NOLL TILL 100 — ENGLISH TRANSCRIPT
Episode 101: 101. Building ecommerce in turbulent times: AI, forecasting and the systems you need

English translation of the full Swedish conversation. Original timestamps preserved.
Original audio: https://www.youtube.com/watch?v=AS49ddvu_Mw

[00:00:01] Welcome to 0 to 100, the podcast about how to take an e-commerce company from an idea to over 100 million kronor in revenue. We are Fedja and Jacob, and we will give you all our best ideas for growing your company. A warm welcome back to a new episode of 0 to 100. As usual, I have Fedja Porobic with me. Thanks, Jacob. Great to be back for yet another new week, as we usually say. New week, new start. Exactly. It is always lovely when it is a new week, but everyone else celebrates when it is Friday. I celebrate when it is Monday.

[00:00:35] I know. Same here. I think, somewhere along the way, you have come to realize that one of the most fun things there is is e-commerce, and growth, and performance, so Mondays are great fun.

[00:00:48] Absolutely. Easter is coming soon. Yes. Undeniably. So, yes, when you hear this episode, it is actually Easter. It is actually Easter or Easter week. But that does not mean very much, because we record a few days before you hear this, and now, just a week or so ago, you also heard our hundredth episode, where we did it live, but that was actually recorded in February. We do not usually record that far in advance. No, that is very unusual.

[00:01:18] But there were some delays there in connection with the Iran war and you getting stuck in Dubai, so yes, I can understand that a bit. But 101 episodes, that is usually said to be, in a way... What did you say? That American textbooks are

[00:01:34] Yes, but I suppose... I think it comes from the American university system, where course 101 is the introductory course in a new era, or whatever you would call it, a new subject.

[00:01:42] Yes, that is... It fits today's episode nicely. We have absolutely loads to talk about, as usual, and a lot is happening in the world around us. We could touch on that a little, because a lot has happened since last time, I think. Especially this business of the major banks raising their interest rates. The Riksbank, after being very positive about the market, is shifting toward a negative outlook.

[00:02:05] A lot is happening around these sponsored terrorist groups in Yemen, which are now threatening to close another strait, and that makes trade deliveries harder, above all. Exactly. But where do we start, the banks perhaps? Yes, well, what has happened is that we have moved from a situation where perhaps there was a consensus that interest-rate cuts would continue this year to there now being a high risk.

[00:02:33] Or I think the Riksbank's position is probably that there will be neither cuts nor increases this year, but the market is starting to price in actual increases this year. And this is not an area of deep expertise, but I mean, that is at least partly on the back of the Iran war, which increases the risk of inflation. And when oil prices rise, for example, that increases prices broadly across society. We also import inflation, both through currency movements and potentially in other ways. So there is a risk that we will see inflation, and then interest rates need to be raised in response to that. And that is the main thesis, I suppose. But one thing it clearly depends very much on is how long the conflict in Iran will continue.

[00:03:23] Yes. And that is highly unclear today, but we are now closer than before to ground troops being deployed. And historically, when the US has had ground troops involved, it has not ended wars very quickly.

[00:03:41] No. Or, yes, they drag on, at any rate. Yes, exactly. But it depends a little on how you look at it. But of course they did not finish Iraq or Afghanistan very quickly in terms of, well, they still had troops on the ground, but there was not necessarily an ongoing war in Afghanistan for as long as the US had troops on the ground.

[00:04:05] Mm. But on the other hand, Iran is big, has an extremely large population, and tricky terrain. So even though there is talk of deploying ground troops, I feel that, given the political consensus in the US around not wanting ground troops in Iran, I find it hard to see them deploying ground troops, possibly with the caveat that they might do so to secure the Strait of Hormuz in some way. That they deploy ground troops locally there and try to sort things out there. But then there is also the aspect of whether Donald Trump manages to convince NATO, for example, to come and help secure the Strait of Hormuz. It is not entirely inconceivable that he

[00:04:52] plays hardball and says he will stop helping Ukraine unless the NATO countries start helping out in the Strait of Hormuz. Mm. Difficult. At any rate, however you look at it, it is a tough situation once again. It feels like we are being hurled into yet another year of chaos.

[00:05:11] No, I agree. It is a bit messy. Once again, when you thought, you know, now we can see light at the end of the tunnel, the tariff worries, these taxes and import taxes and everything last year had calmed down, even though it has been a little chaotic around that this year as well, the market has not reacted very much to it.

[00:05:29] And then we had Greenland as a little interlude there, and that calms down. Then it is like, right, now we are going into yet another war. So, really, ever since COVID in 2020, the market has been very turbulent.

[00:05:45] As far as I know, at least, we have not had, like, 12 months of stability where you can forecast in any normal way, so to speak. What is normal forecasting when this has been going on for six years? It feels like the new normal is volatility.

[00:06:04] Absolutely. No, but it has become very volatile, and parts of the macro... I mean, it is volatile in many ways. For one thing, the stock markets are very volatile

[00:06:14] in a way that perhaps you did not really see before. Then things like tariffs or whatever can suddenly change the playing field in a single afternoon in a way you perhaps did not see as much further back either. So there are several aspects like that which make it tricky to think about how to plan ahead. That is what makes it very difficult as an e-commerce business, in a way, because for many years it has been a global standard that, yes, globalization is the new norm. Everyone is pushing for it. It has become more cost-efficient and better, in a way, to produce things in Asia. The quality has often been better. After COVID, it gave you pause for thought: could we bring the factories back to Europe, at least?

[00:07:09] Yes. And now I almost think that, if you do not have it on the boardroom agenda, something is a bit strange if you have not tried to bring the factories back to Europe, because now we see how exposed we are once again when our sales are, well, our main base is in Sweden and we sell in Europe or the US and have to import things from Asia. When new terrorist groups in Yemen are sponsored by Iran and threaten to close another strait, making the situation considerably worse. But the thing is that, even with production in Europe, there are very often global supply chains further upstream. So when you buy cotton or whatever it might be, perhaps now... Yes, of course it depends on where you get your cotton from, but there is a significant risk that you have exposure to global

[00:08:03] value chains even if you do not at the tier 1 factories. Yes. And the only way around that kind of exposure is to make sure the factories hold stocks of [unclear in the source: en score], really.

[00:08:16] Yes, absolutely. Supply. Absolutely. No. And sometimes you can pay for that yourself, taking that upfront investment, or if you are a loyal customer and have been there a long time, the factories might be willing to hold stock.

[00:08:31] Yes. And take a risk, and make that your main negotiation with your factories this year, in a way. Instead of trying to get a krona off per running meter of fabric or production, it is much better to make sure they have the raw material to produce, or the chips or whatever it is you produce, that they have it.

[00:08:55] One hundred percent. No, but it is a hugely important aspect, obviously, so that is needed. Junipers was largely out of stock here in January and February. That is a shame. Under the circumstances, it still went quite okay. But that was partly because they did not have stocks of raw material, so we had to wait for that to come in, and exactly. And we, yes, finished the year a little stronger than we expected, and the combination of those two things created a squeeze.

[00:09:24] Yes, but since we are talking about Juni anyway, let us leave the macroeconomy for a moment. What we can expect from the macroeconomy is that this year will be volatile. It will be very tough to forecast how it will end. And what we thought might be a comeback for e-commerce could once again become very tough for e-commerce. Even though we see bright spots in Q1, we do not see the effect of what is happening in the market until, like, one or two quarters after it has actually happened.

[00:09:54] Mm. And now it is only a forecast that, yes, okay, the major banks are going to raise interest rates three times this year. But people's finances are not affected immediately when variable interest rates rise. For one thing, some people have fixed rates, and for another, those with variable rates will only see an effect after that.

