you can make that call like instantly, right?

01:14:43: So this is the type of automation that I envision in the sense that We help our team to spend less time on what they would have not enjoyed before, and still we can be compliant and follow the legal guidelines that we have agreed in all the markets.

01:15:00: We have a legal playbook for all the markets that we operate with and particular guidelines for compliance as well.

01:15:08: So every person working on a different market, they stick to slightly different.

01:15:12: but broadly the process is the same within each team, right?

01:15:15: So the social media team doing content on organic social, it's probably one process, same like for the editorial team, same for the S team, right?

01:15:25: And then also collaborate with compliance and legal.

01:15:28: I think this is completely fine.

01:15:33: And I would envision that even in, I mean, I think that even in other industry like the pharmaceutical industry or medical industry.

01:15:43: At the end, you know, it's part of the customer trust exercise, right?

01:15:47: Like, you know, especially in FinTech, you know, you're a new financial provider, you have to get people to trust you with your money, with their money.

01:15:56: And, and what better than making sure that they they're all the signal that that the people can trust you, right?

01:16:04: So, for instance, in our price comparison data, we have quite a lot of quotes from or customers, well, a non-customer that, you know, even though they don't transfer money with Ys, they check Ys price comparison.

01:16:18: And the reason why they do it is probably because, you know, we show the data in the back so that they can see whether that is real or not.

01:16:26: The data is really up to date.

01:16:28: The data is displayed transparently.

01:16:30: We don't make claims that are not correct, right?

01:16:33: So you over time build trust and then that ends up probably being Over time compliant more and more and more because you have been building trust.

01:16:43: often the you know legal and regulator from a compliance perspective They want to make sure that you are protecting the customers right.

01:16:49: so as long as your company mission is aligned to that I Think generally goes okay.

01:16:55: There's some work to do, but the journey goes okay.

01:16:57: Otherwise it becomes a bit tricky, you know if you want to scam customers then Yeah, I wouldn't advise you.

01:17:02: but some companies do right to some extent and And it's their choice, right?

01:17:06: So for them, I'd be more tricky, right?

01:17:08: That makes sense.

01:17:11: a lot for people to take away from that.

01:17:13: Now, I know you left consulting on purpose to build something as remarkable as wise, but I would like you to imagine being put into consulting shoes once again, because I always want to make this podcast like really actionable and practical, and you already delivered a lot on that.

01:17:34: But if you imagine yourself being a consultant to a fast growing FinTech, or SaaS company in like a banking space or financial space.

01:17:44: And they are building their organic growth engine basically from the ground up now.

01:17:49: What would be like your two or three top recommendations for them to take away?

01:17:55: Yeah, yeah.

01:17:57: I haven't done this a lot in the last year, but in the last few years, I have spent times like a very limited time.

01:18:03: We're talking about one hour every few months with some startups and company just.

01:18:08: either through connection that we have, X, Ys and so on.

01:18:11: So I've been in those conversations quite a bit.

01:18:15: So the first aspect that I noticed that I think worked well with us is we tried from early days to separate as much as possible the disciplines to then hire specialists in those disciplines.

01:18:37: I understand that on day one you might need to make a person space across different skill sets, but then we have seen acceleration and growth in a given discipline where we identify what the boundary of the discipline is.

01:18:54: We have given an investment constraint and we have given a specialist to do it.

01:18:59: This has been across all the teams that we run.

01:19:02: We started with the CEO, Then we started to build, we worked on affiliate domains, we're doing like social contents around.

01:19:14: All this time, I literally either executed myself and learned that craft each time in very much details or I have hired to work with me, some specialists in that area to grow that area.

01:19:29: The reason why I mention this is because often, even with regards to traditional SEO, I was often chatting to people and they were asking me, okay, what should we do?

01:19:38: You know, there are some consultants that are quite good in basically outlining you roughly the plan, but then there's an element of execution that you need at least in some disciplines.

01:19:50: specialists to collaborate with those consultants and those people internally that can move this area.

01:19:55: So whether someone runs fully now or runs with a mix of collaborators, like consultants and agencies, at least one specialist on the other side is important.

01:20:06: And sometimes I advise even the agency to help the companies do that, because if you don't have the right person to deal with on the other side, because very tricky.