[00:10:12] Absolutely. Yes, in any case, even if you have a variable rate, so to speak, you have at least three months, a grace period before that. But of course you might be at the end of that, and now when they

[00:10:22] go out and raise them... We should also say that the major banks are raising rates now without the Riksbank having raised them. And that is surely a reflection of how the market is moving, so to speak.

[00:10:34] But it is not yet a given that the Riksbank will raise rates, even if it perhaps looks more likely now than it did before. Yes. No, it is not, but it is a strong indication if the banks raise rates before the Riksbank does. Then they assume the Riksbank will raise rates.

[00:10:51] Yes. And they make money as well. And they make money before it has happened. So the banks are securing their share. Mm. Exactly. But it is not so strange, in a way. And we e-commerce businesses should be thinking like this too. If we are losing margins somewhere, we need to make that back. Now I know there are big discussions about food VAT, which is coming into effect again, or being reduced, and that

[00:11:18] is it April 1 it comes in, or April 1, so that is now, when you are listening to the podcast, food prices in shops should have become 6% cheaper.

[00:11:29] Yes, that is not entirely certain. Some brands, like a company we have invested in, Mylla, we went out and lowered prices and transparently showed how we lowered them by 6%. And then it is up to the producer whether they choose to raise them or not, if they have increased costs elsewhere. But with the food VAT specifically, we show that, okay, this is how we have lowered them. But I find it hard to see grocery stores lowering prices by 6% on average.

[00:11:56] It really is not a perfect market either. I mean, I have a Coop store near me, and that is where I go. It is not because I actually think the Coop store itself is absolutely brilliant. It is just much closer than the nearest ICA store, which is actually a much better store.

[00:12:10] Yes. But above all, I do not go around benchmarking what I pay for fresh basil in every store before making a purchase decision. No, they set the prices they have. Of course, if I noticed one store was insanely expensive compared with the other, perhaps I would avoid it more. But I imagine quite a lot of people have roughly the same relationship with the grocery stores they use.

[00:12:31] Then I think some products are so volatile in themselves. I mean, tomatoes go up and down in price. Cucumbers, then some things sell out, and meat goes up and down in price, and it is difficult to benchmark those particular 6%.

[00:12:43] Absolutely. Perhaps you can see it on some product whose price is stable over time and that is never actually on promotion. No, but no, exactly, because even if my Coop store lowered prices by 6%, I probably would not notice, actually. You might perhaps notice over time, but

[00:13:03] if they lowered prices by 6% today compared with when I shopped there yesterday, I probably would not discover that myself in the store. Perhaps other people would, but I probably would not reflect on it.

[00:13:14] Yes, but the idea of this is that, yes, it was supposed to be a very good year, with parliament doing what it can and the government doing what it can to stimulate the economy, and now

[00:13:24] and interest rates were supposed to come down originally. Interest rates were supposed to fall, and now it is the opposite. But the mortgage repayment requirements are changing now, though. That is also now, that was April 1.

[00:13:35] Yes, they want to do what they can. Yes. So that will mean quite a lot of people, perhaps especially younger people with higher borrowing, will have the opportunity to repay a little less money each year. If you previously had to make repayments of 3% a year, I think the new rules now mean you repay two instead of 3% with a loan-to-value ratio over 70%.

[00:13:56] That can still make some difference. No, it does not make a big difference because it is usually your biggest expense. I mean, as a family with children or a first-time buyer, you have a big expense when you... Yes. Interest. Exactly. Repayments are admittedly a form of saving, but in cash-flow terms you will still feel the difference.

[00:14:17] There will be a difference in itself. But as I said, the government and parliament are doing what they can, and now the market looks set to change, and it will be volatile. So we simply have to prepare ourselves for the next two to three years probably looking the same.

[00:14:33] Yes. So planning two years ahead feels... Of course you should have some kind of budget and scenario for that, but it still feels like you need to revise it monthly, or quarterly at least.

[00:14:47] Yes. And say, is what we said would happen actually true? Has it happened? Should we keep assuming it will be that kind of market this year, or do we need to revise? Yes, no, absolutely. That will probably continue. The US is making a lot of interesting moves in world politics and so on, and you have to say it is new for this administration that so much is supposed to happen all the time, with a lot of people being drawn in. Uh.

[00:15:16] Mm. So that is not really likely to go away. We are only one year into this administration so far. Yes. So there are a few years left, and in that time they can make a great many moves or start more wars.

[00:15:32] Let's see. Let's see. Even though the starting point was, I am not going to start any wars. I am going to go in and stop the war in Ukraine. It will end in, like, what did he say? A week.

[00:15:41] One day, I think. Even one day. And I am going to stop these various wars. Then he has managed to do Venezuela, remove the president, into Iran now, and who knows what is waiting. So what I am thinking, for my own part and for the companies I am involved in, is that I will try to revise every other month and say, is this right? Is what we say we are going to achieve in the budget realistic, or do we need to change what the forecast looks like?

[00:16:10] Yes, because you usually forecast the year before, based on how the market looks. You have a certain tendency to have different scenarios, and you are a little overoptimistic, at least if your name is Fedja Porobic. I know you are much more realistic and calculate things in Excel, whereas I think a lot about, okay, if X, Y, Z happens, if these... If we find a hit ad, then perhaps we grow 100%. What do we need to do to find those ads that grow the brand 100%? So I tend to see it as wanting to be a little more optimistic, but now you need to go back to your forecast quite often and revise it. That is one thing, and then there is the supply chain. Can you negotiate with your factories in some way so that we secure the supply chain

[00:16:58] especially for the core products? The hero products need to be... If you have always in stock, you need to make sure they are always in stock. Yes, absolutely. That is my takeaway, at least. No, but I agree. It sounds like [unclear in the source: alldeles utbanker]. I think generally, forecasting and budgeting and so on will probably become more dynamic as we now move into more AI-based forecasting. Because in the past, or still, I think, in most places, budgeting largely sits with one or two individuals, at least in small organizations. And it is a big task done ahead of the coming year, and you try to arrive at something you believe in.

[00:17:44] But when you can bring more AI into this process, then it is no longer quite as obvious that it should only be done once a year during Q4 and then you stick to it. Perhaps it will become more of a dynamic, more fluid process, where you say, given what has happened so far this year, what do we think about next month? And you are constantly working with an updated forecast

[00:18:05] where you can then choose, as you also indicate, how ambitious you want to be. But it is also pointless having a budget if every month you feel it is completely impossible to achieve it.

[00:18:18] It just demoralizes everyone working with it anyway. So, okay, speaking of forecasting, it is incredibly difficult. As far as I know, I have not tried any app lately that solves, quote unquote, forecasting for the company. Do you know anything good? No, I think the closest you can get now is probably to have all your data connected somewhere and then get help from something like Claude Opus or similar and try to build a forecast through smart prompting. Of course it can be done, and you get fairly reasonable help. But it feels extremely advanced. And first you have to connect all the data, then you have to... What they do is basically connect all the data and then put a wrapper with Claude on top.

[00:19:10] Connecting all the data is not quite as complicated as perhaps it used to be. We have connected our customers' data now through [unclear in the source: Windsore], which then goes into Claude. It still has some problems processing Shopify data, we have noticed, because Shopify spits out data in a very impractical way. The data files get very heavy

[00:19:33] so it struggles with the live calls. Really, you probably need to store that in some database like BigQuery or similar and do that perhaps once a day, which I know Dema, for example, does. No idea whether it is BigQuery, but they use something, at least, where they store data, which means the platforms work relatively quickly after all.

[00:19:53] And that is one reason Lifetimely has become so slow, probably. Yes, because they are trying to handle the live data, and it is incredibly slow, but I find it hard to understand that they have not thought about how to handle that instead.

[00:20:05] They have surely had some requests about it, and everyone is complaining about this now. Yes. So I do not know why they have not solved it. No. Very strange. But no, so that is one way to do it, at least. I have not actually seen any good software specifically for financial forecasting. There is modeling cohorts, making an existing-customer cohort model. That is fairly easy and works fine with Claude Cowork or with many tools. I mean, Lifetimely has okay support for it. It does not work amazingly well, but it is there, at least. You can probably get it out of [unclear in the source: urma] too if you talk to the chatbot there. The harder part is trying to figure out, if we

[00:20:49] spend 600,000 on Meta in May and 200,000 on Google, how many new customers will we bring in? If we spend a little more, is that better for us? If we spend a little less, is that better for us? And generally there are very poor

[00:21:05] yes, systems for that, at least from what I have seen. There are very poor systems for it. And that is something we are actively trying to work on solving internally. You will probably never arrive at a perfect truth, but the fact remains that when you work with a company for a long time, you develop quite a strong gut feeling yourself about how to do this. And I have also built some Excel models on this theme before. If I can do it well for companies I know because of gut feeling, then gut feeling is just a

[00:21:32] type of experience. Yes, it is just a type of experience, but it is data-informed, and that shows, in a way, that obviously this can also be done with pure data.

[00:21:41] Mm. So I think we ought to be able to build something reasonably sensible for that. And it is really only once you add forecasting to data systems that the data system can, in some way, have a chance of telling you whether you are on track or off track

[00:21:59] because otherwise the system becomes very flat. It is just, yes, this happened yesterday, and you are like, yes, okay, great. Even if you add a language model, if it does not know what you wanted, it is pointless. And of course you can add your budget from the year before and so on, but then there are other problems. So this is at least one of the things we are trying to solve but have not yet solved well. But I think we will have a good solution to this before, hopefully before the summer.

[00:22:24] Before the summer, that soon. Yes, I think so. I love having that level of ambition. That is aiming high. Perhaps it will turn into September just because you say that. But I think we have a

[00:22:34] But I still just have to say why it is exciting, because you are trying to tackle an incredibly difficult problem. Yes. If it had been easy, there would be loads of apps. And I mean, there are loads of apps today for what Lifetimely does.

[00:22:46] Yep. It is really visualizing your own data in a nice way so you can see it, and then they have added a few different models and some kind of cohort calculation, as you said. It is not that advanced. This is much more advanced, because you also want to take into account what your budget looks like. How have you performed historically? Do you have seasonality? Yes, well, yes

[00:23:07] exactly. What does the season look like? Do you have enough things in stock? So it has to do a great many calculations and trend lines, and what if something happens, stock runs out in October, causing a dip because of inventory. But there are so many variables in the whole thing. It is not a simple mathematical calculation or a formula or algorithm, but an incredible number of mathematical calculations in parallel. And then how much would it have been if you had had stock in October, when it ran out, and how much will it be in October this year? When do we need...? And I do not know, it is very difficult. One company that does inventory forecasting well is [unclear in the source: Maden]. We have been using [unclear in the source: Madden] at Mockberg since around Black Week now, and it has worked really damn well for us. It has basically

[00:23:56] replaced a buyer, because we were thinking, should we recruit a buyer or bring in a tool, and then we brought in [unclear in the source: med den]. It has worked really damn well. We are nearly into May, and with data behind us it has worked. But it is basically the only tool that does some kind of forecasting relatively well.

[00:24:17] Normally, it has required a full-time position, someone working full-time in purchasing. But even if you do not have [unclear in the source: MEN], or even if you do have [unclear in the source: MEN], you still have the issue that, okay, a lot is happening in the market. How should the budget change? You need to... Yes, everything apart from inventory. Inventory is only a small part of the equation, which of course affects sales

[00:24:41] a great, great deal. But you have other parts of the equation that you have to take into account. But in a way, you have to accept that all models of the future are always wrong.

[00:24:55] That is always the case. I mean, if you make a model and try to predict the future, you will be wrong. But if the model is good enough to give you a tangent, a direction, so that you tend to be wrong by fairly little, then it is probably useful. And that is exactly what an inventory forecasting model tries to do. It tries to be as little wrong as possible so you buy the right inventory, and after a while you feel, well, look, it is basically right most of the time.

[00:25:25] Mm. Then you start trusting it more and more. Then the model is meaningful even though it is wrong to some extent. Yes, of course. And that is exactly the same thing we have to accept when we build this for new customer acquisition: we know it will never be 100% accurate. We know we can use MMM to inform, or, the allocation of spending between Google and Meta.

[00:25:46] Yes, we know we have historical data. We know we can bring in trend analyses. We know we can look at inventory. There are loads of things. In our first version, we will not include anything about incrementality testing. You could argue that would be a possible improvement lever down the line. But then there are also challenges with company size. The bigger the company, the more data there is, the easier it tends to be. The smaller the company, the weaker an MMM becomes, for example, because there are too few data points, and then it falls apart, and the model becomes worse, and in some cases perhaps so bad it is no longer meaningful. And then it is a matter of learning when the model is strong enough because we have

[00:26:30] enough data, because we know enough, when it is not, and how we compensate when new things happen. Say you launch a new product. How do we compensate for that? Can we map it against when you launched a product before? Then you have to guess to some extent. Perhaps it is easiest to use human intelligence there and say, I think this product launch will go about as well as the product launch that happened two years ago.

[00:26:54] But then do you not have to take things into account? How large a share of the tangible market is there for that product? How... Somewhere there is a little data, and then how well does the product you are moving into usually sell online? Perhaps it is a physical retail product. I mean, we at Såpa look at this a lot, and we have launched window cleaner and limescale remover, and they are doing really well. They are fantastic products, but if we launch, say, a dishwashing liquid

[00:27:17] it does much worse. Perhaps because it is much more common to buy dishwashing liquid in retail, and it is something you always pick up. Yes. Whereas window cleaner or deck soap is something people buy infrequently. Yes. So an ad works much better. It has much more impact. So even though the market for deck soap should reasonably be smaller than dishwashing liquid

[00:27:38] it ends up selling more online. Yes. And I think this is very difficult to calculate. Yes, but it is very difficult to calculate, and there I think that, in our V1, or whatever you call it, it will really just be a matter of mapping against

[00:27:54] historical launches and saying, I think this launch will be like that launch. Yes. And you do that manually. Yes, you do that manually, qualitatively. But it is also part of the MMM model that you have to go in and specify all the discounting there has been. That is actually fairly easy to do now too with this [unclear in the source: Windsore] and Claude, because [unclear in the source: Windsore] is connected to Klaviyo. And then you can create a... I can get all the subject lines from all the old emails that have been sent to Klaviyo through [unclear in the source: Winsore] into Claude. And then, for any company, in about two minutes, I can simply say, can you go back and check all the emails that have been sent, and then I try to map out when the campaign period started and when

[00:28:31] the campaign period ended. Because as a rule, you have that in the subject lines in one way or another. Sometimes you do not, but quite often you can extract it that way.

[00:28:40] That is so good, that is spot on. So then you get a historical campaign calendar for the last 48 months, or 24 months. And then you can start with that and come to the customer and say, this is what it looks like, roughly at least, this is what we have extracted [unclear in the source: vi klar], and then you can overlay discount data from Shopify

[00:28:59] and say, here it looks like we had deep discounts, but we have no email. What happened? And then at least you can go in and ask, was there something else happening here that we are missing? Yes, we sent out one of these, but there was never an email, or whatever it was. Yes. Yes.

[00:29:11] So with those two together, at least you get a good basis for discussion with a human team who can tell you what actually happened.