01:20:15: You often be explaining what what your specialist about to someone that is not and your success varies a lot right.

01:20:23: so that's what we do and I spend a bunch of time helping some of these friends in other startup hiring their first hire right.

01:20:30: and to do that I generally advise to to basically get a consultant or get someone.

01:20:36: that is really especially so.

01:20:37: if you never did search don't try to hire our core search person, let another core search person do that.

01:20:44: That's generally the case.

01:20:46: I think that's exercise number one.

01:20:47: Second is, even what's happening with LLMs and the AI, if before, there was the pressure of doing a CEO faster, because you wanted to grow faster and be profitable sooner than later, I think now is even more because there's a little bit of threat that some of the equity or some of the content that you build or some of the user journey that you think that will go through template links will come maybe not in the same way through the same template links.

01:21:19: I still advise some company to invest but then A to be very targeted on where to invest between the spectrum of queries that they cover and two to try to be as fast as possible.

01:21:31: What I mean by that is you know, LLM is not just powering AI overviews and charging PT, but it's also like technology that you can use across the team to really try to scale.

01:21:41: If you're not scaling very fast, I think that today, compared to the past, and if you're not working with someone, maybe that helps you scale really fast, maybe some area of search, you should be giving it up or say, you know, I'm not going to be as fast to capitalize on this, and then maybe I should start diversifying a bit quicker.

01:22:02: to something else right like so by diversifying i mean youtube right like vertical video across all vertical platforms all sorts of investment on doing better crm.

01:22:15: you know for instance the first five years of wise marketing.

01:22:18: we have not invested in crm not in a serious way because we had pretty much we had to acquire.

01:22:27: we didn't have customers we didn't have anything to tell to the existing customer because we didn't have them right.

01:22:32: But then sometimes when you are in the mood of trying to get more and more customers, you discount a little bit how much you can get out of the one that you have in terms of value, and also how much experience you can give to them.

01:22:46: So probably in these days, in age, I would be less relaxed about the traffic that I have.

01:22:56: that doesn't convert.

01:22:58: or that converse, but it's not so valuable because maybe you have not done the right, you know, CRM or the right, you know, engagement tools that you could be building.

01:23:07: So I think this is the other bit.

01:23:09: And to some extent, we're doing this advice as well.

01:23:11: So in the last five years, we have doubled down on like, you know, we sit in a million of visits every month, right?

01:23:17: And so on this ton of traffic that comes to us.com.

01:23:20: And, you know, it's obvious that, you know, it's more impactful in the short term.

01:23:25: to make the most out of that traffic than trying to get a hundred percent uplift in AI.

01:23:31: mention right because we have so much traffic right but I think that the the aspect of doing building your own properties it might be true because in that way so which means the risk in as soon as possible a distribution channel.

01:23:49: right in the past because search kept coming And people kept searching.

01:23:55: We had less need to diversify one distribution channel that was working.

01:23:59: But now the shelf life on some distribution channel have been way lower in time than what it used to be.

01:24:09: So even the wave of growth that people had on Instagram or TikTok, YouTube may be different, but on Instagram and TikTok, it's not that this day TikTok is growing like you know, ten years ago, five years ago, right?

01:24:22: So this was less the case for search, right?

01:24:26: If you were to build an authority with domain twenty years ago, you could still be here now, right?

01:24:31: By doing a good job, right?

01:24:32: Obviously, I keep investing, right?

01:24:33: So I think the aspect of like trying to feed the lies and like, and get the customer to understand better your product, better your messaging, maybe you are not spending enough time on how to pitch.

01:24:48: your your customer what you are doing good your product was doing for them.

01:24:52: Just because you knew that next month there was another ten thousand of them coming that you're the same chance of converting right.

01:24:58: but so I think that's what I would advise just in case the shift of channel mix become become that that relevance right.

01:25:08: one exercise that that we did is is is you know going through companies and see okay if tomorrow.

01:25:17: Google disappears and we are a hundred percent AI mode, HRGPT with no click to the website and no traffic coming through.

01:25:29: Who is it that will not care about it?

01:25:32: People that won't care is Instagram, people that won't care is LinkedIn, people that won't care is a bunch of other apps that we decide to use.

01:25:43: that have built a reason for you to do something with those apps beyond the distribution channel.