[00:29:21] That is damn good. This is... I am looking forward to this tool. You seem to be at the forefront of building tools every time. But I have a few at... It is not me, I have a few people at the office who are incredibly good at this, so I am really just talking about what they are doing rather than building that much myself.

[00:29:37] Yes. No, I think it is so cool. I like Karl at Relevance. They build things like this too; they are on their 17th tool to optimize things. I have always wondered, how can an agency make so damn much profit every year, consistently, year after year?

[00:29:52] Google Ads is special, I mean. And it is because they have built so many tools that they can be so efficient. Yes. And of course, in a way, this is their cutting edge. But I like that you are now starting with this kind of tool. Because this will free up time for you, with your... In a way, your experience is important here, because you sit looking at the account and can surely model this. Yes, but it will take you one to two days to model it, show it, correlate that data, and see that, okay, this works. But if you can streamline that away, then you will be able to spend time on other things.

[00:30:35] Exactly. It is much better to spend... Because there is still a lot of problem-solving around the mathematics, around the statistics and this, but it is better to spend that time trying to structurally solve a tool that can then solve it for all companies. Now it feels a bit more like I solve it for

[00:30:50] one company at a time. Yes, and of course that is really a poor use of time. Yes. And considering what the world around us looks like and what we just talked about, it feels like, just as you finish the loop, it starts again with company one.

[00:31:05] No, but last autumn I came to the conclusion that we cannot have another autumn when... Because [unclear in the source: orange juice] is growing, so we are getting more customers. But last autumn it was very difficult to keep up with everything because we were doing too much manually, and it cannot be like that next autumn, when you have to set budgets for the following year. But this financial side is obviously one thing. The other things we are thinking about, which we have not really started at all yet, include building a strong system around creative ops in one way or another. Right now, I would say we use some AI tools, but it is more like one individual doing something

[00:31:45] very well over there, and another individual doing something well over there, rather than everyone working in the same way. And that probably means some individuals do things very smartly, but others less well. Mm.

[00:31:55] So we do not have a unified way of working, or a uniform way of working that we use overall. And of course there is loads there, everything from creative audits to creative analysis, creative strategy, what should we create going forward? Coming up with new angles, new concepts, all the way to actually creating ads with AI. Your Instagram feed is overflowing with it. So that whole area is of course huge. You could imagine building a big tool around it that can do the entire creative supply chain. And the last one you and I talked a little about before recording is probably, or we should not say last, but the last one we are thinking about in the near term, is the whole ops side.

[00:32:37] For example, sometimes our clients write copy, sometimes our clients create creative material in it. They upload ads somewhere. It has to be matched with copy. We need long names for these so we can analyze them afterward. All of this is fairly mechanical work and takes quite a lot of time, even though it is

[00:32:59] on paper, a fairly simple task. And we are now trying to work out exactly how to reduce it as much as possible, so you go from it taking, I do not know, eight hours per person per week to perhaps two hours per person per week.

[00:33:12] Much better. Um. Because it is also not work that many people find particularly fun to sit doing. No, not anymore, considering how it has developed, how the industry is developing.

[00:33:21] Mm. Some people surely think this is the most fun thing in the world to do. I know, I know that I sometimes talk to... There is such a big culture clash between, like, [unclear in the source: Stockholmina]. Even though we both live in Sweden

[00:33:34] yes, it is incredibly exciting and interesting for people to go to the factory and do the same thing they have done for many weeks, many months, perhaps many years. And in a way it makes you stop and think that I would find that fairly boring.

[00:33:51] Mm. Because I like this variety, but perhaps people think it is secure to do the same thing, and even though it is repetitive, it is work, and they see it more as work that has to be done. As long as you see that you are contributing to an important function, it is fun. And in the same way, media buying is very repetitive after a while, but there are a great many tasks that take unnecessarily long, where you just sit waiting. Really, if you are going to duplicate an ad, you have to wait 10, 15 seconds for Meta to churn through it.

[00:34:22] And you get a bit like, damn it, hurry up, you know, I just want it to happen now. But Meta is also fairly error-prone overall; there are often different kinds of errors, which is quite tiresome.

[00:34:35] Yes, too many. And this can now be automated. I know a colleague of ours built an entire [unclear in the source: cloudbot] and cloned himself and said, now I do not need to work, basically, in... Of course that is a bit of a joke, but

[00:34:46] we are not far from the point where a media buyer is not needed. We have said that before on the podcast. And now we see that, yes, the media buyer's actual role is changing. It has been doing so for a long time, but now it is developing fairly quickly, especially at agencies with 3,000, 5,000 employees. Their only job has really been the actual media buying itself. They have never been creative strategists, they have never done any heavy analysis. They have never really said how you should spend your budget; instead, they receive a budget, receive creatives, upload them and optimize, so to speak. Because what has there been to optimize in recent years? There is very little to optimize.

[00:35:26] Mm. And then they report back figures that tools today actually do much better. Absolutely. We have also built that with Claude now, actually, so I get... Admittedly, I only get Juniper in mine, but our GMs, our growth managers, now auto-generate a

[00:35:43] Slack report with the help of this [unclear in the source: Winswor], because it connects to all the interesting nodes in the system. So you get this... I mean, those Slack reports we used to write manually that took, I do not know, one or two hours to put together.

[00:35:57] Now I get one for Juniper, auto-generated at 06:00 on Monday morning, telling me about last week with a couple of traffic lights for, well, efficiency last week looked bad or good or whatever it is. So you get a weekly report that the team gets too, which is nice. And sure, it is fairly small, you could say, you can go into Lifetimely and all that, but there is some value in it arriving in Slack. Everyone reads it, everyone has it come up.

[00:36:26] You can comment, you can say something like, now we have to tackle this, whatever it is. And it takes no time, which is still incredibly nice. No, I like it. So that, and there are fewer manual errors. Even when we wrote them manually, it took time, because you could have made some... I often saw some [unclear in the source: kopid], you know, this was unreasonable, and then you go in and see, no, this was wrong

[00:36:50] a decimal point one place off, or something, a manual mistake, and then you sit there... So this is a good automation. The risk with the automatic reports, which we talk a lot about internally, is just that people do not read them

[00:37:05] or do not reflect on them. Perhaps you skim more, you simply assume everything is right, and you do not really reflect on, well, what should we do then, or

[00:37:15] so you need to add qualitative reasoning yourself, yes, to actually take in the information, which is still a risk factor with too much automation. Yes. No, I agree. So a summary should still be written by the growth manager in the thread, where perhaps they assign action points based on last week.

[00:37:34] Exactly. Exactly. That is the sensible approach. First of all, a big thank-you for listening to the podcast, which is produced and made possible entirely by my own agency, Porobic Group. One of Sweden's leading fully dedicated e-commerce agencies, working with profitable growth throughout the chain. And if you, like many, many other e-commerce entrepreneurs, have got stuck, plateaued, or just need an active investment from someone like me and want to make the same journey we have made with so many brands from n to 100, do not hesitate to contact me directly on LinkedIn and we will take it from there. And as usual, you are always welcome with questions and challenges. You know I like that. And one last thing: if you have not done it already, make sure to subscribe to this podcast, because that enables us to stay motivated and keep delivering at a really high level

[00:38:31] going forward. Thank you. Thank you very much. Okay, we touched a little on Juniper. You must... March is over, ish. Yes. So we need to know how things have gone now that you have properly restocked and filled out the range. You must have grown something like 50% year over year. Well, we are up 55% year over year to date.

[00:38:53] Damn, what a good guess. Year to date, so March must have gone even better, then. March went great. Yay. Because we were actually down year over year on March 1, I think, or at least something like that. I think we got the bed linen

[00:39:05] I do not know exactly, February 28, or no, perhaps not the 28th, but say February 25. So we had two or three days of bed-linen sales. It is possible that we... I think we were actually down when February closed, and yet now we are up 55%.

[00:39:19] So it made up for a really bad January and February. Exactly. Or we were basically neutral after January and February, perhaps slightly down, and then March went great. And that is

[00:39:31] but then that is triple-digit growth in March. Yes, absolutely. It probably is. You would think I ought to know this. But the report has been generated, and you have not read it.

[00:39:40] No, exactly. [unclear in the source: inte skicka ivägen]. No, but we are up 156% in March compared with last year, so March has been nice, and now you just hope you can carry that momentum into April and May. We were actually supposed to launch a new product line now, at the beginning of April. Instead it will now be the end of April, which

[00:40:00] which, yes, perhaps, but that gives you time to ride out this wave. But still a bit of a shame, because you would have wanted something new now as April starts, so it keeps picking up speed.

[00:40:12] Yes, perhaps. But I think it will pick up speed in May instead, which would otherwise be a natural dip, because it takes about two months now that you have good ads. They will naturally get good traction. April will go fantastically well.

[00:40:24] Let's... I mean, on the back of March going well, April will go really damn well, and then somewhere toward the end of April things will start slowing down, and you come in with new products in May, at the end of April. Perfect timing.

[00:40:36] Yes, but I think it will probably be good. So it feels very promising. Despite things being a bit noisy in the world, at least for the moment it feels like Junipers is ticking along very well, and it would have been... I mean, we only grew 10% last year, and of course it would be quite fun to post somewhat higher growth figures this year

[00:40:56] and take Juniper up to being a bit more of a real company. It is still just a little too small. But 10% is nice considering everything that happened last year.

[00:41:05] Yes. No, but it was completely fine. We were still happy that we grew at all last year. And it is incredibly much more fun to grow 10% than shrink 10%. So if we

[00:41:16] but if we could grow, say, 30, 40% this year, of course that would be great fun. And 10%. But you added B2B, and that has gone quite okay, has it not?

[00:41:26] Mm. B2B is doing well, but it is a small part, though it is doing well. Okay. Is that included in the total 10%, or was it 10% online?

[00:41:36] No, we grew a little online, but 10% overall. Okay, got it. So this year will be online's comeback. Exactly, this year hopefully online will grow quite well.

[00:41:49] Yes. Because we should say that this 156% is online, not the total. Yes. No, the figures must be fantastic this year, also considering how the fragrance products are doing.

[00:42:01] Exactly. But it is still a small part of the total. I know. But a lot of new customers are coming in, as we have talked about. I think that will help you a great, great deal. Much more than you think, because more people become aware of the brand, and it is five times easier to dig where you stand, as I usually say.

[00:42:16] Absolutely. Five times easier. That means it should reasonably be five times cheaper to acquire a customer who has bought a fragrance product and convert them to bed linen. But I agree, the step from fragrance to bedding is a big one, but it is still easier than going from nothing to bed linen. Plus, it is good that our... With the bed linen, it is very anonymous. You do not see that you sleep in it, more or less, even though you might think about it, but you do not see it. But at least the fragrance products are standing in [unclear in the source: ungans], and that also means you become part of people's

[00:42:44] well, I do not know, conscious everyday life in some way. Which I imagine is a positive thing, at least. I think it is a very positive thing. Plus, they look good.

[00:42:53] I think so too. So you want to have them out on display. Exactly. Yes, so we will see. It will probably be good, and I think most of the companies are doing well. I am just trying to think through whether we have any customers now who struggled at the beginning of the year. But I do not really think so; most customers are doing quite well compared with expectations when the year began.

[00:43:16] Compared with expectations is still an important part of that. Was there any customer whose expectation was that they would decline? No, there was not that either. No.

[00:43:25] Actually, wait. Yes, I have one customer whose expectation is down from last year. Wow. And so far that is correct. They are in line with expectations, but down from last year.

[00:43:38] Yes, I see. But there are reasons for that. Yes, there are other internal problems. Yes. So it does not have to be the macroeconomy or anything else. No, exactly. It would have been basically impossible to match last year.

[00:43:51] Okay. Fine. So things are still going okay. Yes, I mean, this could be due to a million different things. It could be anything from running loads of promotions and stopping promotions, building the brand, or something else.

[00:44:01] Yes, whatever. I will leave it anonymous. We will leave it anonymous. In other words, let us talk about someone else who has done really well. So there are all sorts of things, and you are not satisfied. And how are Såpa and Mockberg doing?

[00:44:16] Såpa is finally doing much better. Q1 was what it was, January was a little tough because we were up against very strong numbers, very good numbers from the previous year. Then a lot has happened on Meta. Big changes. We also got a bug in the checkout that we did not notice in January, strangely enough. It sounds very odd, but all the KPIs were very similar, yet we were losing customers coming from Facebook ads because they could not choose invoice payment in the checkout. That solution was bugged, which made it very difficult to discover, really, but sometime in February they discovered it. They fixed it fairly quickly. I think you had the same issue, but I do not know, we have not even talked about this. I do not think you noticed, because you have the same solution we have, do you not?

[00:45:11] Exactly. When we got... We had only received about two customer inquiries about it, so perhaps we did not really think it was such a structural problem. Yes, it was very structural, because when we went in and looked at it, there were several hundred customers. I mean, it affected somewhere up toward 3,000-something customers

[00:45:27] which of course, you catch some customers because they choose Swish or choose a card, but those who really want an invoice will drop out. Yes, there is a risk of that. Absolutely. Unless they switch to desktop or

[00:45:41] exactly, unless they switch to desktop and really want to buy the bed linen, really, hardcore want it. But at least we saw, since these are relatively cheap products, and of course we have a unique brand, but you can still buy window cleaner or limescale remover or deck soap from another brand.

[00:46:00] Mm. So we saw there was some drop-off, and as soon as that bug was fixed, we saw year-over-year growth the very next day.

[00:46:12] Mhm. And it was strong year-over-year growth. So it was a... Yes, but in retrospect you are like, okay, how could we miss this bug?

[00:46:24] Mm. But I basically only look forward, and then we look at how to make sure it does not happen again, because it is very difficult to see anything in the KPIs that was remarkably different. Because rationally, you ought to have seen that, okay, the drop-off rate in the last steps should have increased much more in percentage terms or something. But because it is such a unique situation, and on top of that it was only people coming from iOS.

[00:46:56] Mm. Via Facebook ads, clicking that they want something, going to checkout, and wanting to buy with an invoice. There are very many... Exactly. It is a Venn diagram with a lot of overlaps, so to speak.

[00:47:08] Yes, a lot of overlaps. So it took a while before we discovered it, and then, yes, there were not that many who contacted us; there was a bunch who contacted us. Absolutely. But then we thought, we tried to replicate the problem

[00:47:21] and since you do not go via Facebook ads, because you just go straight to the site, add something, and see, wait a moment, this works, and then we do this from ten different directions, ten different devices, and see that everything works

[00:47:32] and then it is probably just a problem for one individual, and then another individual does it, and you are like, wait a moment, perhaps it is still just two individuals who have this issue.

[00:47:39] Yes, so you continue, but when about ten individuals have the same problem and report it, then the unreported number is usually quite large as well. Also because if you have taken the time to report it, then you usually see that the unreported number is 10 to 12, or 10 to 20 times larger than those who actually took the time.

[00:47:58] So if there are 10 reports, there could be 100 to 200 who have actually had the problem, and that could... Absolutely. Yes. No, but one report people generally perhaps look at a little too little, but which is very good in Shopify, is this funnel. They have a funnel of sessions that is visualized quite nicely, showing how many added to cart, how many went from the cart to checkout, how many went from checkout to purchase.