01:25:48: I'm pretty sure that even Facebook at the time had some distribution through Search or some other way to find distribution, right?

01:25:56: But now they've given customers a reason to engage with their properties beyond that.

01:26:02: I think that's my main advice to people.

01:26:06: In the past, maybe it was a bit easier to not think that are about this and just and just build distribution.

01:26:17: Yeah, makes sense.

01:26:19: Fabrizio, we're already way beyond the hour mark and you already dropped so much wisdom that people have to process.

01:26:29: I always like to direct people to a platform or a. even saw that you went to some conferences where people can follow around and learn more about my guests and see more of what they're doing.

01:26:48: So what is a good place for people to follow you?

01:26:51: or what are maybe conferences where people will meet you?

01:26:56: On the conference, I haven't decided yet where to be next year.

01:27:00: to be honest with you.

01:27:02: I usually go a bit with the flow where my meetings in various offices wise bring me.

01:27:08: I think the best way and generally I share this on socials is to follow me on LinkedIn and ex, right?

01:27:15: But it's a ballerina.

01:27:16: It's my name.

01:27:18: just by googling that you will get to both.

01:27:20: And generally on those platforms, I share either content and where I will be.

01:27:26: physically in terms of events.

01:27:28: So we'll probably start sharing updates on that early next year.

01:27:32: Now we are about to wrap up the year and then see the family back home.

01:27:37: So we have less conferences.

01:27:39: I've just been to two conferences now and that we'll continue next year.

01:27:43: Generally I've been attending the usual SEO conferences that have been going for quite a while in the sense that I still I speak of why it's be out to SEO conference, but also other leads in the team do, right?

01:27:56: The people who lead some of this SEO team under my umbrella of teams and then also other type of conferences in the marketing space because more than before is probably interesting to, you know, get ideas and build a bit the bridge between marketing channels.

01:28:14: So I've been attending a bunch of other conferences that are as SEO specific at times or meetups.

01:28:20: Yeah, and I saw that you're also, I think also currently, but probably also like, from time to time, there are always job openings.

01:28:30: Also, I think from your team, right?

01:28:32: Yeah, I mean, it never stops.

01:28:34: It never stops.

01:28:35: Yeah, so you're always hiring.

01:28:37: So, why is it always hiring?

01:28:39: You know, like, I think there is over six thousand people, if I'm not wrong, at the moment.

01:28:45: My team has been growing for just me and a few engineers to a lot of people.

01:28:49: We mostly hire more because maybe we're growing into our markets, but also growing into our product line, or otherwise just some of the teams are doing well and then we decide to invest more.

01:28:58: So, Wise.Jobs is the best place to be up to date with new openings.

01:29:06: Over the next few months that I'm aware, we will be hiring in our Singapore office, Tallinn and London.

01:29:14: and across various team from influencers to SEO and CRM.

01:29:20: So we have a bunch of open roles across marketing spectrum.

01:29:24: So cool.

01:29:25: So for everybody that is looking maybe for a new career opportunity or that's just inspired by what Fabrizio and his team is building or like the vice mission, I think there's a lot in this mission to really sympathize with like getting fees basically to zero.

01:29:40: So what's not to like about that?

01:29:42: Go check out wise.jobs.

01:29:44: We also put the link in the description and see if any of the open roles fit you.

01:29:51: It has been a pleasure.

01:29:55: You shared so much of the vice journey.

01:29:58: I wish you personally all the best for the future.

01:30:01: I wish also like the whole vice team and the company all the best.

01:30:05: I will continue to be a follower from the sidelines and to see where you guys are heading and if you are still.

01:30:14: like advice, it takes ten years in the future.

01:30:17: We should probably do like an anniversary episode then.

01:30:22: Maybe let's try to get to the five year mark.

01:30:24: Yeah, let's do five years.

01:30:26: Would also be awesome.

01:30:27: I think you're it's a very special thing to be so long at the company that is growing so fast.

01:30:33: So you must have gotten a lot of things right.

01:30:37: So appreciate taking you taking the time today to speak with me.

01:30:41: And yeah, see you around.

01:30:44: Yeah, thanks for having me.

01:30:44: It was great to chat here.

01:30:46: Thanks.

01:30:47: Bye-bye, Fabitsu.

01:30:48: Bye-bye.