[00:48:26] Yes. Then perhaps it is not very easy to find anyway, but at least if you had looked at that year over year or something, perhaps you would still have seen that, well, sessions look good, this looks good, but we are losing a lot in the last step. Then perhaps you might have

[00:48:43] but the problem for us was that we looked at that retrospectively too, of course, but then you also saw that, okay, in January the payment provider had a campaign in 2025 where people could buy now and pay much later

[00:48:57] which drove up the number choosing that invoice solution. I see. And when you looked now, you could not see, okay, is that percentage dropping off because they do not have a campaign this year? And then a lot has happened. We optimized the checkout a great deal during 2025

[00:49:17] so it converts very much better, so far fewer people normally drop off. So looking year over year does not work. Looking period against period, Q1 against Q4, does not work either, for obvious reasons, because you have far more returning customers. You have campaigns once a year, at least at Såpa. So that failed as well, and if you compare week against week, that fails too, because we have different seasons when we start selling deck soap versus not deck soap.

[00:49:46] Mm. So, yes. However you looked at the statistics, it was difficult to read it out, difficult to read it out, and yes, difficult to fix going forward. We have concluded that this will be a problem where we simply need a manual checkbox: okay, look at the checkout, go through it, look at customer service and things like that, are there problems in the checkout

[00:50:13] that we have not discovered yet? And as long as you have third-party solutions, we should say, whatever the third-party solution with Shopify, you run a greater risk than if you are 100% native. But at the same time, you have to pay for the native issue, because Shopify wants a lot of money for its native solution.

[00:50:33] Yes. Yes. So there are pros and cons to everything. That is clear. Yes. Like everything else in life, it feels like there are pros and cons. Yes. So you have to choose a path and then make sure you try to plug those holes.

[00:50:48] Exactly. No, but we have, I have added this to our Slack reports, so you get these four metrics from sessions downward. But just as you say, it is often quite difficult to interpret anyway, because week over week there is often an 80% difference in some form in one of these metrics

[00:51:07] and then you just have to... Up 80%. Yes, right. We had a discount last week that we do not have this week, and then it becomes... So there are very big swings week over week. Year over year, similar kinds of things can happen. Yes. Which also makes it a little difficult, so you

[00:51:21] then the platform also changes how it calculates things. For example, Meta had a click-through ratio which, when they measured it last year, year over year, meant that as soon as you clicked on an ad, it counted as, well, now the click has gone through. Now it is when the site has loaded

[00:51:41] or actually first page load, or whatever it is called. Yes, but it is link... Yes, exactly. Now only link clicks count as click-throughs. Exactly. That is changing today, March 13.

[00:51:51] So things like this happen in the platforms, and sometime in July or August, Shopify also changed how it counts sessions. Yes, a marginal change. And then it also changes in the sense that, okay, ChatGPT gets a lot of search volume, but you do not go on to sites to the same extent as you did through Google. That changes the year-over-year data process, so you get higher-quality traffic today than you did a year ago, but

[00:52:23] yes, for other reasons you get fewer sessions. So a lot happens when you look at reports over time. A lot happens when you look week over week or period against period. So that gut feeling we talked about at the beginning of the episode really comes into its own when you are interpreting data.

[00:52:43] Absolutely. No, but yes, absolutely, to some extent you really need that, and it is not always easy. And the person interpreting the data is not always also the person handling customer service, for example.

[00:52:53] Exactly. And perhaps you do not have full insight into that, and that also creates delays in everyday work. Mm. It feels like that will come too. We are going to test a new customer-service system for Juniper now. We use Gorgias today.

[00:53:07] Mm. What are we going to test? Something called Mimir. Oh, I have not heard of that. Norwegian. Quite small, newly started. Judging by that expression, you do not sound overly enthusiastic.

[00:53:17] No, but we are going to test it because I think it is good. Okay. But it is very AI-first based. And see whether you can get better leverage on the hours you put into customer service by automating more things and so on, and of course the hope is that it will work. And generally they also have the advantage, because it is built in Norway, of having much better connections into the Scandinavian ecosystem. For example, they have a connection to Ongoing, they have connections to many of the systems you are exposed to anyway, which makes it a little easier, whereas when you use, like

[00:53:56] I mean, I know Richpanel, I have never actually tested Richpanel, but Richpanel is an AI-first customer-service system built in the US. They will not have a connection to Ongoing, because nobody uses Ongoing in the US.

[00:54:06] No. So there can also be advantages in it being built a little closer by, yes, in Scandinavia. What will it cost, or is it the same price as something like Freshdesk or Gorgias, or is it 2x, 3x?

[00:54:20] It is about the same price, but then they add a variable component for all customer-service cases that AI solves from beginning to end. It still feels reasonable that

[00:54:37] mm, they get paid for that kind of case. Yes. I am not saying it is wrong. It is just a new part of the payment model. Interesting, where of course the argument is that if AI manages to solve a customer case from beginning to end, the customer is satisfied, and you have to pay a few kronor for it, of course that is a huge saving compared with a human doing that customer case.

[00:55:00] One hundred percent. But then it also has to be at the same level or better. Exactly. Exactly. And that is the other aspect, obviously. But at least there they have a mechanism: they know that if it does not maintain a good standard, people will not continue.

[00:55:13] Mm. True. Normally, these apps most often charge for sales. So if they make an additional sale or an upsell, they take a commission on the sale amount itself. Yes.

[00:55:25] Perhaps not customer-service apps, then. Some do. [unclear in the source: Jag har sett har berättat in på], but other apps in Shopify want to be paid for added value, in a way. Yes. And that can quickly get very expensive. But these instead only charge for having solved a case. Can you then configure it so that the case is solved in a particular, specific way?

[00:55:46] I assume so. But pass. I have not... We are not here, we are not live yet, and we have not even had the first onboarding. Because I am thinking of Twistboxes, where we work a lot with AI. We even have an AI person on the shareholder list, and

[00:56:01] who has worked with AI for a very long time. I was about to say, is it actually an AI on the shareholder list, or...? No, it is really an AI developer who has developed many different things for several different major companies. And because we have that, we have had, you know, the social media person has been an AI agent for over a year now, and it works quite well, and we have had AI-based customer service, and

[00:56:24] what we are tweaking in customer service is that AI should answer so much, I mean, it should help the customer toward a purchase, so it becomes more of a guide, just as a human would have been. Because we see that the questions coming in perhaps are not problems, but usually, how should I think about this, does it work on my car, which cars, and there are a thousand different

[00:56:46] models and adaptations and whatever. Still quite well suited to AI. Quite well suited to AI. Absolutely. So in a way he has now built database models for this, so we will see how it works. But at least that is what you would want to know with Mimir too, that you can configure it to, I want you to sell products in customer... So it becomes more of a sales channel.

[00:57:08] No, absolutely. For example, we have two different weaves of bed linen, and you think like this, so... No, but that is absolutely the hope. And another thing I know they do not do, but which I said they ought to do, is that at the end of every month you would simply want: these were the three most common questions, or the most common question when someone was on this product page, most people asked about this, so you can start covering that gap proactively instead of it happening through customer service. Or, for that matter, post-purchase questions saying, I bought last week and this has happened

[00:57:40] or this is how it is, and then you find out about it, you get a feedback loop. That would be gold, so you can gradually start closing things on the website.

[00:57:48] Yes. Yes. And, or in ads, for example, or... Yes. Because often, when people have those questions repeatedly, you can close them. Mm. Yes. But they are excited about the idea, though I do not think they have developed it.

[00:57:59] No. So, but we will see. If you want to try Mimir, you can write to me and I can tell you how good it has been.

[00:58:07] We will see. I will want to test it. Yes, we will see. We will see. We will want to test it with some company. Yes, I love it when there are innovative companies that want to change how something has worked for a very long time. Now all the companies I have tried have worked well, I think. You have Freshdesk or Freshworks or whatever, they have changed names several times, and then you have Gorgias, and then there has been another one, live chat, what is it called now?

[00:58:33] There is Re:amaze. There are any number of these. Zendesk. Exactly. Yes. And all of them have worked basically equally well. I know. But they are all still a little too much legacy. I mean, if you go to Gorgias's website, the graphics make it feel like an AI company. They say this is AI-first customer service, but it really fucking is not in the end, anyway.

[00:58:56] You feel that the product was built at a time when AI was not at the forefront, and then of course they have tried to integrate loads of AI, but it is still a little slow-moving. And I think that will change now, that there will be a shift as we move toward much heavier support from AI in customer service overall, and I think it will both improve quality and increase speed, it will be better in every way. So that is why I want to test some of the tools popping up now.

[00:59:28] And it is not really that strange, considering they have thousands of customers, perhaps tens of thousands, hundreds of thousands, and if they implemented and made a big change, people would get irritated, because we humans are creatures of habit. Especially in customer service, where they are not normally at the forefront of AI. Who works there? Who do you put in customer service?

[00:59:50] I would rather put our best people there, because it should sell, and they should be very kind people who are good at customer service. But I know a lot of companies outsource their customer service, often offshore somewhere. Usually nowadays with chat solutions, and perhaps they are not the sharpest knives in the drawer. So that is a shame, because I think customer service is one of the most important functions in e-commerce. But if you think that way, they do not want change. And if Gorgias made such a big change and implemented AI-first, even if it were possible, I think many people would rebel a little against it

[01:00:26] probably because they are used to how they wanted it, and then they would almost have needed two platforms. But surely the goal for most companies, at least if you are growing less than 100% a year,

[01:00:40] which I still think covers almost all companies, then in my head it feels like the goal ought to be that you should not have to put more human hours into customer service over the coming years. Obviously it is one thing if you are starting from zero million in revenue and growing; yes, then you probably need to increase. But if you are at 10, 20 million kronor in revenue and you double to 40, then the goal has to be that you should not put in a single extra human hour to cover that gap without losing quality.

[01:01:14] If you go from 40 to 80, the same thing. And it has to be, or obviously it is AI that should cover that gap. Same quality, the same number of human hours. And that ought to be the goal for most companies.

[01:01:26] I agree. In a way, you ought to measure this too, so you can see, okay, how much time have we spent on different stages? And it is not just about having different people doing different things; you can also measure it in terms of how many cases we have solved, what it looks like, and so on and so on. Of course people also become more efficient, so it is

[01:01:46] a little more difficult to interpret than just looking at Excel. You also have to dig a little and understand, okay, have we replaced anyone, has one of the people become more senior, and so on, from very junior and become better, and so on and so on. So there are a few parameters to take into account, but it is not rocket science.

[01:02:06] No, I think it should be possible, and I hope it already is now, so we will see. Okay. We will see. Perhaps we will come back to the question of Mimir. We will see. As I said, we have also been doing well. Very well, actually, strongly with the Mockberg case. Mm.

[01:02:22] We have closed the books for 25, and we are aiming, it feels like we are aiming for the stars in 26 as well, so things should go very well. And above all it is the international markets we are working very hard on and have succeeded in, and I see more Swedish companies, especially in DC, succeeding internationally. I read a lot about ESSNCE now, which won an award at

[01:02:49] D-Congress a month ago. Yes, and where previously we were nominated with Mockberg, a runner-up, you could say. Yes, still great fun. And they have also succeeded very strongly internationally with their perfumes.

[01:03:04] Yes, not fundamentally surprising, but I know they struggled a great deal with Norway before it finally took off. Mm. From one day to the next, basically.

[01:03:16] But yes. No, internationalization really is everyone's focus, or however you put it, on everyone's agenda, you could say. I think it is almost never that I have a discussion with a company turning over more than 25 million, or over 40 million at least, where internationalization is not a top-three priority.

[01:03:35] Yes. And I would still say it is complex in the sense that translating the website still does not work flawlessly. Almost flawlessly. Yes, but we are well on the way. But there is still a little way to go. Translating ads probably works almost flawlessly in practice, or you have it in theory, but in practice it is still difficult to fully find scale in volume and just manage to translate all the ads easily and quickly, both voice and text, and the actual operations around this are still complex, even though if you are only talking about one ad, it works brilliantly. But getting scale in it is fairly difficult. But these things, both of these things, will be solved. Internationalization help

[01:04:23] with localization exists to some extent too; we have talked about it before here on the podcast. So it feels like Europe as a market, at least, which has been one of the hardest markets in the world if you see it as one big market to tackle... I mean, very few American companies have succeeded in Europe, partly because

[01:04:42] there are different cultures in every country, different languages in every country. It is very fiddly, whereas currencies, currency factor, whereas in the US a great many companies have obviously managed to go from state to state as if there were basically no borders at all.

[01:04:57] And hopefully over the next one, two, three years, very much supported by AI, the borders will be erased in Europe too. And internationalization will be much easier. I agree. It feels that way, at least. And partly that, and partly there are more apps tackling different areas that make it easier

[01:05:17] so there have been logistical challenges, because we, so we have payment challenges because you have had things like cash on delivery in some countries, which is still relevant, but the apps tackle those problems in a neater and better way.

[01:05:30] Yes. And it is coming on a broad front, in the same way Shopify democratized e-commerce, in a way. So it is very easy to set up a store today. Still very difficult to succeed after scaling it to a few million. I think a great many people can scale their own store to a couple of million fairly easily, because, you know, you tap into this lowest funnel, the bottom clicks, easily with something like Google Shopping and last-click attribution. Very easy. The first ad you upload on Meta, provided it is reasonably okay, you have reasonable pricing,

[01:06:07] usually generates some sales, and you are like, yes, and you think, I am so good at e-commerce. So the first two, three, five, 10 million can be relatively easy. It has been democratized. It was much harder 10, 15 years ago, when you needed to invest a lot in the technology.

[01:06:23] The difference now is that it is more expensive on the marketing side. It is more expensive. Absolutely. They do not cancel each other out. It is still relatively easy to buy the last click, because when people have an interest, and as long as you have a good product-market fit, you will get some velocity.

[01:06:38] It is much easier. On the other hand, it gets harder to scale after that level, and that is where people most often have this challenge of, okay, how do we go from here, now that I have reached product-market fit but perhaps stagnated at 8 million for three years in a row? Mm.

[01:06:54] What should I do? Because here I actually need to know... You need a certain level of operational excellence, or whatever you want to call it, marketing excellence. That is also where I think this comes in even more, that marketing has become structurally more expensive. So it is harder to scale the marketing and continue [unclear in the source: sexäla] profitability.

[01:07:12] Yes. And once you crack that nut, you get to the next level, and then you need to crack that nut again. So there are a few levels that become a dealbreaker. Now with this internationalization, since it is being democratized from several directions, we have an app we use for translations that has now integrated with ChatGPT, this Weglot

[01:07:34] on Shopify, which has actually become good. Weglot has an extra payment option with ChatGPT because they need to pay for tokens, and then the site becomes very good. But of course you might want to double-check your terms and conditions, policies and so on, so you do not break a local law in a local country.

[01:07:52] Yes. Yes, of course. So you do not have something wrong there. But otherwise, that takes a day, and then you have done it reasonably well, say 99.9%. I think even if you had outsourced this to a freelancer in the country in question, you would also have those same few errors then.

[01:08:11] Yes. Yes. Because that person does not know your brand, that person does not know the background, the problems you have had, and does not have this experience, the gut feeling you have. So of course, if you translate something from Swedish to English, you will be able to translate the spirit of your own copy.

[01:08:30] Yes. AI will be able to do it at least as well as a freelancer you would normally have paid something like $25 an hour or 30 and accepted that it was correct

[01:08:41] and accepted that it is correct. And of course that last one per mille, that 0.1%, you only get by, say, hiring a head of Germany for however many tens of thousands of dollars, or euros for that matter, a month, who then hires a team or a local agency to interpret the ads and completely redo them

[01:09:05] yes, for the German market. And if you are not at two, three, 400 million in revenue, it is very difficult to get there. Yes. But that said, it is becoming much easier, the bottom line is that selling in other European countries is being democratized with the progress of these new apps.

[01:09:25] Yes. One hundred percent. No, but it is a very enjoyable development. It will hopefully mean more Swedish companies can make it out. We are usually quite good at branding and positioning and so on. So hopefully it means we can get out into the world more.

[01:09:44] And perhaps, for that matter, we will see more international brands in Sweden too. But in a way I do think so. There are a great many who want to sell to the Swedish market. Partly because we have had [unclear in the source: höghandels], or good GDP growth. Good, even though in recent years it has been worse compared with many European countries. But historically it has been very strong. So we are a nation with quite a lot of purchasing power, and if you think about all 200 countries, we are always in the top tier. If we fall on that list too,

[01:10:14] then you have to do the calculations. No, but economic development has been worse than in many other countries, especially in percentage terms, and it does not look set to change in the next few years, but depression

[01:10:30] we have a good tech and startup ecosystem and so on. I think there is quite a lot to be hopeful about. They did some study the other day, or that I read the other day at least, from... I think it was one of the founders of Lovable who shared it and said, well, Silicon Valhalla has the second-most unicorn startups in the entire world per capita, beaten only by Tel Aviv.

[01:10:54] Okay. Which right now is obviously, yes, in turmoil. In turmoil, to say the least. It is a... Yes. But in that sense, yes, because it is actually damn good, but there are not that many e-commerce companies on that list, pure e-commerce companies; usually they are AI-driven, tech-heavy companies that have disrupted a market. Another thing on the internationalization side was that Google actually launched Google Shopping in another 15 markets in Europe.

[01:11:25] Yes, I do not know, there is not that much to talk about, but still interesting. It is interesting that it has taken so long. I have to admit I did not know about this, so the news for me was not... I do not know, it was not news I was sitting waiting for, but it still stuck.

[01:11:36] I know. Yes, I knew about this. I just did not understand why they did not have it, because the technology has obviously been there, and adoption, people's behavior in those countries, has been there. So I have always questioned, for example, why does it not work, why is there no Google Shopping in Serbia? Yes,

[01:11:54] now I have many brands that do sell in Serbia, because I obviously come from Serbia, so it is still a market that... Serbia, Croatia is in the EU. Serbia is much harder because it is outside the EU, but it is still an interesting market. Even though there is a certain culture clash, I can bridge it naturally.

[01:12:13] Yes. Anyway, now Google Shopping is there. I think, or I say think, I know this will boost e-commerce sales in a great many countries in parallel now. Fifteen countries getting access to perhaps Google's most powerful tool. If you think about it like this,

[01:12:31] I would probably say, if you think about how large a share of e-commerce sales through Google comes from Google Shopping, it is usually somewhere between 80 and 90%. I would say it is probably actually their most powerful tool. It sounds a little boring in a way. They would probably have liked you to

[01:12:47] answer PMax or something. But that is not really how it is in practice. Or something like YouTube, or somewhere where... YouTube is a different world, just not for e-commerce; YouTube is not really as big there.

[01:12:58] No. In e-commerce, Google Shopping is enormous, I mean, it has gained huge adoption across practically the entire Western world where Google Shopping has existed for a very long time now. And as you say, I have been a little surprised that there were so many countries where it was not active.

[01:13:19] Yes, I knew about some, but okay. But not others. No, but because... Yes, so it is nice. Nice addition. Do you think people should take action if they have been present in some countries, just set this up ASAP? It is an action that is fairly

[01:13:30] yes, it is a very clear action. Especially if you sell in Europe and have something like rest of world that you measure, and have certain countries that are

[01:13:41] countries in your calculations, then you should be thinking, okay, how do we set this up and localize it so we make sure it looks good in the different countries? Exactly. Exactly. Correct.

[01:13:54] I mean, it really is such a simple quick fix, and you will get a substantial boost in some countries, especially over time as people understand how it works in those countries, when they start clicking through and seeing, yes, how nice.

[01:14:06] But with that said, perhaps it is starting to be time to wrap up. We are wrapping up this week's episode, and this is a special episode. Do you want to explain why? Yes, this will be the last episode that at least I am part of on 0 to 100. We recorded, or released, the 100th episode last week, and this is the 101st, potentially perhaps a fresh start for something new. But we will see whether you choose to keep running the podcast. But I think you have a few episodes in the pipeline ahead, at least.

[01:14:41] And then we will see where it goes in the longer term. There are a few episodes I have committed to. Then, as you say, we will see; I have no new concept. Jacob leaves big shoes to fill. One hundred and one episodes together is quite a lot. We have been podcasting since 2023, right, late autumn 2023.

[01:15:02] So it has been a kind of Monday therapy for me, coming in here on Monday. We get things off our chests, and what has happened in the world? How are the brands doing? So it has still been a nice Monday podcast session. I have even booked those where we have had... This morning I had a board meeting with Twistboxes. I was like, no, I have the podcast with Jacob. This takes priority.

[01:15:27] No, but I agree. This journey has been very exciting, and I do not know, I felt it was time, time to do something else or something new, and it takes a fair amount of time. And perhaps, we will see. I actually like podcasting. Perhaps there will be another podcast in the longer term, but here and now it felt like the right time to pause. After all, we have talked, and I have been the guest in I do not know how many of these 101 episodes, perhaps 93.

[01:15:54] Yes. So it feels like, to some extent, you have said what you want to say. Yes, very much so, you are right on the substance, and I have also been thinking about it for several months. For one thing, I have been pushing all of you listening to go in and review, you know, give us more feedback, comment, so we get more energy too. And of course we see the listener numbers increasing, but that is not the whole thing. You want to see engagement increasing, and at least I am very driven by emotion around... I do not make any, we do not make any money from the podcast. We pay money for the podcast, for cutting, editing. We have to sit here, you know, equipment, studio, all sorts of things. So we pay money to share our best tips, everything that is happening in

[01:16:42] the world around us, and do a lot of work on it. And because of that, I have also been thinking about this for several months now. But in a way I like the tradition it has been. So now, as we turn a page in the podcast's history, we will see. There are a few more episodes coming that are already planned, obviously, and I am committed to those. So you will hear 0 to 100 with a few guests coming up, and then we will see how well, or how we take this forward. So if you listening now have any tips, now really is do or die, or now or never. Send them in. Send feedback. Comment, would you like a new person sharing their journeys from zero to a hundred? That person needs to be here every Monday.

[01:17:34] I am set on recordings being on Mondays, publication on Thursdays, because it feels like that gives our editors enough time to fix everything. Yes. So a big thank-you to Jacob Wibom Westerberg, who has been part of 100 episodes.

[01:17:56] A big thank-you. A big thank-you, and hats off. Thank you very much. The physical hat off, and we will see each other on the other side. Yes, indeed. Yes, we will. And thanks to everyone who has listened. Fedja will be back next week with a new episode.

[01:18:13] Cool. See you next week. Bye. Thanks for listening. We will be back next time to tell you our thoughts on what is happening in the landscape and share our latest tips for growing your company and becoming more profitable.

[01:18:28] And if you like this episode, share it with a friend, and please give the podcast five stars and follow us wherever you listen to podcasts so you do not miss the next episode.